Article

Differentiation fate of a stem-like CD4 T cell

controls immunity to cancer

https://doi.org/10.1038/s41586-024-08076-7 Maria A. Cardenas1, Nataliya Prokhnevska1, Ewelina Sobierajska1, Petra Gregorova1,

Christopher B. Medina2, Rajesh M. Valanparambil2, Rachel Greenwald1, Luke DelBalzo1,
Received: 4 December 2023
Mehmet Asim Bilen2,3, Shreyes Joshi1,4, Vikram Naryan1,4, Viraj A. Master1,4, Martin G. Sanda1,4
Accepted: 18 September 2024 & Haydn T. Kissick1,2,4,5 ✉

Published online: xx xx xxxx

Check for updates The T cell response to cancer controls disease progression and response to
immunotherapy1–3. Despite extensive knowledge regarding CD8 T cells, how CD4
T cells contribute to this process is less well understood. Here we identified a
population of PD1+TCF1+ CD4 T cells with stem-like properties that are capable of
self-renewal and differentiation into canonical CD4 effector cells. Primarily residing
in tumour-draining lymph nodes (TDLNs), these tumour-specific CD4 T cells are
restricted by T regulatory (Treg) cells to a stem-like fate that predominantly generated
induced Treg (iTreg) cells, limiting effector CD8 T cell responses to the tumour. By
contrast, upon Treg depletion, stem-like CD4 T cells differentiated into T helper 1 (TH1)
cells, and via IFNγ production induced robust effector differentiation from TCF1+ CD8
T cells in TDLNs, a state we defined as ‘active’. Notably, enforcing TBET expression in
transferred stem-like CD4 T cells was sufficient to overcome the established restricted
T cell state. Despite the presence of Treg cells, endogenous stem-like CD4 T cells
actively generated TH1 cells, which were required to restore TDLN effector CD8 T cell
differentiation, enhance tumour control and rescue response to immunotherapy.
In agreement, TH1 differentiation in patients with kidney cancer predicted successful
immunotherapy responses and improved progression-free survival. Together,
these findings identify a stem-like CD4 T cell population that through alternative
differentiation fates controls the switch between restricted and active T cell states
with implications for cancer immunotherapies.

CD4 T cells are associated with successful immunotherapy responses, T cells infiltrating human kidney tumours. We identified three clus-
adoptive T cell therapies, and vaccination strategies across cancer ters that expressed genes related to antigenic stimulation and tissue
types4,5. The expansion of conventional CD4 T cells following check- migration (Fig. 1a and Extended Data Fig. 1a,b). Cluster 1 cells had a Treg
point therapy, or their presence in tertiary lymphoid structures often phenotype (expressing FOXP3 and IL2RA) and were significantly
predicts favourable clinical outcomes4,6–9. Moreover, the differentiation enriched for a Treg signature (Fig. 1b and Extended Data Fig. 1b,c).
of CD4 T cells into TH1 or T follicular helper (TFH)-like subsets appears Cluster 2 cells expressed EOMES and various cytotoxic molecules,
to be critical for effective tumour control10–14. However, CD4 T cells can with minimal TBX21 (encoding TBET) expression or enrichment for a
also be found as Treg cells or dysfunctional phenotypes, and these states TH1 signature (Fig. 1b and Extended Data Fig. 1b,c). Notably, cluster 3
are frequently associated with disease progression5,15,16. Thus, although cells did not express any of the major lineage-defining transcription fac-
activated CD4 T cells are a significant component of the cancer immune tors (BCL6, TBX21, GATA3 or RORC). These CD4 T cells instead expressed
response, the mechanisms that determine whether CD4 T cell subsets high levels of TCF7 (Fig. 1b), genes related to the stem cell programme,
will enhance anti-tumour immunity or contribute to tumour progres- and low expression of genes encoding inhibitory receptors and effec-
sion remain poorly understood. Here we examine the differentiation tor molecules (Extended Data Fig. 1b). Furthermore, TCF7-expressing
fate of tumour-specific CD4 T cells to better understand their functional CD4 T cells were significantly enriched for a human stem-like CD8 T cell
roles in the anti-tumour response. gene signature17 (Fig. 1c) and had the lowest enrichment score for cell
proliferation and CD4 helper signatures, suggesting a more quiescent
state (Extended Data Fig. 1c and Supplementary Table 4).
PD1+TCF1+lin− CD4 T cells infiltrate tumours We extended these findings to 125 kidney, 17 bladder and 6 pros-
To examine the heterogeneity of the CD4 T cell response, we first per- tate tumours using flow cytometry. We found a wide range of CD4
formed single-cell RNA sequencing (scRNA-seq) on PD1+CD45RA− CD4 T cell infiltration between patients and cancer types, with most cells

1
Department of Urology, Emory University School of Medicine, Atlanta, GA, USA. 2Emory Vaccine Center, Emory University School of Medicine, Atlanta, GA, USA. 3Department of Hematology
and Medical Oncology, Emory University School of Medicine, Atlanta, GA, USA. 4Winship Cancer Institute of Emory University, Atlanta, GA, USA. 5Department of Microbiology and Immunology,
Emory University School of Medicine, Atlanta, GA, USA. ✉e-mail: haydn.kissick@emory.edu

Nature | www.nature.com | 1
Article
a Sorted PD1+CD45RA–CD4+ c Human stem-like
T cells from RCC tumours CD8 signature e TCF1+
<0.001 EOMES+
<0.001 0.073
d Gated on CD4 T cells FOXP3 (PE) EOMES (eFluor660) TBET (BV605) BCL6 (PE-CF594) Treg

Per cent of PD1+CD45RA– cells

1 1 83.2 0.93 TH1 (TBET+)
11.5 2.25 18.8 13.0 5.43 1.24 0.33 0.98

Tumour PD1 (UV737)

3 104 103 103 100

Transcription factor
TFH (BCL6+)

Enrichment score
4
103 103
80
UMAP 2

103 102
0 0.5 60
102 102
102
100 40
−4 102 100 100
2 0 100 100
13.1 2.78 38.9 47.4 25.9 42.4 41.2 52.1 40.5 58.2 20
−4 0 4 100 103 104 102 103 102 103 102 103 104 102 103 104 0
+

+
g
re
F1

ES
T
UMAP 1 CD45RA (BV786) TCF-1 (AF488)
TC

M
EO
0.015
b Cluster 1 Cluster 2 Cluster 3 f <0.001
<0.001
<0.001
<0.001
<0.001
<0.001 <0.001
FOXP3 EOMES TCF7 104 104 104 104

CD127 MFI

CD39 MFI

CD38 MFI
CD26 MFI
Tumour
103 103 103 103

102 102 102 102

100 103 104 100 100 103 100 100102 103 104 100 100 103 104 100
CD26 (Pecy5) CD127 (BV711) CD39 (BV421) CD38 (BV650)
0 6 12
g PD1+CD45RA–CD4+ T cells i Sorted TCF1+ (CD26+CD127+) Effectors (CD39+) k PD1+CD45RA–CD4+ T cells
CTV label and FACS CTV label and FACS
Unstimulated TH1 stimulation TH1 stimulation TBET+ (%) TBET+ (%)
TCF1+(CD26+ Treg/EOMES TCF1+(CD26+ Treg/EOMES
<0.001 0.063 <0.001 0.020
CD127+CD39–) (CD26–CD127–CD39+) 0 0.97 52.9 0.37 0 4.76 100 100 CD127+CD39–) (CD26–CD127–CD39+)
TBET (BV605)

80 80
103 103 103
60 60
5 days in culture +
40 40
102 Patient matched Co-culture 1:1
102 102 20 20
Unstimulated Bead stimulation 100 100 100 tumour DCs 5 days
0.97 98.1 44.1 2.57 0 95.2 0 0 (HLA-DR+ CD11c+) 20 U ml–1 IL-2
10–20 U ml–1 IL-2 10–20 U ml–1 IL-2
100 103 104 105 100 103 104 105 100 103 104 105 U S U S U TH1 Treg
+
Cytokines for helper CTV stim stim
differentiation condition j Unstimulated Treg stimulation Treg stimulation (%) FOXP3+ FOXP3+ (%) l 1:1 DC:TCF1+ CD4 T cells 0.006
h <0.001 0.310
20
<0.001 0 0.15 34.9 5.42 2.78 20.1 100
<0.001 0.322
100
<0.001 60.0 33.6
100

Per cent divided

100 104
0.011 80
Replication index

PD1 (UV737)
80 80 80
15
FOXP3 (PE)

60
Divided (%)

103 103 103

60 60 60 103
10 40
40 40 40
102 102 102 20
20 5 20 20 100
100 0 0
99.8 10 49.2
0
10.5 10 9.03 68.1 4.78 1.69 0
0 0 0 0

9+
100 103 104 100 102 103 104 100 102 103 104 0 103 104

F1
U S U S U S U S U TH1 Treg
+

9+

D3
F1

CTV

TC
CTV
D3

C
stim stim
TC

Fig. 1 | PD1+TCF1+ stem-like CD4 T cells are the predominant population in TCF1+lin− (n = 11) or CD39+ (n = 9) CD4 T cells after five days under unstimulated
human tumours and TDLNs. a, scRNA-seq of sorted tumour PD1+CD45RA− (U) or stimulated (S) TH0 conditions. Data are median ± 95% confidence interval;
CD4 T cells from patients with renal cell carcinoma (RCC) (n = 2). b, Normalized two-sided unpaired Mann–Whitney U test. i,j, Representative plot showing
expression of transcription factors that define each cluster. c, VISION gene CTV dilution and expression of TBET and FOXP3 transcription factors after
set enrichment analysis (GSEA) using the human cancer stem-like CD8 T cell five days in unstimulated, T H1 (i) or Treg ( j) stimulating (stim) conditions
signature. Enrichment scores are shown as violin plots; horizontal bars (TH1 stimulation, n = 18; Treg stimulation, n = 11). Data are median ± 95% confidence
show the mean; one-way analysis of variance (ANOVA) with Tukey’s multiple- interval; two-sided unpaired Mann–Whitney U test when comparing U versus S
comparison test. d–f, Representative flow cytometry phenotype (d) and within the same population; Kruskal–Wallis test with Dunn’s multiple-
frequency (e) of activated (PD1+CD45RA−) CD4 T cell populations in kidney comparison tests for comparing three groups. k, Experimental design for
tumours, and expression of selected markers (f). Data are mean ± s.d.; Kruskal– dendritic cell (DC) and PD1+ CD4 T cell co-cultures. l, Representative plot of
Wallis test with Dunn’s multiple-comparison test (n = 125 patients with kidney sorted TCF1+lin− CD4 T cells after 5 days of 1:1 co-culture with donor-matched
cancer). MFI, geometric mean fluorescence intensity. g, Experimental design dendritic cells (n = 7 patients for TCF1+lin− and n = 4 patients for CD39+). Medians
to test the proliferative and differentiation capacity of TCF1+lin− and CD39+ CD4 shown and analysed by two-sided unpaired Mann–Whitney U test.
T cells from primary tumours. h, Frequency and replication index of sorted

expressing PD1 (Fig. 1d and Extended Data Fig. 1d,e). In agreement most dominant cell in the PD1+ CD4 T cell pool in tumours and TDLNs
with the scRNA-seq analysis, around 35% of activated (PD1+CD45RA−) across various human cancer types.
CD4 T cells expressed FOXP3 or EOMES, with very few cells express-
ing TBET or BCL6 in most patients with kidney cancer (Fig. 1d,e and
Extended Data Fig. 1f). The remaining tumour CD4 T cells were nega- TCF1+lin− cells are precursors of CD4 helpers
tive for lineage-defining transcription factors, expressed TCF1 and We next tested the proliferative capacity of these cells by sorting cell
various co-stimulatory molecules and cytokine receptors, and were trace violet (CTV)-labelled TCF1+lin− or CD39+ CD4 T cells from primary
consistently the most frequent PD1+ population (Fig. 1d–f and Extended kidney tumours (Supplementary Data 1a) and cultured them in vitro
Data Fig. 1f–h). Furthermore, despite their activated state, these cells for five days under unstimulated (U) or stimulated (S) TH0 conditions
were negative for effector molecules and inhibitory receptors (Fig. 1f (Fig. 1g and Methods). Upon stimulation, TCF1+lin− CD4 T cells under-
and Extended Data Fig. 1g,h). By contrast, FOXP3+ and EOMES+ cells went extensive proliferation and retained their activated but uncom-
expressed CD39 and molecules associated with their respective phe- mitted phenotype, whereas CD39+ cells remained undivided (Fig. 1h
notypes (Fig. 1f and Extended Data Fig. 1g,h). and Extended Data Fig. 2a,b).
Finally, we examined the phenotype of CD4 T cells in non-metastatic The lack of lineage transcription factor expression, concomitant
TDLNs from 12 patients with kidney cancer. We found 20–60% of the with prior studies finding plasticity among CD4 lineages18,19, prompted
total CD4 T cell pool in a PD1+CD45RA− activated state, with TCF1+ us to investigate whether TCF1+lin− cells could differentiate to canon-
lineage-negative (lin−) CD4 T cells being the most frequent popula- ical CD4 effector programmes. PD1+TCF1+lin− or CD39+ CD4 T cells
tion (Extended Data Fig. 1i–k). These cells expressed similar patterns were sorted from human tumours as described above and cultured
of co-stimulatory molecules, cytokine receptors, and effector mol- in cytokine polarization conditions towards TH1, Treg, TFH or EOMES
ecules as the TCF1+lin− CD4 T cells found in the tumour (Extended Data lineages. TCF1+lin− CD4 T cells downregulated TCF1 and upregulated
Fig. 1j–l). Together, these data identified a TCF1+lin− population as the TBET and GZMB in response to TH1 cell polarization (Fig. 1i and Extended

2 | Nature | www.nature.com
Data Fig. 2c), whereas they upregulated FOXP3 and CD25 in response to of cells retaining a TCF1+lin− phenotype (Fig. 2a,b and Extended Data
Treg cell polarization (Fig. 1j and Extended Data Fig. 2c). TCF1+lin− CD4 Fig. 3c). The high frequency and phenotype of PD1+TCF1+lin− CD4
T cells were also capable of upregulating EOMES and BCL6-related T cells was conserved across B16F10-GP, MC38 and both subcuta-
programmes, in non-classical TH1 and TFH polarizing conditions, respec- neous and orthotopic RENCA-HA tumour models (Extended Data
tively (Extended Data Fig. 2d–f). It is worth noting that although TH1 Fig. 3d–k). Furthermore, transfer of naive CD4 T cells from SMARTA
and TFH CD4 T cells were rarely found infiltrating kidney tumours, mice—which express a TCR with specificity for LCMV GP—with a large
PD1+TCF1+lin− CD4 T cells had the capacity to acquire these programmes (500,000) or small (10,000) number of precursors resulted in a simi-
upon stimulation with the appropriate conditions. By contrast, CD39+ lar in vivo differentiation trajectory of antigen-specific CD4 T cells
CD4 T cells retained expression of their respective transcription fac- from naive to TCF1+lin− in mice inoculated with TRAMPC1-GP cells
tors and effector molecules in unstimulated conditions and across all (Extended Data Fig. 4a–i). Seven days post transfer, around 95% of
stimulated conditions (Fig. 1i,j and Extended Data Fig. 2a–f). the transferred SMARTA CD4 T cells expressed PD1 and Ki67 and were
To test whether PD1+TCF1+lin− CD4 T cells responded to physiologi- TCF1+lin− within TDLNs and tumours (Extended Data Fig. 4b,c). Over
cal levels of stimulation, we sorted and co-cultured these cells with the four- to five-week tumour response, 20% and 85% of transferred
donor-matched CD11c+ major histocompatibility complex class II SMARTA CD4 T cells expressed FOXP3 within TDLNs and tumours,
(MHCII)-expressing dendritic cells at a 1:1 ratio (Fig. 1k and Supple- respectively, with minimal expression of TBET, RORγt or BCL6
mentary Data 1a). PD1+TCF1+lin− CD4 T cells underwent extensive pro- (Extended Data Fig. 4c–i). Together, these data show the generation of
liferation, maintained CD26 expression, and downregulated markers an activated, undifferentiated population of tumour-specific TCF1+lin−
inversely associated with T cell receptor (TCR) stimulation (Fig. 1l CD4 T cells in TDLNs and tumours and suggest that over time this
and Extended Data Fig. 2g–i). Moreover, upon addition of IL-12 to the population can potentially give rise to iTreg cells across mouse cancer
co-culture, TCF1+lin− CD4 T cells upregulated TBET and acquired a TH1 models.
programme upon division (Extended Data Fig. 2j). By contrast, minimal To examine the proliferation and differentiation capacity of TCF1+lin−
to no proliferation was observed for co-cultured CD39+ effector CD4 CD4 T cells to the GP antigen in a different environment, we sorted
T cells (Extended Data Fig. 2g–i). PD1+CD39+ cells underwent prolifera- these cells from TDLNs of 5-week TRAMPC1-GP tumour-bearing mice
tion only upon administration of a high dose of IL-2 (4,000 U ml−1), a (Supplementary Data 1b) and transferred them into naive mice that
dose previously shown to induce effector CD8 T cell proliferation20 were immediately infected with LCMV Armstrong (Fig. 2d). Eight
(Extended Data Fig. 2k,l). Together, these results show that despite days after infection, CD45.1+ SMARTA TCF1+lin− CD4 T cells exhibited
existing in a quiescent state in vivo, TCF1+lin− CD4 T cells retain exten- more than 20-fold expansion in blood, lymphoid and non-lymphoid
sive proliferative and differentiation potential into various lineages of organs, with a similar distribution as the endogenous virus-specific
CD4 T cells upon stimulation, whereas CD39+ CD4 T cells are in more GP66+ CD4 T cells (Extended Data Fig. 4j). Notably, TCF1+lin− SMARTA
terminally differentiated states. T cells differentiated into TH1 and TFH cells in response to LCMV, with
The differentiation capacity of TCF1+lin− CD4 T cells led us to inves- minimal Treg differentiation, analogous to endogenous virus-specific
tigate the relationship between CD4 T cell subsets in tumours from CD4 T cells across all tissues examined (Fig. 2e,f and Extended Data
patients with kidney cancer using paired scRNA-seq and single-cell Fig. 4j,k). Thus, tumour-specific TCF1+lin− CD4 T cells have stem-like
TCR sequencing data. Among the clonally expanded TCRs for both properties with extensive capacity to proliferate and differentiate into
patients, we found shared clonal overlap between the TCF1+lin− popula- different effector lineages of CD4 T cells depending on their environ-
tion and Treg or EOMES clusters. Additionally, a small number of clones ment. On the basis of these observations, we refer to TCF1+lin− CD4
overlapped between all three CD4 subsets (purple) in both patients cells as ‘stem-like’ CD4 T cells.
(Extended Data Fig. 2m,n). These shared TCR clonotypes suggest a
lineage relationship between TCF1+lin− and the Treg cells and EOMES+
CD4 T cells in kidney tumours. Collectively, these data indicate that Treg cells inhibit stem-like differentiation to TH1 cells
PD1+TCF1+lin− CD4 T cells have functional stem-like properties and Given that stem-like CD4 T cells had the capacity to differentiate into
act as a precursor to other differentiated effector CD4 T cells within helper CD4 lineages outside the tumour environment, we next inves-
human tumours. tigated potential mechanisms that negatively regulated their differ-
entiation in cancer. First, we examined the role of PD1, given its high
expression on stem-like CD4 T cells and its inhibitory mechanisms that
TCF1+lin− CD4 T cells arise within TDLNs regulate differentiation of stem-like to effector CD8 T cells23–25. After
To address antigen specificity, differentiation kinetics and functional two weeks of PDL1 blockade therapy, both antigen-specific and bulk
relevance of TCF1+lin− CD4 T cells, we turned to mouse models. We first PD1+ CD4 T cells maintained a stem-like and Treg phenotype in TDLNs
used TRAMPC1 cells expressing the lymphocytic choriomeningitis and tumours across refractory TRAMPC1-GP and responsive MC38 and
(LCMV) glycoprotein (GP) to track the activation and differentiation of RENCA-HA models (Extended Data Fig. 4l–p), indicating that PD1 does
antigen-specific (GP66+) CD4 T cells throughout tumour progression. not regulate stem-like CD4 T cell differentiation in cancer.
Following inoculation with TRAMPC1-GP, GP66+ CD4 T cells underwent Our data show that Treg cells make up a large fraction of the PD1+
expansion in TDLNs and persisted throughout the five-week response CD4 T cell pool in TDLNs and tumours, so we next examined whether
(Fig. 2a and Extended Data Fig. 3a). Notably, the majority of GP66+ CD4 Treg cells suppress stem-like CD4 T cell differentiation to canonical
T cells acquired a TCF1+lin− phenotype in TDLNs within the first week, helper lineages. Indeed, FOXP3+ Treg cell depletion three to four weeks
suggesting CD4 T cells are rapidly skewed towards this activated state after TRAMPC1-GP tumour inoculation in DEREG mice12 (denoted
(Extended Data Fig. 3a). Five weeks after tumour inoculation, GP66+ as FOXP3-DTR) resulted in significant tumour-specific GP66+ CD4 T cell
CD4 T cells in TDLNs all expressed activation markers and primarily expansion and TH1 differentiation in TDLNs and tumours (Fig. 3a–d
exhibited a TCF1+lin−, FOXP3+ or BCL6+ phenotype, with very few cells and Extended Data Fig. 5a–c). Differentiated GP66+ TH1 CD4 T cells
expressing TBET, RORγt or EOMES (Fig. 2b and Extended Data Fig. 3b). exhibited downregulation of TCF1 and increased expression of TBET
Phenotypically, GP66+TCF1+lin− CD4 T cells expressed high levels of the and other known TH1 markers, along with increased production of
tolerance-associated markers FR4 and CD73 (refs. 15,21,22), regulators IFNγ and IL-2 (Fig. 3e and Extended Data Fig. 5a,b). This shift in CD4
of T cell differentiation, and limited expression of effector molecules T cell differentiation peaked on day 5 and was maintained until day
(Fig. 2c and Extended Data Fig. 3b). In contrast to TDLNs, GP66+ CD4 25 post depletion, despite the return of host Treg cells (Fig. 3c,d and
T cells in tumours were primarily Treg cells, with a smaller proportion Extended Data Fig. 5b). TH1 accumulation was not exclusive to the GP66

Nature | www.nature.com | 3
Article
a b Gated on GP66+CD44+ week 5 TRAMPC1-GP
Gated on CD4 T cells TCF1 Treg TFH TH1 EOMES
week 5 TRAMPC1-GP FOXP3 (PE-e610) TBET (Pecy7) BCL6 (BV421)
100 104
105 4
7.69 10 0

log10(total population)
105 17.8 0.62 0 26.8

log10(no. of GP66+ cells)

104

GP66+ CD4 T cells

0.26
104 104 80

Percentage of
104 103
103 103 60
TDLN

103

TDLN
103 102 103
40 102
101 102
102
100 100 20
100 100 100 101
4.00 70.5 22.2 77.2 17.8 55.4
10–1 0 0 1 2 3 4 5
100 103 104 100 103 104 100 103 104 100 103 104 105 Weeks after tumour inoculation

LN

r
ou
100
104
TD
70.0 0 2.00 1.00 0 0.69

log10(total population)
105 104

Tu
GP66+ (% of CD4 T cells)

GP66+ CD4 T cells

105 10 80
104

Percentage of
2.14

transcription factor
104 103
104 103 60
Tumour

Tumour
103
103 40 102
GP66 (APC)

103 5
100 100 100 20 101
100 4.00 26.0 72.0 25.0 86.1 13.2
0 0 1 2 3 4 5
0 100 103 104 105 100 103 104 105 100 103 104 Weeks after tumour inoculation
100 103 104 TCF1 (PE)
LN

r
ou

CD44 (UV737)
TD

Gated on GP66+CD44+ week 5 after tumour inoculation from TDLN

Tu

c Activation/proliferation TH1-like markers Positive regulators TFH-like Ectonucleotidase

PD1 (UV737) Ki67 (BV650) PSGL1 (Percp-e710) SLAMF1 (BV605) CD226 (FITC) SLAMF6 (BV605) FR4 (Percp-cy5.5) CD73 (AF700)
19.3 76.2 105 3.80 16.2 105 18.1 28.4 21.0 31.0 2.78 52.8 38.3 59.2 17.9 76.4 19.5 79.7
104 104 104
104 104
104 104 103
103 103 103
103 103 103 103
102
Molecule

102
100 100 100 100 100 100
100 102
1.16 3.30 16.7 63.4 50.2 3.32 47.2 0.74 6.94 37.5 2.07 0.38 2.17 3.52 0.54 0.27
100 103 104 100 103 104 100 103 104 100 103 104 100 103 104 100 103 104 100 103 104 105 100 103 104 105
TCF1 (PE)

d Sort and transfer: e f Tumour Transferred Endogenous

TDLN CD44+PD1+CD25– Gated on SMARTA T cells (GP66+CD45.1+CD44+) week 5 SMARTA T cells GP66+ cells
Gated on CD4 T cells
SLAMF1–IL18R1–CD39– spleen LCMV day 8 0.0003 0.0006
TRAMPC1-GP TBET (Pecy7) FOXP3 (PE-e610)
CD226+ SMARTA T cells 0.0002 >0.9999
2.5 × 106 100 0.0112 0.966 100
24
2.49
49 56.3 18.0 2.38 0.26

FOXP3+ (% of population)
0 62
0.62

TBET+ (% of population)
Transcription factor

5 Weeks 100 104 80 80

Analysis 103
GP66 (APC)

103 60 60
8 days 103
Day 0 102 102 40 40
0.5 × 106 100
SMARTA T cells 104
6.08 19.6 59.3 38.1 20 20
LCMV Armstrong
100 103 104 100 103 104 100 103 104 0
0
CD45.1 (UV395) TCF1 (PE)

Fig. 2 | Identification of a TCF1+lin− stem-like CD4 T cell population in mouse marker). d, Experimental design to test the functional capacity of sorted
cancer models. a, Representative I-AbGP66 tetramer staining in TDLNs (top) PD1+TCF1+lin− SMARTA T cells from TDLNs (n = 40 pooled mice as donors
and tumour (bottom) five weeks after TRAMPC1-GP inoculation. Summary from 2 independent experiments). e, Representative phenotypic analysis of
plots show the frequency and number of GP66+ CD4 T cells in each tissue. recovered SMARTAs and endogenous I-AbGP66 CD4 T cells in the spleen eight
b, Phenotypic analysis of GP66+ CD4 T cells in TDLNs (top) and tumours days after LCMV Armstrong infection. f, Frequency of SMARTA T cells or
(bottom) five weeks after TRAMPC1-GP inoculation. Summary plots show the virus-specific endogenous GP66+ cells expressing TBET and FOXP3. ‘Tumour
frequency and number of GP66+ CD4 T cells that express the respective lineage week 5’ represents SMARTA T cells in the tumour from donor mice prior to
transcription factor in each tissue. Data are median or mean ± s.d. (n > 5 mice sorting. Data are representative of 2 independent experiments (n = 8 recipient
per timepoint). c, Phenotypic characterization of GP66+ CD4 T cells in TDLNs mice); median ± 95% confidence interval; Kruskal–Wallis test with Dunn’s
of TRAMPC1-GP tumour-bearing mice after five weeks (n > 5 mice for each multiple-comparison tests.

antigen, as bulk PD1+ CD4 T cells were similarly skewed towards the Tbx21 and several effector TH1 molecules (Extended Data Fig. 5j–h).
TH1 lineage in TDLNs, blood and tumours across all models (Extended Notably, TCRβ sequencing showed that whereas in untreated condi-
Data Fig. 5c–g). tions, stem-like CD4 T cells shared clonality with Treg and TFH popula-
We next examined the transcriptional changes on stem-like CD4 tions in TDLNs (Extended Data Fig. 5m), stem-like CD4 T cells from
T cells after Treg depletion, and their relationship to the emerging Treg-depleted mice shared more than 90% of clonotypes with the clon-
TH1 cells. We performed scRNA-seq on bulk CD44+PD1+ CD4 T cells ally expanded TH1 cluster (Fig. 3f). Together, these data indicate that
from TRAMPC1-GP TDLNs untreated or 5 days after Treg depletion, the stem-like CD4 T cell population is a precursor to differentiated
with naive (CD44−CD62L+) CD4 T cells from lymph nodes included as TH1 cells in cancer.
a control. Transcriptional analysis identified five clusters of activated To further examine this relationship, we sorted tumour-specific PD1+
CD4 T cells, with high expression of markers related to antigen expe- SMARTA stem-like or Treg cells from TDLNs of 3-week TRAMPC1-GP
rience (Extended Data Fig. 5h,i). Consistent with our flow cytometry tumour-bearing mice (Supplementary Data 1d) and transferred
data, stem-like CD4 T cells (cluster 1) were present in both untreated these cells into congenically distinct tumour-matched wild-type or
and Treg-depleted mice. These cells were characterized by a unique FOXP3-DTR recipients (Fig. 3g). All recipient mice received diphtheria
transcriptional profile, with high expression of Tcf7, Slamf6, Lef1 and toxin two days after transfer and the phenotype of transferred cells
Gpr183, and significant enrichment for the chronic LCMV-specific was examined five days later. Transferred stem-like SMARTA T cells
CD4 T cell precursor signature26 (Extended Data Fig. 5j,k and Supple- underwent a greater expansion in both conditions compared with
mentary Table 4). Upon Treg depletion, TCF1+lin− CD4 T cells retained Treg cells (Fig. 3h) and were primarily found within TDLNs (Extended
their stem-like programme, but upregulated genes associated with Data Fig. 6a,b). Phenotypically, stem-like CD4 T cells re-established
TH1 differentiation (Extended Data Fig. 5l and Supplementary Table 2). the stem-like pool in TDLNs, with a small fraction differentiating into
In comparison, differentiated TH1 cells (cluster 3) comprised almost Treg and TFH cells in wild-type mice (Fig. 3i and Extended Data Fig. 6a).
exclusively of activated CD4 T cells from Treg-depleted mice expressed In Treg-depleted mice, stem-like CD4 T cells also re-established the

4 | Nature | www.nature.com
 a c Gated on GP66+ CD4 T cells from TDLN
TCF1 Treg TH1 TFH d TCF1 Treg TH1 TFH
Untreated Treg dep day 5 100
105

log10(GP66+ population)
WT or 104 0.65 0.65 104 54.1 1.82 DT

GP66+ CD4 T cells

DT 80
FOXP3-DTR

Percentage of
Day 28, 29 Analysis 104
Day 5, 11, 17 and 25 103 103 60
post depletion 103
3 weeks

TBET (Pecy7)
40
102 102 20 102
100 100
2.5 × 106 67.1 31.6 42.7 1.36 0 101
TRAMPC1-GP
100 103 104 100 103 104 0 15 30 45 60 0 15 30 45 60
FOXP3 (PE-e610) Days post tumour inoculation Days post tumour inoculation

b e CD4 T cells TDLN day 5 post depletion, GP61–77 peptide restim f Top 10 clonally expanded TCRs in TH1 cluster
Untreated Treg dep day 5 IFNγ+ (% CD4 T cells) IL-2+ (% CD4 T cells)
150 WT TCF1+ TH1 TFH Cycling
<0.0001 <0.0001 0.0006
9.67E-3 0 0.12 0.19 100 100
Treg dep 105 105
Tumour size (mm2)

100
104 104
10–1 10–1
DT
50 103 103
IFNγ (APC)

10–2 10–2
100 100
0 99.9 0.065 99.5 0.23
0 10 20 30 40 50
100 103 104 105 100 103 104 105 10–3 10–3 55 49 43 42 38 33 29 27 25 24
Days post tumour inoculation WT Treg dep WT Treg dep
IL-2 (PE)
>0.9999
g Sort and transfer: h TDLN stem CD4 T cells TDLN stem CD4 T cells TDLN Treg cells 0.0289 0.1166
TDLN CD45.1+ stem SMARTA 12–20,000 105 Stem CD4 T cells
WT WT WT DTR WT DTR

log10(no. of SMARTA T cells)

or WT to WT
TRAMPC1-GP TDLN CD45.1+ Treg SMARTA 5–15,000 0.022 0.25 0.17 Stem CD4 T cells
CD45.2 2.5 × 106 104
104 104 104 WT to DTR
3 weeks Treg cells
Analyse
103 WT to WT
CD44 (UV737)

Day 5 post depletion 103 103 103

Treg cells
2 weeks WT to DTR
200,000 100 100 100 102
CD45.1
CD45.2 WT or
SMARTA cells
FOXP3-DTR DT 100 103 104 100 103 104 100 103 104 101
CD45.1 (UV395)

TDLN phenotype j Tumour phenotype

i TDLN stem CD4 T cells TDLN stem CD4 T cells TDLN Treg cells >0.9999 0.0009 >0.9999
WT WT WT DTR WT DTR 0.0207 0.0060 100 0.2863 0.0952 100 0.0055 0.0346
50 Stem CD4 T cells

FOXP3+ (% of SMARTA T cells)

TBET+ (% of SMARTA T cells)

TBET+ (% of SMARTA T cells)

4.80 0.53 31.2 11.7 3.22 41.8 WT to WT
104 104 104 80
40 80 Stem CD4 T cells
WT to DTR
103 103 103 30 60 60 Treg cells
TBET (Pecy7)

WT to WT
20 40 40 Treg cells
100 100 100
WT to DTR
83.3 11.4 41.1 16.0 3.96 51.0 10 20 20
100 103 104 105 100 103 104 105 100 103 104 105
FOXP3 (PE-e610) 0 0 0

Fig. 3 | Regulatory T cells actively suppress differentiation of stem-like median; two-sided unpaired Mann–Whitney U test. f, Cluster distribution of the
CD4 T cells to T H1 cells. a, Experimental design. Treg cells were transiently ten most dominant TCR clonotypes in the TH1 cluster from Treg-depleted mice.
depleted in TRAMPC1-GP tumour-bearing mice after three to four weeks by The number of cells sharing the respective TCR clonotype is indicated below
intraperitoneal administration of diphtheria toxin (DT) on two consecutive the graph (n = 4 mice pooled for sorting). g, Experimental design. PD1+ stem
days. WT, wild type. b, Tumour kinetics shown as mean tumour diameter ± s.e.m SMARTA T cells or Treg SMARTA T cells from TRAMPC1-GP TDLNs were
in untreated (n = 10) or Treg-depleted (dep) (n = 13) mice. Analysed by two-sided transferred into tumour-matched wild-type or FOXP3-DTR recipients. h, Total
unpaired Mann–Whitney U test at endpoint. c,d, Phenotype of GP66+ CD4 number of recovered stem SMARTA T cells or Treg SMARTA T cells in wild-type
T cells in TDLNs of untreated mice or at various timepoints after Treg depletion. or Treg-depleted mice in TDLNs. i,j, Phenotype of stem and Treg SMARTA T cells
Summary plots show the frequency (c) or total number (d) of GP66+ phenotypes in TDLNs (i) and tumours ( j) after transfer. Data are representative of two
in TDLNs at various timepoints prior to and after Treg depletion. Data are independent experiments (25–30 pooled mice for each experiment, n = 3 or 4
mean ± s.e.m, n ≥ 5 mice per timepoint. e, Representative IFNγ and IL-2 staining recipient mice per experiment for each condition). Data are median; Kruskal–
after in vitro restimulation (restim) with GP61–77 peptide in CD4 T cells from Wallis test with Dunn’s multiple-comparison tests.
TDLNs in untreated (n = 9) mice or 5 days after Treg depletion (n = 12). Data are

stem-like CD4 T cell pool while simultaneously differentiating into

TH1 cells in TDLNs and tumours (Fig. 3i,j and Extended Data Fig. 6a,b). Requirement of CD4 T cells for CD8 T cell
By contrast, SMARTA Treg cells primarily retained their FOXP3+ phe- differentiation
notype in both conditions across all tissues examined (Fig. 3i,j and Although we found that stem-like CD4 T cells underwent TH1 differ-
Extended Data Fig. 6a,b). Together, these data show that stem-like CD4 entiation after Treg depletion, it was unclear whether this process con-
T cells can give rise to TH1 cells in the cancer environment, but Treg cells tributed to effective anti-tumour immunity. We found that stem-like
actively suppress their differentiation. Treg suppression of stem-like CD4 T cells were required, given that TRAMPC1-GP and B16F10-GP
CD4 T cell differentiation also has therapeutic relevance. Despite Treg tumours were not controlled after total CD4 T cell depletion compared
cells exclusively expressing high levels of CTLA4 in TDLNs and tumours with Treg depletion alone (Fig. 4a,b and Extended Data Fig. 7a,b). This
(Extended Data Fig. 6c), only the depleting anti-CTLA4 clone 9H10 CD4-dependent effect was not mediated by direct cytotoxicity, as CD8
(refs. 14,27,28), significantly induced a TH1 response in both tissues, T cell depletion in Treg-depleted mice abrogated tumour control in two
leading to sustained tumour control in TRAMPC1-GP and orthotopic of the three models examined (Extended Data Fig. 7a–c). These find-
RENCA kidney cancer models (Extended Data Fig. 6d–j). Together, ings highlight the requirement of stem-like CD4 T cells and intact CD8
our findings indicate that tumour-specific stem-like CD4 T cells act T cell responses. Therefore, we next examined the effect of Treg cells
as a precursor population to helper or regulatory T cells during the and stem-like CD4 T cell help on tumour-specific CD8 T cells, using
cancer response. Furthermore, these data suggest that differentiation our Treg cell or full CD4 T cell depletion models (Fig. 4a). We found that
of stem-like CD4 T cells to TH1 cells might be crucial for effective tumour GP-specific (GP33+) and SPAS1+ (an endogenous tumour antigen) CD8
control and response to immunotherapy. T cells expanded and underwent robust effector differentiation within

Nature | www.nature.com | 5
Article
a b Untreated Anti-CD4 (GK1.5)
c Gated on GP33+CD44+ CD8 T cells - from TDLN day 5 post depletion
0.563 0.155
DT Anti-CD4 (GK1.5) Untreated Treg dep CD4 dep
150
WT or
Analyse
0 5.62 45.2 1.60 6.55 12.1 100 <0.001 <0.001 100 <0.001 0.0325
104 104 104

Tumour size (mm2)

Day 28, 29 Day 5, 11, 17

GZMB+ (% of GP33+)

Ki67+ (% of GP33)
FOXP3-DTR 0.3437

GZMB (AF700)
and 25 80 80
3 weeks post depletion 100
103 103 103 60 60
50 40 40
2.5 ×106 100 100 100
TRAMPC1-GP 5.06 89.3 27.8 25.4 9.22 72.1 20 20
0
0 10 20 30 40 50 100 103 104 100 103 104 100 103 104 0 0
Days post tumour inoculation TCF1 (PE)
WT SMARTA T cells TetON SMARTA T cells
d e 200–500,000
f WT SMARTA T cells TetON TBET SMARTA T cells + anti-PDL1 + anti-PDL1
+ anti-PDL1 + anti-PDL1
WT SMARTA T cells <0.001 0.001
100

(% of SMARTA T cells)

(% of SMARTA T cells)
SMARTA × rtTA TBET T2A VEX or 5.96 0.87 26.2 1.79 50
TRE TetON TBET SMARTA T cells 80 40

TBET (Pecy7)
TetON TBET SMARTA T cells 103 103

TBET+
rtTA 60 30

IFNγ+
Doxycycline
BM HSCs 8 weeks 10 days Analyse
14 days day 14 40 20
102 102
in culture WT 100 100 20 10
88.2 4.94 66.1 5.90
2.5 × 106 Anti-PDL1 0 0
TRAMPC1-GP 100 103 104 100 103 104
Sort HSCs FOXP3 (PE-e610)

i WT SMARTA g WT SMARTA T cells TetON TBET SMARTA T cells

h WT SMARTA T cells TetON TBET SMARTA T cells

TIM3+CD44+PD1+ CD8 T cells)

log10(no. of TBET+GP66+ cells)
WT SMARTA + anti-PDL1 + anti-PDL1 + anti-PDL1 + anti-PDL1 + anti-PDL1
104 <0.001 105 <0.001
TetON SMARTA

CD8 T cells from TLDN

6.74 3.93 28.3 2.58 7.75 10.2 16.3 2.78
TetON SMARTA + anti-PDL1
104 104
TBET (Pecy7)

TIM3 (BV421)

log10(no. of
150 103 103 103 104

CD44+PD1+
Tumour size (mm2)

Endogenous GP66+ CD4 T cells - TDLN

TRAMPC1-GP
<0.0001

102 102 103 103

0.0003

100
0.0407

102 103
50 100 100 100 100 70.0
82.6 6.74 57.9 11.2 4.05 78.0 10.9
101 102
0 100 103 104 100 103 104 100 103 104 100 103 104
0 10 20 30 40 FOXP3 (PE-e610) TCF1 (PE)
Days post tumour inoculation

CD8 T cells per g tumour)

endogenous CD4 T cells)
<0.001

log10(TIM3+ endogenous
104 105 0.0091

CD8 T cells from tumour

j WT no transfer WT SMARTA 105 0.21 0.030 105 0.19 0.47 45.0 2.07 40.7 0.64
log10(no. of IFNγ+ 104 104
IFNγ (APC)

TetON TBET SMARTA 104

TIM3 (BV421)
CD44+PD1+
120 104 104 103
Tumour size (mm2)

0.0100 0.0584

B16-F10-GP 103 103 103

80 103 103
<0.0001

102
102
40 100 100 100 100
99.5 0.27 97.7 1.66 40.5 12.4 42.8 15.9
101 101
100 103 104 100 103 104 100 103 104 100 103 104
0
0 5 10 15 IL-2 (PE) TCF1 (PE)
Days post tumour inoculation

Fig. 4 | Stem-like CD4 to T H1 differentiation is sufficient to promote to GP61–77 peptide stimulation. g, Phenotype and cytokine production of
effector CD8 T cell responses in TDLNs in the presence of Treg cells. endogenous GP66+CD44+ CD4 T cells in TDLNs for each group 14 days after
a, Experimental design. Treg cells or total CD4 T cells were transiently anti-PDL1 therapy in response to GP61–77 peptide stimulation. h, Phenotype
depleted in TRAMPC1-GP tumour-bearing mice after three to four weeks. of bulk activated (CD44+PD1+) CD8 T cells in TDLNs or tumours for each
b, Tumour kinetics as shown by tumour diameter for untreated or total CD4 group. i, Tumour growth kinetics in TRAMPC1-GP mice for each group. Data
T cell-depleted mice (untreated, n = 10; CD4-depleted, n = 9). Data are are mean ± s.e.m. Data in f–i represent 3 independent experiments (n = 4–6
mean ± s.e.m.; unpaired two-sided Mann–Whitney U test. c, Phenotypic mice per group for each experiment). Medians are shown in each summary plot;
analysis of GP33+ CD8 T cells in TDLNs 5 days after Treg or total CD4 T cell Mann–Whitney U test or Kruskal–Wallis test with Dunn’s multiple-comparisons
depletion (n = 11–16 mice per group). Data are median; Kruskal–Wallis test tests where appropriate. j, Tumour growth kinetics in B16F10-GP mice
with Dunn’s multiple-comparisons tests. d, Schematic of construct and that received no SMARTA T cells, wild-type SMARTA T cells or TetON TBET
process to make TetON TBET SMARTA T cells using a bone marrow (BM) SMARTA T cells. Data are mean ± s.d.; representative of two independent
haematopoietic stem cell (HSC) chimera lentiviral system. e, Experimental experiments; Kruskal–Wallis test with Dunn’s multiple-comparison tests
design. f, Phenotype and cytokine production of transferred wild-type or (n = 5–7 mice per group for each experiment).
TetON TBET SMARTA T cells in TDLNs 14 days after PDL1 therapy in response

TDLNs five days after Treg depletion (Fig. 4c and Extended Data Fig. 7d,e), T cells, we also investigated whether promoting differentiation
a process previously limited to the tumour29–31. By contrast, total CD4 of stem-like CD4 T cells to TH1 cells could overcome resistance of
T cell depletion resulted in expansion of stem-like CD8 T cells in TDLNs PDL1 blockade in non-responsive models. To test this, we generated
(LN-stem CD8 T cells) with significantly reduced effector differentiation tetracycline-inducible TBET-overexpressing SMARTA T cells (Fig. 4d
across all models examined (Fig. 4c and Extended Data Fig. 7d–i). In and Supplementary Data 1c). Upon transfer, these TetON SMARTA
agreement, transcriptional analysis showed that LN-stem CD8 T cells T cells expanded and retained a stem-like phenotype in TDLNs prior
from Treg-depleted mice exhibited unique upregulation of genes related to doxycycline administration, similar to their wild-type counterparts
to effector differentiation (Id2, Tbx21, Gzmb, Ifngr1 and Cxcr6) and (Extended Data Fig. 8a). Upon doxycycline administration, TetON
IFN signalling compared with wild-type and total CD4-depleted mice SMARTA T cells expressed TBET and produced IFNγ in TDLNs, which
(Extended Data Fig. 7j–p and Supplementary Table 2). These data sug- was further augmented in response to anti-PDL1 (Fig. 4f and Extended
gest a model in which stem-like CD4 T cells or their differentiation into Data Fig. 8a). TetON SMARTA T cell transfer resulted in expansion of
TH1 cells is required to induce effector output from tumour-specific endogenous GP66+ and bulk PD1+ stem-like CD4 T cells and TH1 dif-
LN-stem CD8 T cells. ferentiation with increased IFNγ and IL-2 production in TDLNs and
tumours. Notably, this effect was fivefold higher with anti-PDL1 therapy
than in untreated and anti-PDL1-treated wild-type SMARTA groups
TH1 cells are sufficient to induce effector CD8 T cells (Fig. 4g and Extended Data Fig. 8b–d). Furthermore, TetON SMARTA
We next examined whether differentiation of stem-like CD4 T cells T cell transfer in combination with anti-PDL1 resulted in significant
to TH1 cells was sufficient to promote effector CD8 T cell differentia- expansion and effector differentiation of total PD1+ and GP33+ CD8
tion in TDLNs, even in the presence of Treg cells. Additionally, given T cells in all tissues, leading to tumour control in the unresponsive
the increased expression of cytotoxic molecules on effector CD8 TRAMPC1-GP model (Fig. 4h,i and Extended Data Fig. 8e–h). TetON

6 | Nature | www.nature.com
 a b c Gated on CD44+ SMARTA T cells from TDLN d WT untreated, no SMARTA T cells
Naive TetON Naive TetON TetON TBET Ifng-KO TetON TBET TetON + anti-PDL1
Sender Receiver
TBET SMARTA T cells TBET SMARTA T cells SMARTA T cells + anti-PDL1 SMARTA T cells + anti-PDL1 Ifng-KO TetON + anti-PDL1
LN-stem CD8 T cells

IFNγ+ (% of SMARTA T cells)

CD4 T cells 40 0.003 240
Cd8a gRNA (WT) Ifng gRNAs (KO)
Ccl5 Ccr2 105 4.63 8.78 105 0.35 3.15

Tumour size (mm2)

Ifng Ifngr1 200
30

LN-stem CD8 T cells

TH1

Lgals3 Spn 160

Transfer 104 104

IFNγ (APC)
Anxa1 Cxcr6
400,000 Doxycycline

0.036
20 120

0.003
Il21 Il2rg rtTA 103 103
Il2 Il2rb Doxycycline 80
Analyse 10
TFH

Il4 Il2rg 18 days 100 100

Day 14 40
Gpi1 Atp5b 61.3 25.3 54.9 41.6
Tnfsf8 Tnfrsf8 0 0
100 103 104 100 103 104 WT KO 0 10 20 30 40
2.5 × 106 Anti-PDL1 IL-2 (PE)
TCF1+

CD226 Pvr Days post tumour inoculation

TRAMPC1-GP
Ligand Receptor
e TetON TBET Ifng-KO TetON TBET
TetON + anti-PDL1 Ifng-KO TetON + anti-PDL1 f TetON TBET Ifng-KO TetON TBET

TBET+ (% of CD44+PD1+ cells)

TBET+ (% of CD44+PD1+ cells)

SMARTA T cells + anti-PDL1 SMARTA T cells + anti-PDL1 SMARTA T cells + anti-PDL1 SMARTA T cells + anti-PDL1
40 0.043 40 0.024 0.029
104 104 104
CD44+PD1+ CD4 T cells

18.4 7.65 7.92 8.87

Total PD1+ CD8 T cells

104 104 26.8 1.53 8.47 1.69

log10(GZMB+ cells
per g tumour)
30 30

GZMB (AF700)
TBET (Pecy7)

from tumour
from tumour

103 103 103

103 103
20 20

100 100 103

10 10
100 100
23.3 50.7 9.11 74.1 53.1 18.6 63.1 26.7
0 0 102
100 103 104 100 103 104 100 103 104 100 103 104
Tumour TDLN Tumour
FOXP3 (PE-e610) TCF1 (PE)

g h WT P14 T cells Ifngr1-KO P14 T cells 0.009 WT P14 T cells Ifngr1-KO P14 T cells <0.001

GZMB+ (% of P14 T cells)

107 100
Doxycycline

log10(no. of P14 T cells)

Analyse 105 0.22 105 0.044 104 20.6 26.1 104 5.56 12.2
WT or

GZMB (AF700)
DT 106 80
CD44 (UV737)

Day 5
FOXP3-DTR
post depletion 104 104
105 103 103 60
3 weeks 103 103 104 40
100 100
100 100 103 20
99.8 99.9 4.12 49.2 2.22 80.0
2.5 × 106 WT or 102 0
TRAMPC1-GP Ifngr1-KO P14 T cells 100 103 104 100 103 104 WT KO 100 103 104 100 103 104 WT KO
CD45.1 (UV395) TCF1 (PE)

i 0.022 j Gated on PD1+CD45RA– CD8 T cells from RCC tumour 0.026 0.011 k 1.00
TH1 high
101 101

Progression-free survival
+
102
TH1 low TH1 high CD39+GZMB+ (% of total) + + + + + + ++ + n = 13 TH1 low
100 101 100 0.75
CD4 (% of total)

32.0 2.02 50.3 3.28

CD8 (% of total)

10–1 10–1
100
GZMB (AF700)

103 103 10–2 0.50

10–2 10–1 +++

10–3
10–3 + + +
10–2 0.25
102 102 10–4 + +
n = 34
10–4
100 100 10–3 10–5 P = 0.001
10–5 56.2 9.83 38.9 7.55 0
10–4 10–6
gh
w

hi 0 400 800 1,200 1,600

gh gh
lo

102 103 104 102 103 104

w

1 hi hi
lo

lo
1

T TCF1 (AF488) 1 1 Days after start of immunotherapy

H

1
T

T T
H

H
T

T
H

Fig. 5 | TH1 cell-derived IFNγ is required for differentiation of LN-stem CD8 Data are median; two-sided unpaired Mann–Whitney U test (n = 6–8 mice per
T cells to effector CD8 T cells in TDLNs. a, NicheNet analysis showing top group). g, Experimental design to test the intrinsic requirement for IFNγ on
ligand–receptor interactions between CD4 T cell populations and LN-stem CD8 tumour-specific stem-like CD8 T cells. h, Phenotype of wild-type and inducible
T cells from Treg-depleted mice. b, Experimental design to test the requirement Ifngr1-KO P14 CD8 T cells in TDLNs five days after Treg depletion. Data are
of TH1 cell-derived IFNγ for effector CD8 T cell differentiation. gRNA, guide representative of two independent experiments (n = 5–6 mice per group
RNA. c, Cytokine production in response to GP61–77 peptide stimulation for for each experiment). Data are median; two-sided unpaired Mann–Whitney
transferred wild-type TetON TBET or Ifng-KO TetON TBET SMARTAs in TDLNs U test. i,j, Total frequency of tumour CD4 T cells and phenotype of tumour
14 days after anti-PDL1 therapy. d, Tumour growth kinetics in TRAMPC1-GP PD1+CD45RA− CD8 T cells in a cohort of patients with kidney cancer who
mice. Data are mean ± s.d.; Kruskal–Wallis test with Dunn’s multiple-comparison received immunotherapy. Data are median; two-sided unpaired Mann–Whitney
tests between groups (untreated, n = 10; TetON, n = 8; Ifng-KO TetON, n = 6). U test. k, Disease progression after start of immunotherapy in patients with
e,f, Phenotype of bulk PD1+ CD4 (e) and CD8 (f) T cells in tumours 14 days after primary kidney cancer, stratified into those with high or low TH1 CD4 T cell
transfer of TetON or Ifng-KO TetON SMARTA T cells with anti-PDL1 therapy. infiltration in the primary tumour based on optimal cut methods (n = 47 patients).

TBET-overexpressing SMARTA T cells also induced robust endogenous in the absence of helper CD4 T cells (Extended Data Fig. 7d–p). Further-
PD1+ TH1 differentiation, resulting in tumour control without the need more, we found reduced effector CD8 T cell differentiation in TDLNs
for checkpoint therapy in B16F10-GP tumours (Fig. 4j and Extended accompanied by accelerated tumour growth after Treg depletion when
Data Fig. 8i–k). Together, these results indicate that shifting differ- IFNγ, but not IL-12, was systemically blocked in both TRAMPC1-GP and
entiation of stem-like CD4 T cells to TH1 cells is sufficient to promote B16F10-GP tumour-bearing mice (Extended Data Fig. 9a–g).
effector CD8 T cell differentiation, enhance tumour control and rescue Given these observations, we next examined whether TH1 cell-derived
the response to anti-PDL1 therapy, despite the presence of Treg cells. IFNγ was required to stimulate effector CD8 T cell differentiation and
rescue the response to PDL1 blockade. To test this, we generated Ifng-
knockout (KO) TetON TBET SMARTAs using a CAS9 electroporation
IFNγ drives CD8 effector T cell differentiation system followed by transfer into mice with established TRAMPC1-GP
Given the notable effect of inducing stem-like CD4 to TH1 differen- tumours. Compared to TetON SMARTAs, Ifng-KO TetON SMARTAs
tiation on restoring effector CD8 T cell generation in TDLNs, we next were unable to control TRAMPC1-GP tumours despite the addition of
investigated potential mechanisms. NicheNet sender–receiver analysis anti-PDL1 therapy (Fig. 5d and Extended Data Fig. 9h,i). Ifng-KO TetON
predicted IFNγ from TH1 CD4 T cells as a top candidate ligand for induc- SMARTAs did not induce endogenous GP66+ or bulk PD1+ CD4 T cell
tion of transcriptional changes related to effector differentiation on differentiation to TH1 cells in TDLNs and tumours (Fig. 5e and Extended
LN-stem CD8 T cells (Fig. 5a and Supplementary Table 3). Indeed, IFNγ Data Fig. 9j), suggesting that IFNγ also has a role in altering the fate of
receptor (IFNGR1) expression and interferon signalling were upregu- stem-like CD4 T cells. Notably, GP33+ and bulk PD1+ CD8 T cells showed
lated in LN-stem CD8 T cells after Treg depletion, which was not observed a trend for reduced effector differentiation in TDLNs and significantly

Nature | www.nature.com | 7
Article
reduced effector CD8 T cell accumulation in the tumour (Fig. 5f and propose two possible fate choices of stem-like CD4 T cells that dictate
Extended Data Fig. 9k–l). Together, these findings suggest that TH1 the outcome of the cancer response (Extended Data Fig. 10n). Early after
cell-derived IFNγ is involved in promoting stem to effector CD4 and tumour inoculation, CD4 T cell differentiation is rapidly dominated by
CD8 T cell differentiation, which mediates tumour control and response Treg cells that through unknown mechanisms restrict tumour-specific
to anti-PDL1 therapy. CD4 T cells to stem-like or iTreg states. In this restricted fate, stem-like
Finally, to examine whether direct IFNγ signalling on stem-like CD8 CD4 T cells provide limited help, and antigen-specific CD8 T cell effector
T cells caused effector differentiation, we generated inducible Ifngr1-KO differentiation is constrained to the tumour29–31. Alternatively, when Treg
CD8 T cells from P14 mice (Extended Data Fig. 10a and Supplementary cells are depleted, stem-like CD4 T cells generate a robust TH1 response
Data 1c). Wild-type and Ifngr1-KO P14 CD8 T cells were left for three and promote differentiation of LN-stem CD8 T cells to effector CD8
days to similarly activate within TRAMPC1-GP TDLNs and establish a T cells through IFNγ, a model that we defined as active (Extended Data
stem cell pool (Extended Data Fig. 10b–d). After three days, mice were Fig. 10n). Of note, we showed that even in the presence of Treg cells,
treated with doxycycline to knock out Ifngr1 followed by Treg deple- endogenous stem-like CD4 T cells can generate TH1 cells, and that this
tion (Fig. 5g). Five days after Treg depletion, wild-type P14 CD8 T cells switch in differentiation fate is sufficient to promote effector output
underwent robust effector differentiation and upregulated IFNGR1, from LN-stem CD8 T cells and rescue response to immunotherapy.
GZMB and TIM3, with the accompanying loss of TCF1. By contrast, Finally, we found that human patients with kidney cancer with a robust
Ifngr1-KO P14 CD8 T cells underwent a significantly smaller expansion TH1 population have a higher frequency of effector CD8 T cells and
and retained a stem phenotype in TDLNs (Fig. 5h and Extended Data are highly responsive to immunotherapy. Together, these data high-
Fig. 10e–h). Endogenous GP33+ CD8 T cells within the same environ- light that stem-like CD4 T cell fate controls the anti-tumour response
ment as Ifngr1-KO P14 CD8 T cells underwent expansion and effector by regulating the switch between restricted and active T cell states.
differentiation in TDLNs (Extended Data Fig. 10i,j), highlighting the Given these observations, we speculate that targeting the large pool
direct requirement for IFNγ on LN-stem CD8 T cells. Together, these of PD1+ stem-like CD4 T cells and promoting their differentiation to
data suggest a model in which stem-like CD4-to-TH1 differentiation is TH1 cells will have important therapeutic implications for enhancing
sufficient to promote tumour-specific effector CD8 T cell responses in anti-tumour immunity.
TDLNs. Mechanistically, TH1 cell-derived IFNγ is intrinsically required
by LN-stem CD8 T cells to generate cytotoxic effector cells that medi-
ate tumour control and improve response to anti-PDL1 therapy. Thus, Online content
stem-like CD4 T cell fate acts as a switch that regulates restricted or Any methods, additional references, Nature Portfolio reporting summa-
active LN-stem CD8 T cell differentiation states. ries, source data, extended data, supplementary information, acknowl-
edgements, peer review information; details of author contributions
and competing interests; and statements of data and code availability
TH1 differentiation predicts immunotherapy response are available at https://doi.org/10.1038/s41586-024-08076-7.
The data from mouse models predict that patients with cancer who
have a robust TH1 response would have a greater number of cytotoxic
1. Galon, J. et al. Type, density, and location of immune cells within human colorectal
effector CD8 T cells and be more responsive to PD1 blockade. Although
tumors predict clinical outcome. Science 313, 1960–1964 (2006).
the majority of activated CD4 T cells within tumours had a TCF1+lin− 2. Herbst, R. S. et al. Predictive correlates of response to the anti-PD-L1 antibody
phenotype (Fig. 1), around 15% of patients had a significant TBET+ TH1 MPDL3280A in cancer patients. Nature 515, 563–567 (2014).
3. Schreiber, R. D., Old, L. J. & Smyth, M. J. Cancer immunoediting: integrating immunity’s
CD4 T cell population (Extended Data Fig. 10k). In a cohort of 47 patients roles in cancer suppression and promotion. Science 331, 1565–1570 (2011).
with kidney cancer who were receiving immunotherapy, those with 4. Borst, J. et al. CD4+ T cell help in cancer immunology and immunotherapy. Nat. Rev.
robust TH1 responses in their resected primary tumours had a greater Immunol. 18, 635–647 (2018).
5. Speiser, D. E. et al. CD4+ T cells in cancer. Nat. Cancer 4, 317–329 (2023).
frequency of total and effector GZMB+ CD8 T cells, with significantly 6. Cohen, M. et al. The interaction of CD4+ helper T cells with dendritic cells shapes the
better progression-free survival after immunotherapy (Fig. 5i–k and tumor microenvironment and immune checkpoint blockade response. Nat. Cancer 3,
Supplementary Table 5). Of note, patients with TH1 responses did not 303–317 (2022).
7. Magen, A. et al. Intratumoral dendritic cell–CD4+ T helper cell niches enable CD8+ T cell
exhibit reduced total Treg or stem-like CD4 T cell infiltration within differentiation following PD-1 blockade in hepatocellular carcinoma. Nat. Med. 29,
tumours (Extended Data Fig. 10k,l). Additionally, TH1 CD4 T cell fre- 1389–1399 (2023).
quency in non-metastatic TDLNs also correlated with effector GZMB+ 8. Cui, C. et al. Neoantigen-driven B cell and CD4 T follicular helper cell collaboration
promotes anti-tumor CD8 T cell responses. Cell 184, 6101–6118.e13 (2021).
CD8 T cell frequency in patients with kidney cancer (Extended Data 9. Sharma, P. et al. Immune checkpoint therapy—current perspectives and future directions.
Fig. 10m). Together, these data indicate that human TH1 CD4 T cell dif- Cell 186, 1652–1669 (2023).
ferentiation correlated with a greater number of effector CD8 T cells 10. Binnewies, M. et al. Unleashing type-2 dendritic cells to drive protective antitumor CD4+
T cell immunity. Cell 177, 556–571.e16 (2019).
and improved checkpoint therapy response in patients with kidney 11. Kim, J. M., Rasmussen, J. P. & Rudensky, A. Y. Regulatory T cells prevent catastrophic
cancer. This occurs in the presence of Treg cells, supporting the model autoimmunity throughout the lifespan of mice. Nat. Immunol. 8, 191–197 (2007).
that generation of TH1 cells from stem-like CD4 T cells can induce 12. Lahl, K. et al. Selective depletion of Foxp3+ regulatory T cells induces a scurfy-like
disease. J. Exp. Med. 204, 57–63 (2007).
productive effector CD8 T cell responses. 13. Spitzer, M. H. et al. Systemic immunity is required for effective cancer immunotherapy.
Cell 168, 487–502.e15 (2017).
14. Wei, S. C. et al. Distinct cellular mechanisms underlie anti-CTLA-4 and anti-PD-1 checkpoint
blockade. Cell 170, 1120–1133.e17 (2017).
Discussion
15. Alonso, R. et al. Induction of anergic or regulatory tumor-specific CD4+ T cells in the
Here we identified a tumour-specific PD1+TCF1+ lineage-negative CD4 tumor-draining lymph node. Nat. Commun. 9, 2113 (2018).
T cell with stem-like properties across cancer models and human 16. Nishikawa, H. & Sakaguchi, S. Regulatory T cells in tumor immunity. Int. J. Cancer 127,
759–767 (2010).
disease. Found primarily within TDLNs, these cells retained exten- 17. Jansen, C. S. et al. An intra-tumoral niche maintains and differentiates stem-like CD8
sive proliferation and differentiation capacity, and shared pheno- T cells. Nature 576, 465–470 (2019).
typic, transcriptional and functional similarities with stem-like CD8 18. Zhou, L., Chong, M. M. & Littman, D. R. Plasticity of CD4+ T cell lineage differentiation.
Immunity 30, 646–655 (2009).
T cells17,24,29,30,32–34 and precursor CD4 T cells26,35–38, highlighting parallels 19. Zhu, J., Yamane, H. & Paul, W. E. Differentiation of erffector CD4 T cell populations. Annu.
in CD4 T cell differentiation between cancer and other chronic dis- Rev. Immunol. 28, 445–489 (2010).
eases. We showed that tumour-specific CD8 T cells are not inherently 20. Brummelman, J. et al. High-dimensional single cell analysis identifies stem-like cytotoxic
CD8+ T cells infiltrating human tumors. J. Exp. Med. 215, 2520–2535 (2018).
dysfunctional but can effectively generate effector cytotoxic cells 21. Hong, S.-W. et al. Immune tolerance of food is mediated by layers of CD4+ T cell
when appropriate CD4 T cell help is provided. Based on our data, we dysfunction. Nature 607, 762–768 (2022).

8 | Nature | www.nature.com
22. Kalekar, L. A. et al. CD4+ T cell anergy prevents autoimmunity and generates regulatory 32. Alfei, F. et al. TOX reinforces the phenotype and longevity of exhausted T cells in chronic
T cell precursors. Nat. Immunol. 17, 304–314 (2016). viral infection. Nature 571, 265–269 (2019).
23. Barber, D. L. et al. Restoring function in exhausted CD8 T cells during chronic viral infection. 33. Gearty, S. V. et al. An autoimmune stem-like CD8 T cell population drives type 1 diabetes.
Nature 439, 682–687 (2006). Nature 602, 156–161 (2022).
24. Im, S. J. et al. Defining CD8+ T cells that provide the proliferative burst after PD-1 therapy. 34. Eberhardt, C. S. et al. Functional HPV-specific PD-1+ stem-like CD8 T cells in head and
Nature 537, 417–421 (2016). neck cancer. Nature 597, 279–284 (2021).
25. Siddiqui, I. et al. Intratumoral Tcf1+ PD-1+ CD8+ T cells with stem-like properties promote 35. Moguche, A. O. et al. ICOS and Bcl6-dependent pathways maintain a CD4 T cell population
tumor control in response to vaccination and checkpoint blockade immunotherapy. with memory-like properties during tuberculosis. J. Exp. Med. 212, 715–728 (2015).
Immunity 50, 195–211.e10 (2019). 36. Shin, B. et al. Effector CD4 T cells with progenitor potential mediate chronic intestinal
26. Xia, Y. et al. BCL6-dependent TCF-1+ progenitor cells maintain effector and helper CD4+ inflammation. J. Exp. Med. 215, 1803–1812 (2018).
T cell responses to persistent antigen. Immunity 55, 1200–1215.e6 (2022). 37. Zou, D. et al. CD4+ T cell immunity is dependent on an intrinsic stem-like program. Nat.
27. Du, X. et al. A reappraisal of CTLA-4 checkpoint blockade in cancer immunotherapy. Cell Immunol. 25, 66–76 (2024).
Res. 28, 416–432 (2018). 38. Kratchmarov, R. et al. TCF1–LEF1 co-expression identifies a multipotent progenitor cell
28. Simpson, T. R. et al. Fc-dependent depletion of tumor-infiltrating regulatory T cells (TH2-MPP) across human allergic diseases. Nat. Immunol. 25, 902–915 (2024).
co-defines the efficacy of anti–CTLA-4 therapy against melanoma. J. Exp. Med. 210,
1695–1710 (2013). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
29. Prokhnevska, N. et al. CD8+ T cell activation in cancer comprises an initial activation published maps and institutional affiliations.
phase in lymph nodes followed by effector differentiation within the tumor. Immunity 56,
107–124.e5 (2023). Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this
30. Connolly, K. A. et al. A reservoir of stem-like CD8+ T cells in the tumor-draining lymph article under a publishing agreement with the author(s) or other rightsholder(s); author
node preserves the ongoing antitumor immune response. Sci. Immunol. 6, eabg7836 self-archiving of the accepted manuscript version of this article is solely governed by the
(2021). terms of such publishing agreement and applicable law.
31. Li, Z. et al. In vivo labeling reveals continuous trafficking of TCF-1+ T cells between tumor
and lymphoid tissue. J. Exp. Med. 219, e20210749 (2022). © The Author(s), under exclusive licence to Springer Nature Limited 2024

Nature | www.nature.com | 9
Article
Methods transcription factors and various molecules. For experiments where
exogenous IL-12 was provided, 10 ng ml−1 IL-12 was added to the den-
Human sample processing and flow cytometry dritic cell–CD4 T cell well at the time of plating.
Patients were recruited in accordance with Emory University Institu- For high-dose stimulation assays, 4,000 U ml−1 IL-2 was used in com-
tional Review Board protocol (IRB00055316), with all patients pro- bination with CD3/CD28/CD2 beads at a ratio of 1 bead for 2 CD4 T cells.
viding informed consent. Tumours and TDLNs were collected from CD39+ CD4 T cells were analysed five days after stimulation by flow
patients undergoing partial or radical nephrectomy, prostatectomy, cytometry.
or transurethral surgery for resection of kidney, prostate or bladder To calculate the frequency of original cells that underwent division,
tumours. No statistical methods were used to predetermine sample we used the following calculation: X = (sum of percentage of cells in divi-
size. Samples were maintained in Hank’s Balanced Salt solution until sion i/2i), where i = 1:5. To obtain the final percentage of original cells we
processing. Samples were then cut into small pieces, digested with a used the formula X/(X + Y), where Y is the percentage of undivided cells.
collagenase/liberase enzyme cocktail, and homogenized using a MACS
dissociator. Digested tumours were then washed with buffer through Mice
a 70-μm filter into a single-cell suspension. Samples were then lysed Animal experiments were conducted and designed in accordance
using red blood cell ACK lysis buffer or ice-cold water followed by an with National Institutes of Health and the Emory University Institu-
equal volume of 1.8% of NaCl solution, followed by a 44% Percoll/RPMI tional Animal Care and Use Committee guidelines and approved under
gradient. Single-cell suspensions were then either used fresh or frozen PROTO201800261. Mice were housed in a 07:00–19:00 light cycle,
in freezing medium (FBS + 10%DMSO) at −80 °C for future use. 22 °C and controlled 40–50% humidity in clean pathogen-free rooms.
For flow cytometry, single-cell suspensions were stained with anti- C57BL/6 J mice (000664) and Pep Boy mice (B6.SJL-Ptprca Pepcb/
bodies listed in Supplementary Table 1 for 30 min at 4 °C or on ice in BoyJ, 002014) were purchased from Jackson laboratories between
fluorescence-activated cell sorting (FACS) buffer. Fixable near-IR or the ages of 7–10 weeks. LCMV DbGP33-specific TCR transgenic P14
aqua dead cell staining kit (Invitrogen) was used for live/dead staining. mice were a gift from the laboratory of R. Ahmed and were bred and
For intracellular staining, cells were permeabilized using the FOXP3 maintained at Emory University. LCMV GP66-77-specific SMARTA
Fixation/Permeabilization kit (eBioscience) in fixation buffer for 45 min mice (030450) were purchased from Jackson laboratories and were
at 4 °C. Cells were then stained for transcription factors and intracel- bred and maintained at Emory University. FOXP3-DTR mice (DEREG,
lular molecules in permeabilization buffer for 30 min at 4 °C or on ice. 032050-JAX) expressing the human diphtheria toxin receptor and the
After washing, samples were acquired in a Symphony instrument (BD GFP reporter12 were purchased from Jackson laboratories and used
Biosciences) and analysed using FlowJo (v10). for Treg depletion experiments. rtTA (Rosa26-CAGs-rtTA3 knock-in,
029627) mice were purchased from Jackson laboratories, bred in
Human in vitro stimulations, co-cultures and functional house and crossed with transgenic SMARTA mice for TBET overex-
analyses pression experiments. iCas9 (B6;129S4-Gt(ROSA)26Sortm1(rtTA*
Single-cell suspensions from fresh or frozen human tumour sam- M2)Jae Col1a1tm1(tetO-cas9)Sho/J, 029415) mice were purchased
ples were stained with CTV according to manufactures instructions from Jackson laboratories, bred in house and crossed with rtTA mice
(Thermo) at a concentration of 1 μl CTV per 10 million cells in PBS. If for inducible knockout experiments. For tumour experiments, male
frozen samples were used for stimulation, cells were rested in 10% sup- C57BL/6J mice were used for the TRAMPC1-GP cell line, while female
plemented RPMI at 37 °C for 4 h prior to sort. CTV-labelled stem-like and male C57BL/6J mice were used for B16F10-GP and MC38 experi-
and CD39+ Treg cells/EOMES+ CD4 T cells were sorted according to the ments. Female Balb/c mice were used for RENCA-HA or orthotopic
gating strategy in Supplementary Data 1a in the Becton Dickinson FACS RENCA-HA-luciferase experiments.
Aria II Cell Sorter. Stem-like CD4 T cells were defined as live CD4+PD1+C
D45RA−CD28+CD26+CD127+CD39−, CD39+ effectors were defined as live Tumour cell lines, subcutaneous injection and orthotopic
CD4+PD1+CD45RA−CD28+CD26−CD127−CD39+ (representative plots for surgical renal implants
sorting strategy shown in Supplementary Data 1a). For differentiating Tumours were scored according to Emory University tumour burden
between Treg cells and EOMES CD4 T cells, CD25 and CD38 were used policy 303, where any individual tumour >20 mm in any diameter was
to distinguish these two populations. Sorted CD4 populations were considered endpoint, as indicated by PROTO201800261. Mice were
cultured in 96-well U bottom plates in supplemented T cell medium randomized to experimental groups to normalize for tumour sizes
(RPMI, 10% FBS, 1% penicillin-streptomycin, 1% l-glutamine, 1% sodium prior to the start of treatments. No statistical methods were used to
pyruvate, 1% non-essential amino acids, 0.0005% 2-mercaptoethanol) predetermine sample size. Investigators were not blinded to group
and 10–20 U ml−1 IL-2 (Peprotech). Stimulating conditions were per- allocation during experimental setup, data collection, or analysis.
formed using anti-CD3/CD28/CD2 beads (Miltenyi Biotech) at a ratio TRAMPC1 cell line was obtained from American Type Culture Collection
of 1 bead for 2 CD4 T cells and polarization cytokines according to the (ATCC) and full length LCMV glycoprotein was added using lentiviral
condition. TH0 condition: 10–20 U ml−1 IL-2, TH1 condition: 10–20 U ml−1 transduction29. B16 F10 was obtained from ATCC and the full length
IL-2, 10 ng ml−1 IL-12; Treg condition: 10–20 U ml−1 IL-2, 10 ng ml−1 TGFβ, LCMV glycoprotein was made by lentiviral transduction in the labora-
1 μg ml−1 anti-IFNγ; EOMES condition: 10–20 U ml−1 IL-2, 10 ng ml−1 IL-12, tory of R. Ahmed. B16F10-GP cell line was a gift from the laboratory of
50 ng ml−1 IL-4, 1 μg ml−1 anti-IFNγ; TFH condition: 50 ng ml−1 activin A, R. Ahmed. The MC38 cell line was obtained from ATCC. The RENCA cell
5 ng ml−1 IL-12. Samples were analysed 5 days after stimulation by flow line was obtained from ATCC and transduced using a lentivirus con-
cytometry for proliferation as well as expression of transcription fac- taining the influenza haemagglutinin (HA) for stable RENCA-HA trans-
tors and various molecules. duction. RENCA-HA-luciferase was made in house by an additional
For dendritic cell co-cultures, CTV-labelled PD1+ stem-like, CD39+ transduction of a lentiviral plasmid containing luciferase and neo­
CD4 T cells and bulk dendritic cells (CD3−HLA-DR+CD11c+) were sorted mycin. Cells were selected on neomycin resistance for 10 days. All cells
from matched patient tumours as shown in Supplementary Data 1a. were cultured at 37 °C in 5% CO2 in appropriate media. All cell lines
Sorted dendritic cells were irradiated with 10 Gy prior to being placed were tested annually for mycoplasma infection and tested negative. For
in culture. Co-cultures were plated at one dendritic cell per one CD4 tumour inoculation, cells were detached from culture using 0.05%
T cell population for as many cells as possible based on the number of trypsin and saturated with respective media. Cells were then washed
cells recovered for each patient. Samples were analysed five days after with PBS twice and resuspended in PBS at different concentrations and
stimulation by flow cytometry for proliferation as well as expression of every mouse was injected with 100 μl subcutaneously at the following
concentrations: TRAMPC1-GP 2.5 × 106 cells, B16F10-GP 2.5 × 105 or of the crosses iCAS9 × P14 rtTA SMARTA mice were collected. Sple-
5 × 105 cells, MC38 2.5 × 105 cells, and RENCA-HA 2.5 × 105 cells. nocytes were processed in sterile conditions as described above. No
RENCA-HA-luciferase cells were prepared as described above. Prior collagenase digestion was performed for adoptive transfers. CD8 and
to surgical implant, RENCA-HA-luciferase cells were mixed at a 1:1 ratio CD4 T cells were isolated using EasySep mouse CD8 isolation (19853)
with Matrigel (356231 Corning) for a final 1 × 105 cell concentration per or naive CD4 negative selection (19765) kits (StemCell), respectively.
mouse in 20 μl. For orthotopic RENCA-HA-luciferase tumour experi- For tumour experiments, 250,000 to 1 million isolated P14 CD8 T cells
ments, Balb/c female mice were placed under isoflurane anaesthetic and 200,000–500,000 naive SMARTA CD4 T cells were transferred
and received 0.5 mg kg−1 bupenorphrine SR (sustained release) and intravenously unless otherwise specified for an experiment. For LCMV
6 mg kg−1 of lidocaine subcutaneously. Mice were shaved along the right Armstrong infection experiments, 1,000–5,000 P14 CD8 T cells were
flanks and skin was disinfected with Prevantics swabs. A small verti- transferred intravenously.
cal incision was made on the right lateral side of the mouse above the For TDLN SMARTA re-transfer experiments, activated stem-like
kidney. The kidney was then lifted upon the body surface and the mix- SMARTA CD4 T cells were sorted according to Supplementary Data 1b
ture of cells and Matrigel was injected using a low-dose insulin syringe and 400–3,000 sorted cells were transferred intravenously into naive
into the subcapsular space of the kidney. 6-O absorbable sutures were mice that were subsequently infected with LCMV Armstrong. Mice
used for musculature closure and 5-O non-absorbable synthetic sutures were infected with 2 × 105 pfu of LCMV Armstrong via intraperitoneal
were used to close the skin. After 7 days, Balb/c mice were injected with injection as previously described39.
200 μl d-luciferin potassium salt/PBS solution into the intraperitoneal For TDLN re-transfer experiments into tumour-matched mice,
space for 8 min for in vivo bioluminescence imaging to confirm tumour CD45.1+PD1+ stem-like or Treg SMARTA cells were sorted according
growth prior to the start of treatment. to Supplementary Data 1d. A total of 5,000–20,000 cells was trans-
TRAMPC1-GP cells were cultured in DMEM with glucose supplemented ferred intravenously via tail vein into 2- to 3-week-old TRAMPC1-GP
with 10% FBS (Corning), 1% l-glutamine, 1% penicillin-streptomycin, tumour-bearing CD45.2+ WT or FOXP3-DTR mice. SMARTAs were left
dehydroisoandrosterone (1.65 μg), and insulin (2.5 mg). B16F10-GP cells untreated for 2 days followed by 1 μg of diphtheria toxin administra-
were cultured in DMEM with glucose supplemented with 10% FBS (Corn- tion. Mice were analysed 5 days after diphtheria toxin administration
ing), 1% l-glutamine, 1% penicillin-streptomycin and 1% sodium pyru- and 7 days after transfer.
vate. MC38 cells were cultured in DMEM with glucose supplemented
with 10% FBS (Corning), 1% l-glutamine, 1% penicillin-streptomycin, In vivo treatments and depletions
1% sodium pyruvate and 1% 100× non-essential amino acids and 10 mM Treg cells were depleted in FOXP3-DTR (DEREG) mice bearing tumours
HEPES. RENCA-HA and RENCA-HA-luciferase cells were cultured in RPMI by intraperitoneal injection of 1 μg of diptheria toxin (Sigma) diluted in
with 10% FBS (Corning), 1% l-glutamine, 1% penicillin-streptomycin, endotoxin-free PBS on two consecutive days. CD4 T cells were depleted
1% sodium pyruvate and 1% 100× non-essential amino acids. All cell by intraperitoneal injection of 300 μg anti-CD4 antibody (GK1.5)
lines used were negative for Mycoplasma and other infectious agents. diluted in endotoxin-free PBS. For cytokine blocking experiments,
anti-IL-12p40 (250 μg, BioXcell) or anti-IFNγ (250 μg, BioXcell) antibod-
Mouse tissue processing and flow cytometry staining ies were injected intraperitoneally every other day for the duration of
Tumours, lymph nodes, spleens and lungs were collected and digested the experiment. CD8 T cells were depleted with 250 μg per mouse of
in collagenase D (2 mg ml−1) shaking for 25 min at 37 °C. Inguinal, axil- anti-CD8b (Lyt 3.2) antibody intraperitoneally every other day. For
lary and brachial lymph nodes on the side of tumour inoculation were PDL1 or CTLA4 blocking experiments, 200 μg of anti-PDL1 (10 F.9G2,
pooled as TDLNs. All digested tissues were washed with RPMI supple- BioXcell), anti-CTLA4 (4F10, BioXcell) or anti-CTLA4 (9H10, BioXcell)
mented with 2–5%FBS through a 70-μm filter into single-cell suspen- were administered intraperitoneally every three days. Doxycycline
sion. Tumours and spleen were RBC lysed using ACK lysis solution and hyclate (Sigma) was administered with one dose (25 mg kg−1) of intra-
resuspended in 2–5% RPMI. Tumours went through an additional 44% peritoneal injection followed by administration in drinking water at
Percoll/RPMI gradient for 10 min to remove excess fat prior to staining. a concentration of 2 mg ml−1 supplemented with 2% sucrose in sterile
Livers and lungs went through a 44% and 67% Percoll gradient for 20 min conditions to induce CAS9 or to overexpress TBET.
to remove excess fat and hepatocytes. Mouse tissues were stained with
antibodies listed in Supplementary Table 1. For extracellular staining, 10x scRNA-seq and analysis
cells were stained for 30 min at 4 °C on ice in FACS buffer. Cells were then For human scRNA-seq, tumour single-cell suspensions were stained
washed and fixed in FOXP3 Fixation/Permeabilization kit (eBioscience) and sorted on a Beckton Dickinson FACS Aria II Cell Sorter. Activated
for 25 min at room temperature or overnight at 4 °C. For intracellular CD4 T cells were sorted based on live CD3+CD4+CD8−PD1+CD45RA− from
staining, cells were stained in permeabilization buffer for 30–45 min at kidney tumours from two separate patients. For mouse scRNA-seq,
room temperature if fixation was performed at room temperature or activated CD4 T cells were sorted based on live CD4+CD8−PD1+CD44+
on ice if cells were fixed overnight. For intracellular cytokine staining, CD19−B220−HLADR−, activated CD8 T cells were sorted based on live
total cells from TDLNs were cultures in RPMI supplemented with 10% CD4−CD8+PD1+CD44+CD19−B220−HLADR− from TDLNs of wild-type
FBS in the presence of 2 μg ml−1 of LCMV GP61–80 peptide (GLKGPDIY (n = 12 pooled mice), Treg-depleted (n = 4 pooled mice) and CD4-depleted
KGVYQFKSVEFD), 1 μg ml−1 of Brefeldin A (BD) and 2 μg ml−1 of Monesin (n = 4 pooled mice) mice. Naive CD4 and CD8 T cells (PD1−CD44−CD62L+)
(BD) for 5 h at 37 °C before staining. Cells for intracellular staining were were spiked into each respective sample. scRNA-seq libraries were
staining for live/dead and surface proteins. Fixation was performed made using the Chromium single-cell 5′ Library and Gel Bead kit
using the BD Cytofix/Cytoperm kit according to the manufacturer’s (10x Genomics). Sorted cells were sorted and captured into the Gel
instructions. After 20 min at 4 °C, IFNγ, TNF and IL-2 (BD) were stained Beads-in-emulsion (GEMs). After the reverse transcription GEMs were
in BD permeabilization buffer for 30 min on ice. Data was acquired on a disrupted and cDNA was isolated and pooled. The barcoded cDNA was
Beckton Dickinson LSRII or a Symphony instrument (BD Biosciences). fragmented, end repaired, and A-tails were added followed by sample
All CD4 and CD8 tetramers were acquired from the NIH tetramer core index PCR. The purified libraries were sequenced to 50,000 reads per
facility at Emory University, supported by contract 75N93020D0000. cell on a HiSeq300 (Illumina) with 26 cycles for 1,8 cycles for index (i7)
and 91 cycles for read 2.
Adoptive transfers and LCMV Armstrong infection For data processing, samples were aligned, filtered and counted for
For adoptive transfers, spleens from congenically mismatched barcodes and unique molecular identifiers using Cellranger v3.1. Data
GP33-transgenic P14 mice, GP61–77 transgenic SMARTA mice, or any were further analysed using R v4.1.2 and the Seurat package v4.0.3
Article
(ref. 40). For human scRNA-seq, cells with a percentage of mitochon- Expanded HSCs were then spinfected in retronectin (20 μg ml−1)/
drial genes below 7% were included. Cells with more than 3,000 or fibronectin coated plates at 1,700g for 90 min at 32 °C with a lentivirus
fewer than 200 genes were considered outliers and were excluded from carrying the single guide RNA (sgRNA) of interest with the fluores-
downstream analysis. For the mouse scRNA-seq, cells with a percent- cent reporter VEX to mark infected cells. For TBET overexpression
age of mitochondrial genes below 7% were included. Cells with more experiments, expanded HSCs were infected with a lentivirus carrying
than 4,000 or fewer than 500 genes were considered outliers and were a tetracycline promoter and a shortened N-terminus TBET sequence,
excluded from downstream analysis. Samples from different groups producing a protein of 513 amino acids, that was attached to T2A and
or different patients were merged using the FindIntegrationMarkers the VEX reporter. Transgene positive (VEX+) HSCs were then sorted to
function in the Seurat package. Principal components analysis was reconstitute lethally irradiated recipient mice (Supplementary Data 1c).
performed, and the top eight to ten most significant components were Mice were irradiated with 2 doses of 5.75 Gy 6 h apart. For TBET over-
used for clustering. Differentially expressed genes within each cluster expression experiments, bulk LSK+ cells were sorted to reconstitute
or treatment were identified by the Seurat function FindAllMarkers irradiated mice, given that VEX expression was under the control of
for volcano plots. Differentially expressed genes between CD8 T cell doxycycline. Mice were kept on sterile cages under administration
and CD4 T clusters are provided in Supplementary Table 2. Gene set of neomycin antibiotics in their water for 3-weeks after irradiation.
enrichment analysis was performed using the VISION R package v.3.0 Chimerism was assessed 8–10 weeks after transfer by bleeding
(ref. 41) using the signatures in Supplementary Table 4 for human and experiments were performed between 10 and 14 weeks after
CD8 T cell stem-like signature, Treg signature, TH1 signature, cell cycle reconstitution.
signature, human TFH signature, mouse IFNγ signalling signature and
mouse CD4 precursor TCF1+ signature26. Signatures of human CD4 Lentiviral constructs and guide RNA design
stem-like, EOMES+ and Treg tumour-infiltrating lymphocytes from The sgRNAs for knockout experiments were designed using the CHOP-
our study from differentially expressed genes on our scRNA-seq are CHOP design tool45. sgRNAs were cloned into the pXPR_053 vector
included in Supplementary Table 4. For TCR analysis, TCR clonotypes (Addgene 113591) using a BsmBI restriction digest. sgRNA sequences
were aligned using CellRanger v3.1 and V, D, J gene segments were are listed in Supplementary Table 1 for IFNGR1. For TBET overex-
aligned using MixCR v3.0 (ref. 42). TCRα chains were filtered and pression experiments, we generated the TBET-T2A-VEX construct
unique clonotypes between populations were defined by matching (Supplementary Data 1d and Supplementary Table 1) and inserted the
CDR3 beta sequences. full sequence into the pTet-IRES-EGFP plasmid (Addgene #64238) using
PmeI and SalI cloning sites.
NicheNet analysis
NicheNet ligand–receptor analysis was performed between TDLN CD4 CRISPR–CAS9 gene editing and adoptive transfer
and CD8 T cell populations in wild-type and Treg-depleted mice. In brief, To generate TetON TBET Ifng-KO SMARTAs, sgRNAs targeting mouse
separate Seurat objects containing CD4 T cells and CD8 T cells were Ifng (5′-GGCTTTCAATGACTGTGCCG-3′, Mm.Cas9.IFNG.1.AA) and
included to contain cells derived from each treatment group: untreated, (5′-AAGAGATAATCTGGCTCTGC-3′, Mm.Cas9.IFNG.1.AC) were obtained
Treg depletion and total CD4 T cell depletion. We set the different helper from IDT (Supplementary Table 1). A sgRNA targeting mouse Cd8a
CD4 T cells (clusters 1–4) as sender populations and stem-like CD8 was used as a control (5′-CGTCCCACGTTATCTTGTTG-3′, Mm.Cas9.
T cells as the receiver population for each group. For target ligands, CD8A.1.AA). In brief, naive TetON TBET SMARTA T cells were isolated
we performed differentially expressed gene analysis (average log fold from the spleen of chimeric mice using the EasySep naive CD4 negative
change > 0.25 and P < 0.05) between wild-type and Treg-depleted CD4 selection kit (StemCell). Guides were mixed with a tracrRNA (1075928,
T cell subsets and chose the top significant ligands within each CD4 IDT) and incubated at 95 °C for 4 min. Guide or trans-activating CRISPR
T cell population. For the receiver stem-like CD8 T cell population, we RNA (tracrRNA) was left to return to room temperature prior to adding
focused our analysis on the most differentially expressed receptors CAS9 Nuclease (1081058, IDT). Naive CD4 T cells were then prepared in
(average log fold change > 0.25 and P < 0.05) between Treg-depleted the primary cell nucleofector solution (V4XP-3032, Lonza) and mixed
mice when compared to wild-type and CD4-depleted mice that were with the CAS9–RNP mixture and transferred to the 4D-Nucleofector
associated with effector differentiation and analysed the top ten ligand– 96-well shuttle. Cells were electroporated using a mouse unstimulated
receptor pairs. Scoring of the predicted ligand–receptor pairs based on T cell programme in the 4D-Nucleofector unit. Cells were allowed to
the NicheNet vignette Pearson correlation analysis43. NicheNet scores rest for 30 min in complete RPMI at 37 °C followed by adoptive transfer
are provided in Supplementary Table 3. via tail vein intravenous injection.

HSC expansion and bone marrow chimeras Quantification and statistical analyses
Chimeras for knockout and overexpression functional experiments Statistical analysis was performed with Prism (v9.0, GraphPad).
were made using a haematopoietic stem cell bone marrow system Two-tailed unpaired Mann–Whitney U tests, One-way ANOVA with
(Supple­mentary Data 1c). Femurs, tibias and hips were isolated from Tukey’s multiple-comparison test or Kruskal–Wallis test with Dunn’s
donor mice. Bones were cleaned and bone marrow was extracted multiple-comparison tests were used when appropriate and as indi-
by flushing the bones with a syringe and RPMI with 2% FBS. Red blood cated in each figure legend for human and mouse data. For survival
cells in the cell suspension were then lysed using ACK lysis buffer and analysis, follow up time was calculated as the number of days from the
cells were surface stained in FACS buffer for 30 min with antibodies date of surgery to an event (disease progression or death) or to cen-
listed in Supplementary Table 1. Lin−SCA1+cKIT+ (LSK) cells were sorted sorship. Patients who had not progressed or were not deceased were
and plated in fibronectin coated plates (R&D and Corning). Sorted censored, and the number of days is calculated from the date of surgery
donor HSCs were expanded for 2- to 3-weeks in albumin-free culture to 5 January 2023. The TH1 high and low patient groups were stratified
F12 medium supplemented with 1% Insulin-Transferrin-Selenium- based on the optimal cut-off value. Investigators were not blinded
Ethanolamine (Thermo), 1% penicillin-streptomycin-gentamycin during outcome assessment. Statistical significance was defined as
(Thermo), 10 mM HEPES (Thermo), 100 ng ml−1 mouse thrombopoi- P < 0.05 and individual P values are listed for each summary graph.
etin (Pepro, Fugifilm), 10 ng ml−1 mouse SCF (Pepro), and 0.1% polyvinyl
alcohol as previously described44. Donor bone marrow HSCs varied Reporting summary
according to the experiment from the following mice: iCAS9 × P14 or Further information on research design is available in the Nature Port-
rtTA × SMARTA. folio Reporting Summary linked to this article.
 Acknowledgements This work was supported by the James M. Cox Foundation, and
Data availability J. C. Kennedy, funding from the Prostate Cancer Foundation and pilot funding from the
Winship Cancer Institute supported by the Dunwoody Country Club Senior Men’s Association;
Raw fastq files and associated scRNA sequencing have been uploaded H.T.K. was supported by the Cancer Research Institute Lloyd J. Old STAR program. NCI grants
to the NCBI Gene Expression Omnibus (GEO) database under identifier 1R01CA280069 (to H.T.K.) and U01-CA113913 (to M.G.S.); DOD grants W81XWH-20-1-0525
(to H.T.K. and V.A.M.) and HT9425-23-1-0318 (H.T.K.). H.T.K. was supported by ARPA-H funding
GSE274801. Other relevant data are available from the corresponding (1AY1AX000001) C.B.M. was supported by the Howard Hughes Medical Institute Hanna
author upon reasonable request. Source data are provided with this H. Gray Fellowship (GT16001), the Burroughs Wellcome Fund – PDEP Fellowship (1022362)
paper. and the Cancer Research Institute Bristol Myers Squibb Fellowship (CRI4061). The authors
acknowledge the Yerkes NHP Genomics Core (P51 OD011132, NIH S10 OD026799) and the
Emory Flow Cytometry Core (UL1TR002378). All Tetramers were acquired from the NIH
tetramer core, supported by contract 75N93020D0000.
39. Matloubian, M., Concepcion, R. J. & Ahmed, R. CD4+ T cells are required to sustain
CD8+ cytotoxic T-cell responses during chronic viral infection. J. Virol. 68, 8056–8063 Author contributions M.A.C. and H.T.K. designed the study, analysed data and composed
(1994). the manuscript. M.A.C. conducted experiments with support from N.P., E.S., P.G. and R.M.V.
40. Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587.e29 R.G., L.D. and P.G. collected human tissue samples and organized clinical data. M.A.B., S.J.,
(2021). V.N., V.A.M. and M.G.S. provided clinical samples. N.P., E.S., C.B.M. and R.M.V, provided critical
41. Jones, M. G., Rosen, Y. & Yosef, N. Interactive, integrated analysis of single-cell expertise and assisted in result interpretation. All authors reviewed the manuscript.
transcriptomic and phylogenetic data with PhyloVision. Cell Rep. Methods 2, 100200
(2022). Competing interests The authors declare no competing interests.
42. Bolotin, D. A. et al. MiXCR: software for comprehensive adaptive immunity profiling. Nat.
Methods 12, 380–381 (2015). Additional information
43. Browaeys, R., Saelens, W. & Saeys, Y. NicheNet: modeling intercellular communication by Supplementary information The online version contains supplementary material available at
linking ligands to target genes. Nat. Methods 17, 159–162 (2020). https://doi.org/10.1038/s41586-024-08076-7.
44. Wilkinson, A. C. et al. Long-term ex vivo haematopoietic-stem-cell expansion allows Correspondence and requests for materials should be addressed to Haydn T. Kissick.
nonconditioned transplantation. Nature 571, 117–121 (2019). Peer review information Nature thanks the anonymous reviewer(s) for their contribution to the
45. Labun, K. et al. CHOPCHOP v3: expanding the CRISPR web toolbox beyond genome peer review of this work.
editing. Nucleic Acids Res. 47, W171–W174 (2019). Reprints and permissions information is available at http://www.nature.com/reprints.
Article
a) Patient Cluster b)
1 PDCD1 CD69 BATF JUNB TOX CXCR3 CXCR6 ICOS

All Clusters
Distribution

100

2 3

Proportion (%)
75
IL2RA IKZF2 IL1R1 CTLA4 TNFRSF18 CCR8 ENTPD1 TIGIT

Cluster 1
12
50
9
6
25
3
0
0 GZMK FASLG IFNG CCL4 GZMA APOBEC3G NKG7 CD38 Normalized

Cluster 2
Expression
Pt1 Pt2

● TCF1+ ● EOMES+ ● Tregs

DPP4 IL7R GPR183 CD40LG KLF6 CD226 BACH2 ANXA1

Cluster 3

<2e-16 0.1548 <2e-16 <2e-16

c) Treg signature
1.5
<2e-16 <2e-16 TH1 signature 0.278 0.888 Cell cycle signature <2e-16 0.0002 TFH signature <2e-16 <2e-16
1.5 1 1.5

Enrichment score
Enrichment score

Enrichment score
Enrichment score

Enrichment score
Max 1
0.5 0.5
0.5 0.5

0 0
0 0
0
TCF1+ Treg Eomes TCF1+ Treg Eomes TCF1+ Treg Eomes TCF1+ Treg Eomes

d) e) f) ● TCF1+ ● TH1-like ● Tregs ● TH1 ● TFH g) Gated on PD1+ CD45RA-

(EOMES+) (TBET+) (BCL6+)
2 100 100 100 100
10 16.1 8.44 0.0013
PD1+ CD4 TILS (%)

KI67+ (% of population)
CD4+ (% of total)

1 80 80
10 80 80
% of PD1+ CD45RA-

% of PD1+ CD45RA-
3

KI67 <BV650>
TUMOR

10
0
10 60 60
60 60
-1
10 40 40
-2
40 10
2 40
10
20 20 0
-3 20 20
10 30.5 44.9
0 0
-4 0 102 103 104 0
10
K B P K B P Bladder cancer Prostate cancer TCF-1 <AF488> TCF1+ CD39+

h) Gated on PD1+ CD45RA- CD4 T cells - TUMOR l) Gated on PD1+ CD45RA- CD4 T cells - TDLNs
<0.001
<0.001 0.002 0.002 <0.001 5
3 10 0.0042
4
15.6 49.7 5.97 19.4 21.1 19.3 1.80 31.7 15.9 8.34
10 0.001
3
4 4 10 100 10
10
10 <0.001 4 <0.001 0.0950
10 4
0.001
4
4
GZMK <Pecy7>

CD39 <BV421>

CD39 <BV421>
CD26 <Pecy5>

CD26 <Pecy5>

10 10
10

CD26 MFI
GZMK MFI

3 2
3
3

CD39 MFI
3
10
10
CD39 MFI
CD26 MFI

10
10
10
50 3
2 3 2 3 10
10 10
10 10
2 1 2
10
2 2 0 10 10 10
10 0
0 0 1
0 1 0 10 <0.001
21.2 13.5 1 63.7 11.0 57.3 2.27 6.84 59.6 10 74.6 1.15
10
10 3 10 4 0 102 103 0 10 3 102 103 104 0 103 104

<0.001 3
4 <0.001 3 <0.001 0.030
14.8 51.0 10 3.01 15.6
10 <0.001
9.46 17.8 4 5.06 53.8 3 0.0008 0.88 1.74 10 0.0140
10 10 0.094
CD127 <BV711>

CD127 <BV711>

3 3
<0.001
GZMB <AF700>

0.017
CCR8 <BB700>

CCR8 <BB700>
3 10 10
2 3 0.020

CCR8 MFI
CD127 MFI

10
3 10 10 3 3
CCR8 MFI

10
CD127 MFI

10 10
GZMB MFI

2
2 1 2
2 2 10
2 2 10 10 10
10 10
10 10 0 2 2
0
10 1 10 10
1
0 10 0
0 1
10 1 10
24.3 9.89 65.0 16.3 69.3 3.45 10.9 30.2 10 82.3 15.0

103 104 0 102 103 0 10 3 102 103 104 0 103 104

FOXP3 <PE>
0.043 <0.001
31.6 57.5
<0.001 1.58 10.5 4 0.029 12.6 19.1
<0.001
4 10 4 4 4.37 83.4 3 <0.001 0.17 2.03 0.0001
4
10 <0.001 10
10 200
CTLA4 <BV786>

0.021 10 <0.001
CD226 <Pecy7>

CD226 <Pecy7>

GZMB <AF700>

10
3
Perforin <PE>

3 3 10
3
CD226 MFI

150
CD226 MFI

3
3
CTLA4 MFI

3 10
10 10 10
Perforin MFI

GZMB MFI
10
10 3
2 10 2 100
2 10 10
2 2 10 2
10
10 2 10
1 2 50
10
0 10 0.005 10

1 0 1 0 0 1 0
0
8.61 2.30 10 80.2 7.67
10 66.7 1.63 3.79 8.45
10 87.7 10.1

103 104 0 103 0 10 3 102 103 104 0 103

TCF1 <AF488>
0.0009 0.0146 <0.001 4 0.059 4 0.1955
3 39.3 60.6 4.53 13.9 25.0 19.3 26.4 11.9 10 88.0 12.0 10
TIGIT <PE-Dazzle>

10 4 4 3
4
10 0.0001 10 0.012
10
10
CD28 <BV510>

CD38 <BV650>

CD38 <BV650>

CD28 <BV510>

<0.0001 3 4
CD28 MFI

10 10
CD38 MFI

3
10 3 3
2 3
TIGIT MFI

10 10
CD38 MFI

2
CD28 MFI

10 2 10 3 10
3 2 10 10
3 10
10 10 1
2
10 2
0 10
1 0 0.025 10 0
2
2 0
10 0
0
10
0.078 0.013 10 73.9 7.62 53.7 1.36 10 54.1 7.58 0 0.012

10 2 10 3 104 0 103 0 10 3 0 103 0 103

TCF1 <AF488> EOMES <eFluor660> FOXP3 <PE> BCL6 <PE-CF594> EOMES <eFluor660>

i) Gated on CD4 T cells j) Gated on PD1+ CD45RA- CD4 T cells - TDLNs k) ● TCF1+ ● TH1-like ● Tregs ● TH1 ● TFH
RCC Non Metastaic TDLNs FOXP3 <PE> EOMES <eFluor660> TBET <BV605> BCL6 <PE-CF594> (EOMES+) (TBET+) (BCL6+)
100 <0.0001
100 4
4 28.5 1.88 4 7.83 15.3 1.89 11.2 2.04 9.39 0.25 19.2 10 0.0200
% of PD1+ CD45RA-

10 10
PD1+ CD4 s(%)

3 80 0.0004
Transcription Factor

3
10
80 10 3
PD1 <UV737>

10
60
PD1 MFI

3 3
10 60 10

40 3
2 2 10
40 2
10 10
0 0
10 0 20
0
20 0
32.1 37.5 5.10 71.8 9.84 77.0 13.3 75.2 15.1 65.5 0
0 2
0 103 104 102 103 104 102 103 104 102 103 104 102 103 104 10
K TDLNs TCF-1 <AF488>
CD45RA <BV786>

Extended Data Fig. 1 | See next page for caption.

Extended Data Fig. 1 | Phenotypic characterization of PD1 + CD45RA- CD4 Mean ± s.d. are represented in each summary plot and were analyzed by
T cells in kidney, bladder and prostate cancer patients. a) ScRNAseq analysis Kruskal-Wallis test with Dunn’s multiple-comparison tests (n = 14–125 kidney
of PD1 + CD45RA- CD4 T cells in kidney cancer patients (n = 2). Bar plot showing cancer patients for each marker). i) Quantification of PD1 + CD4 T cell
cluster distribution between patients. b) UMAP projections of selected genes infiltration in kidney tumor draining lymph nodes, shown as the percent of
across all clusters. c) VISION GSEA using defined human CD4 T cell lineage total CD4 T cells (n = 12). j-k) Representative lineage transcription factor and
(Treg, Th1, Tfh) and proliferation (cell cycle) signatures. UMAP projections PD1 expression in PD1 + CD45RA- CD4 T cell populations in kidney tumor
show the top 10% scoring cells and the enrichment score for the signature is draining lymph nodes (n = 9–12). Mean ± s.d. are represented in each summary
represented as violin plots for each cluster. Means shown for every violin plot plot and were analyzed by Kruskal-Wallis test with Dunn’s multiple-comparison
and were analyzed by one-way ANOVA with Tukey’s multiple-comparison tests. tests. l) Phenotype of activated CD4 T cells in kidney tumor draining lymph
d-e) Total quantification (d) and frequency of PD1+ (e) CD4 T cell infiltration in nodes (n = 5–12 patients per marker). Mean ± s.d. are represented in each
kidney (K n = 125), prostate (P n = 6) and bladder (B n = 17) tumors. f) Frequency summary plot and were analyzed by Kruskal-Wallis test with Dunn’s multiple-
of activated CD4 T cell populations based on transcription factor expression in comparison tests. Each point represents an individual patient. The activated
human bladder (n = 17) and prostate (n = 6) tumors. g) Representative TCF1 and populations were defined based on the following gates: Tregs: CD28 + FOXP3 +
KI67 expression in activated CD4 T cells infiltrating human kidney tumors TBET- EOMES-, EOMES: CD28 + FOXP3- TBET-EOMES +, TFH: CD28 + FOXP3-
(n = 21). Mean ± 95% CI are represented and were analyzed by two-sided TBET-EOMES-TCF1 + BCL6 +, Th1: CD28- FOXP3-TBET +, TCF1+lin-: CD28 +
unpaired Mann Whitney U test. h) Representative plots of various phenotypic FOXP3-TBET-EOMES-TCF1 + BCL6-.
markers expressed in activated CD4 T cells in human kidney tumors.
Article
a) Sorted TCF1+ (CD26+ CD127+) TILS Sorted Effectors (CD26- CD127- CD39+) TILS b) <0.001 c) 100
<0.001 0.142 <0.0001 0.9048
100 100
Unstimulated Th0 Stimulation Unstimulated Th0 Stimulation

cells undergoing divison

4 4 <0.001 0.856 80 80 80
10 0 0 10 14.6 0.46 0 15.9 0 36.6 100

% of original

GZMB+ (%)

CD25+ (%)
FOXP3 <PE> 60 60 60
3 3 80

FOXP3+ (%)
3 3
10 10 10 10
60 40 40 40
2 2
10 10 20
40 20 20
2 2
10 10
1 1 20 0 0 0
10 10 0 0
0.48 99.5 76.0 8.93 0 84.1 3.47 59.9 U S U S U S U S
0 TCF1+ CD4s - Th0 Stim
0 103 104 105 0 103 104 105 0 103 104 105 0 103 104 105
U S U S Th1 Stimulation Treg Stimulation
CD39+ CD4s - Th0 Stim
<0.001 0.479
0 0.97 24.8 2.00 0 2.24 0 1.78 100
TBET <BV605>

80
3 3
10
3
10
3
10
f) Sorted TCF1+ Sorted Effectors

Tbet+ (%)
10
60 (CD26+ CD127+) TILS (CD26- CD127- CD39+) TILS
40 100 100
2 2 2 2
10 10 10 10
20

TCF1+ (% in condition)

TCF1+ (% in condition)
0 0 0 0 80 80
0.97 98.1 62.8 10.3 2.88 94.9 5.92 92.3
0
0 103 104 105 0 103 104 105 0 10 3 10 4 10 5 0 103 104 105 U S U S 60 60

d) Unstimulated EOMES Stimulation Unstimulated EOMES Stimulation 40 40

<0.001 0.533
100
EOMES <eFluor660>

0 2.66 38.9 1.28 0 22.7 0.65 27.9

3 3
10
3
10
3 20 20
10 10
80

EOMES+ (%)
0 0
60
2 2 2 2 100
10 10 10 10 40 100
0 0 0 0

Tbet+ (% in condition)

Tbet+ (% in condition)
20 80 80
0.36 97.0 55.7 4.15 0.32 77.0 2.60 68.8
0 60
0 10 3 10 4 10 5 0 103 104 105 0 103 104 105 0 103 104 105 U S U S 60

4 0
10
0.28 4
10 18.3 0.29 4
10 1.27 15.2 4
10 1.54 12.3 100 <0.001 0.125 40 40
FOXP3 <PE>

80 20

FOXP3+ (%)
3 3 3
10
3 20
10 10 10
60
2 2 2 2 0 0
10
10 10 10 40
100
100
1 1 1 1 20

FOXP3+ (% in condition)
10

FOXP3+ (% in condition)
10 10 10
20.5 79.2 73.1 8.27 0 83.5 12.3 73.8 80
0 80
0 10 3 10 4 10 5 0 103 104 105 0 103 104 105 0 103 104 105 U S U S
60 60
e) Unstimulated TFH Stimulation Unstimulated TFH Stimulation
100 0.071 0.881 40 40
0 0.20 54.6 4.03 0 0 0 0
BCL6 <PE-CF594>

3 3 3 3
10 10 10 10
80 20 20
BCL6+ (%) 60
0 0
2
10
2 2 2 40
10 10 10

20 100 100

EOMES+ (% in condition)

EOMES+ (% in condition)
2.62 97.2 24.0 17.5 0 100 2.50 97.5
0 80 80
0 103 104 0 103 104 0 103 104 0 103 104 U S U S

0 0 5.71 0.27 0 0 0 0.60 100 0.071 0.392 60 60

TBET <BV605>

4 4 4 4
10 10 10 10
80 40 40
TBET+ (%)

3 3 3 3
10 10 60 20
10 10 20

2
40
2
10
2
10 10
2
10 0 0
0 0 0 0
20

● ● ●Th1 Stimulation
3.22 96.8 77.5 16.5 0 100 0 99.4
0 Unstim Th0 Stimulation
●Tregs Stimulation ● EOMES Stimulation ●TFH Stimulation
0 103 104 0 103 104 0 103 104 0 103 104 U S U S
CTV

1:1 DC:TCF1+ CD4s

g) 1:1 DC:TCF1+ CD4s 1:1 DC:CD39+ CD4s 1:1 DC:CD39+ CD4s
h) 1:1 DC:TCF1+ CD4s 1:1 DC:CD39+ CD4s 1:1 DC:TCF1+ CD4s 1:1 DC:CD39+ CD4s i)
0.0061
64.7 29.6 0.68 0.68 100 4.19 1.65 4.79 35.6 100 0.0061 100 0.5273 50 0.0061

cells undergoing divison

4 4 4 104
CD26 <Pecy5>

10 10 80 10 80 80 40
FOXP3 <PE>

%EOMES+

% of original
% FOXP3+
%CD26+

60 3
60 60 30
3 3 10 103
10 10
40 40 40 20
0 0
0
20 0 20 20 10
2.68 3.00 15.1 83.6 62.0 32.2 8.90 50.7
0 0 0 0
0 10 3 10 4 0 10 3 10 4 0 10 3 10 4 0 10 3 10 4
CTV CTV

1:1 DC:TCF1+ CD4s 1:1 DC:CD39+ CD4s 1:1 DC:TCF1+ CD4s 1:1 DC:CD39+ CD4s
j) 1:1 DC:TCF1+ CD4s 1:1 DC:TCF1+ CD4s 1:1 DC:TCF1+ CD4s
10 ng/mL IL-12 +IL-12
4 31.9 1.03 4 0.63 1.10 100 0.0242 800 4 40.7 1.29 2500
10 10 10
0.006 0.006 0.012 0.024 0.023
Geometric Mean TBET
Geometric Mean IL7R
TBET <BV605>

TBET <BV605>

80 2000
600
3 3 0.0061
%TBET

3
10 10 60 0.00830.0061 10 1500
400 0.0167
40 1000
2 2 2
10 10 10
200
0 0 20 0 500
38.5 28.6 8.98 89.3 25.8 32.2
0 0 0
103 104 103 104 10 3 10 4
0 0
0 1 2 3 4 5+ 0
0 1 2 3 4 5
CTV Division CTV Division

k) PD1+ CD45RA- CD4 T cells l) Low IL-2 (10 U/ml) High IL-2 (4000 U/ml) m) Clonotypes from Eomes+ cluster
Clonotypes from Treg cluster
n) Number of shared clonotypes
CTV label and FACS Sort Patient 1 Patient 2
4.72 35.4 16.6 23.8 100 0.0571
4 4
FOXP3 <PE>

10 10 80 TCF1+ NA 17 12 NA 17 13
TCF1+
Tregs/Eomes
(CD26- CD127- CD39+) 60
% Divided

3 3
10 10
40 EOMES NA 16 EOMES NA 7

5 days in culture
0 0 20
Bead Stimulation 8.66 51.2 29.8 29.8 Tregs NA Tregs NA
0
0 10 3 10 4 0 10 3 10 4
Low High
CTV 41 22 20 TCF1+ EOMES Tregs TCF1+ EOMES Tregs
10 U/ml IL-2 4000 U/ml IL-2
Patient 1
Top 3 TCR clonotypes

Extended Data Fig. 2 | See next page for caption.

Extended Data Fig. 2 | Tumor PD1 + TCF1+lin- CD4 T cells retain proliferation g-h) Representative plots of various markers on sorted TCF1+ lin- or CD39 +
and differentiation capacity. Representative CTV staining and expression CD4 T cells after 5-days of 1:1 co-culture with patient matched dendritic cells
of selected differentiation markers after five days of culturing TCF1+ lin- and (n = 7 patients for TCF1+ lin- and n = 4 patients for CD39+). Medians are shown
CD39 + CD4 T cells in unstimulated (U) or a) Th0 (green, TCF1+ n = 11, CD39+ and analyzed by two-sided unpaired Mann–Whitney U test. i) Frequency of
n = 8) d) Eomes (red, TCF1+ n = 9, CD39+ n = 6) and e) TFH (pink, TCF1+ n = 6, original plated cells undergoing division for TCF1+ or CD39 + CD4 T cells after
CD39+ n = 6) stimulating conditions. Summary plots show the frequency of DC co-cultures. Medians are shown and analyzed by two-sided unpaired
cells within each population positive for the indicated marker on day 5. Each Mann–Whitney U test. j) Representative plot of TBET expression on TCF1+
point represents an individual patient. Multiple conditions were performed lin- CD4 T cells in DC co-cultures with exogenous IL-12 (no IL12 n = 7, IL12 n = 4).
with the same patient depending on cell sorting numbers. Median ± 95% CI are Mean ± s.d. are shown and analyzed by two-sided unpaired Mann–Whitney
represented and analyzed by two-sided unpaired Mann–Whitney U test. U vs. U test between each division. k) Experimental design to test the capacity of
S within the same population were compared across conditions. b) Frequency PD1 + CD39 + CD4 T cells to proliferate with 4000U/ml of exogenous IL-2. l)
of original plated cells undergoing division for TCF1+ or CD39 + CD4 T cells in Proliferation and FOXP3 expression on sorted CD39 + CD4 T cells after 5-days
Th0 conditions. Median is represented and analyzed by two-sided unpaired of CD3/CD2/CD28 bead stimulation with 10U/ml (low, n = 3) or 4000U/ml
Mann-Whitney U test. c) Frequency of TCF1+ or CD39 + CD4 T cells expressing (high, n = 4) of exogenous IL-2. Medians are shown and analyzed by two-sided
the indicated marker on day 5 under Th1 (TCF1+ n = 13, CD39+ n = 2) or Treg unpaired Mann–Whitney U test. m) UMAP projections of the distribution
stimulation (TCF1+ n = 10, CD39+ n = 5). Median ± 95% CI are represented and of TCR clonotypes corresponding to cells in the Treg (blue) or Eomes (red)
analyzed by two-sided unpaired Mann–Whitney U test. f) Summary plots clusters. Summary bar graph shows the cluster distribution of the 3 most
represent the frequency of cells within each sorted population (TCF1+ lin- or dominant TCR clonotypes across all clusters for one patient. The number of
CD39+) expressing the indicated transcription factor after 5 days of culture cells sharing the respective TCR clonotype is indicated below. n) Correlation
in each of the respective conditions tested. Cytokine stimulation conditions matrix showing the number of clonotypes shared by each population in the
were the following: Th0 (IL-2), Th1 (IL-2 and IL-12), Treg (IL-2, TGF-beta, and tumor for both patients.
anti-IFN gamma), Tfh (Activin A, IL-12), or Eomes (IL2, IL-12, IL-4, aIFNg).
Article
a) Gated on PD1+ CD44+ CD4 T cells - TDLN - 1-week TRAMPC1-GP
TCF1+ Tregs
Gated on CD4 T cells FOXP3 <PE-e610> TBET <Pecy7> BCL6 <BV421>

GP66+ Ki67+ (% with phenotype)

6 100 100
10

W1 TRAMPC1-GP
0.25 7.46 2.03 1.02 1.36 0.44 8.06

% of GP66+ CD4 T cells

4

Transcription Factor
10
4 5 4 80 80

GP66 <APC>
10
10 10

GP66+ (Log10)
3
10
4 3 60 60
10
3 10 3
10
10
3 40 40
10 10
2

0 2 0
0
0 20 20
10
5.42 85.1 11.9 85.8 15.5 76.5
0 0
10 3 10 4 10 5 103 104 103 104 103 104
0
W1 W2
0 0 0
0 1 2 3 4 5

Tr +
s
H
1
CD44 <UV737> TCF1 <PE>

eg

Th
Weeks post tumor inoculation

F1

TF
TC
b) Gated on GP66+ CD44+ CD4 T cells - TDLN - 5-week TRAMPC1-GP
EOMES <Percp-eFluor710> RORgT <BV421> TOX <PE> CXCR3 <FITC> ICOS <BV711> IL2RA <UV395> IL7R <PE> CD39<SB702>
5
0 0.40 0.75 2.16 3.96 70.3 4 16.0 35.3 4 47.6 30.6 4 4.80 24.7 23.2 1.48 4 12.0
0 15.2
2
W5 TRAMPC1-GP

10 4 10 4
10 10 10 10 10
4
10
4
10
3 3 3 3
3 10 3 10
10 10 10 10
3
3
10 10
Molecule

Molecule
0 0 0 0
0 0 0
0
18.2 81.4 15.4 81.7 9.23 16.5 7.91 40.8 20.7 1.11 63.5 7.01 65.3 9.96 70.0
7 0.0
0 2.76
6

0 10 3 10 4 0 10 3 10 4 0 10 3 10 4 10 5 0 10 3 10 4 0 10 3 10 4 0 10 3 10 4 0 10 3 10 4 10 5 0 10 3 10 4

TCF1 <PE> FOXP3 <PE-e610>

c) Gated on GP66+ CD44+ CD4 T cells - TUMOR - 5-week TRAMPC1-GP

PD1 <BV786> KI67 <BV650> TOX <PE> RORgT <BV421> FR4 <Percp-cy5.5> SLAMF6 <BV605> CD39 <SB702> PSGL1 <Percp-e710>
5 5 5
4 64.0 25.0 10 30.9 9.88 4 17.9 29.5 10 13.0 3.00 68.4 18.9 17.0 26.0 68.0 2.00 10 73.0 19.0
10 10
4 4 4
10 10 10
4 4 4
10 10 10

3 3
10 10 3
3 3 10
3 3 10 10 3
10 10 10
Molecule

0 0 0
0 0 0 0
0

10.0 1.00 43.2 16.0 36.8 15.8 61.0 23.0 9.47 3.16 57.0 0 6.00 24.0 1.00 7.00

0 103 104 105 0 103 104 105 0 103 104 105 0 103 104 105 0 103 104 105 0 103 104 105 0 103 104 105 0 103 104 105

TCF1 <PE>

d) e)
Gated on PD1+ CD44+ CD4 T cells - TDLN - 5-week TRAMPC1-GP
Gated on CD4 T cells FOXP3 <PE-e610> TBET <Pecy7> BCL6 <BV421> 5
6

FOXP3+ (% of CD44+ PD1+)

10 100

Total # of population (Log10)

10
W5 TRAMPC1-GP

104
12.3 8.26 20.7 24.0 0.72 2.13 104 0.87 29.2
CD44+ PD1+ (Log10)

80
Transcription Factor
CD44 <UV737>

4 5 104
10
10
103

103 4 60
3 4 10
10 10 103
40
102
3
10 0
0
0
20
0
78.5 0.89 3.11 52.2 23.1 74.1 24.6 43.7 3
2 10
10 0
0 103 104 103 104 103 104 103 104
TDLN Tumor
0 0 0
0 1 2 3 4 5 TDLN LN Spleen
PD1 <BV786> TCF1 <PE>
Weeks post tumor inoculation LCMV Arm.

f) g)
Gated on GP66+ CD44+ CD4 T cells - TDLN - D12 B16-GP
Gated on CD4 T cells 5 FOXP3 <PE-e610> TBET <Pecy7> BCL6 <BV421>
10 100

% of CD44+ PD1+ CD4 T cells

100 100
5
0.083 10.4 15.6 0 12.6 0 8.33
% of GP66+ CD4 T cells
10

Ki67+ (% of GP66+)
Transcription Factor

4 80
GP66+ (Log10)

10 10
4
10
4
80 80
GP66 <APC>

4
B16F10-GP

10 3
10

3 60 60 60
3 10 10
3
10
3
10
40 40 40
2
10 0
0
0 0
20 20 20
1 1.48 72.6 11.1 76.3 18.8 72.9
10 0 0 0
0 10 3 10 4 10 5 0 10 3 10 4 0 10 3 10 4 0 10 3 10 4
D12 TDLN

Tr +
s
H

1
Tr +
s
H
1

TCF1 <PE>

eg

Th
eg

F1
TH
F1

TF
TF

TC
CD44 <UV737>
TC

h)
Gated on CD44+ PD1+ CD4 T cells - TDLN - D14 MC38 i) Gated on GP66+ PD1+
Gated on CD4 T cells FOXP3 <PE-e610> TBET <Pecy7> BCL6 <BV421> CD4 T cells LCMV Arm D8
6
10
% of CD44+ PD1+ CD4 T cells

100 100
13.4 9.05 22.7 31.9 0.90 1.98 0.18 25.6 4.68 34.2

% of GP66+ CD4 T cells

CD44+ PD1+ (Log10)

5
Transcription Factor

BCL6 <BV421>

10
CD44 <UV737>

10
4
10
4
80 80
3 3
4 10 3 10
MC38

10 10
60 60
3 3
10
3 10
10 40 40
2
10
2 0
0
0 10 0 0 20 20
77.0 0.60 1.18 44.2 23.4 73.7 23.7 50.5 44.1 17.0
1 0 0
10
0 103 104 0 103 104 0 103 104 0 103 104 0 10 3 10 4
TDLN Tumor
Tr +
s
H
1

Tr +
s
H
1
eg
eg

TCF1 <PE>

Th
Th

F1
F1

TF
TF

PD1 <BV786> TCF1 <PE>

TC
TC

j) Gated on CD44+ PD1+ CD4 T cells - TDLN - D21-28 RENCA-HA Gated on PD1+ CD4 T cells
Gated on CD4 T cells FOXP3 <PE-e610> TBET <Pecy7> CXCR5 <PE> RENCA-HA Tumor
6 100
% of CD44+ PD1+ CD4 T cells
Total # of population (Log10)

% of CD44+ PD1+ CD4 T cells

10 100
FOXP3 <PE-e610>

5 5
10 9.66 5.69 16.9 39.1 2.30 0.46 10 0.29 13.4 81.5 0.99
80
CD44 <UV737>

10
4 10
3
10
4 80
5
RENCA-HA

4 4
10 10 10
60
Molecule

60
3 3
10 3 10
10
3
4 10
2 10 40 40
10 0

0
0 0
20 0 20
84.1 0.54 2.01 42.0 33.4 63.4 18.9 67.3 4.69 12.82
3 0
10 0
0 103 104 0 103 104 105 0 103 104 0 103 104 105 0 103 104
TDLN Tumor
PD1 <BV786>
Tr +
s
H
1

TCF1 <PE>
Tr +
s
H
1

eg
eg

TH
TH

F1
F1

TCF1 <PE>
TF
TF

TC
TC

k) Gated on CD44+ PD1+ CD4 T cells - TUMOR- Orthotopic RENCA-HA-Luciferase D15

Gated on CD4 T cells FOXP3 <PE-e610> SLAMF6 <BV605> CD73 <AF700> KI67 <BV650> CD25 <UV395>
6
10
Total PD1+ CD4T cells/

5 5 5 4 5
75.3 1.55 34.7 9.73 13.4 6.86 56.4 6.82 64.9 2.45
10
41.9 10 10 10 10
RENCA-HA-Luc

g of tumor (Log10)
CD44 <UV737>

5 4
4 10 4 4 4
10
10 10 10 10
3
4 10

3 10 3 3 3 10
3
10 10 10 10
Molecule

3 0
0
10 0 0 0 0

15.5 7.60 53.2 2.37 74.5 5.24 31.5 5.28 23.0 9.65
2
10 10 3 10 4 10 5 10 3 10 4 10 5 10 3 10 4 10 5 10 3 10 4 10 5
0 103 104 0 103 104 0 0 0 0
Tumor
PD1 <BV786> TCF1 <AF488>

Extended Data Fig. 3 | See next page for caption.

Extended Data Fig. 3 | Tumor specific CD4 T cells activate in TDLNs and experiments (n = 5–7 mice per group for each timepoint). f-g) Phenotype of
rapidly acquire a TCF1+lin phenotype. a) Representative I-AbGP66 tetramer GP66+ (f) or bulk PD1+ (g) CD4 T cells 12-days after B16-GP inoculation in
staining and phenotype of GP66 + CD4 T cells in TDLNs one-week after TDLNs. Data are representative of 2 independent experiments (n = 10).
TRAMPC1-GP inoculation. b-c) Phenotypic characterization of GP66 + CD4 h) Representative CD44 and PD1 staining, and phenotype of bulk PD1 + CD4
T cells in TDLNs (b) or tumors (c) of 5-week TRAMPC1-GP bearing mice. Data are T cells in TDLNs of day 14 MC38 bearing mice. Data are representative of
representative of 3–5 independent experiments (n > 5 mice for each marker). 2 independent experiments (n = 11). i) Representative TCF1 and BCL6 staining
d) Representative CD44 and PD1 staining and phenotype of bulk PD1 + CD4 in GP66 + CD4 T cells on D8 LCMV Armstrong infected mice in the spleen.
T cells in TDLNs of 5-week TRAMPC1-GP bearing mice. Summary plot shows Summary plot shows the frequency of virus specific CD4 T cells in each
the total numbers of activated (CD44 + PD1 +) CD4 T cell populations in each phenotype. Data are representative of 2 independent experiments (n = 10).
tissue. Kinetics plot shows the total number of activated CD4 T cells in each j) Phenotype of endogenous CD44 + PD1 + CD4 T cells in subcutaneous D21–28
phenotype within TDLNs (n > = 5 mice for each timepoint for each individual RENCA-HA TDLNs and tumor (n = 8–12 mice). k) Phenotype of PD1 + CD4 T cells
experiment). e) Frequency of PD1 + CD4 T cells expressing FOXP3 in TDLNs in orthotopic RENCA-HA-Luciferase in tumors 15-days after orthotopic implant
1-week after tumor inoculation or secondary lymphoid tissues 8-days after (n = 9 mice).
LCMV Armstrong infection. Data are representative of 2 independent
Article
a) c) W1 W5 W5 TCF1+ Tregs TH1 TFH

CD44+ PD1+ (% of SMARTAs)

TRAMPC1-GP 4 0.079 0.72 4 17.2 11.8 4 0.068 0.61 100 100
10 10 10
2.5 mil W1 W3 W5
80 80

% of SMARTAs
TDLN
3 3
3
10 10 10 60 60
40
40
2
0
10 20
500k SMARTAs 0
0 20
0.52 98.7 0.14 70.9 18.1 81.2 0
0
b) 0 103 104 0 103 104 0 103 104 1 2 3 4 5 W1 W3 W5
6
5 10 100

CD44+ PD1+ (% of SMARTAs)

10 1.53 4 0.28 0.28 4 85.9
10
6.10 4 5.56 1.49 100
10 10
5
10 80
CD44 <UV737>

SMARTAs (Log10)

FOXP3 <PE-e610>
4 80

Tumor

% of SMARTAs
10
4 3 3 3
60
10

TBET <Pecy7>
10 10 10 60
3
10 3 40
10 40
2
2 10 20
0 10 0 0
0 20
1 6.84 92.6 4.74 3.25 85.1 7.86 0
10 0
0 103 104 0 103 104 0 103 104 0 103 104
TDLN Tumor 1 2 3 4 5 W1 W3 W5
CD45.1 <UV395> TCF1 <PE>
Weeks post tumor inoculation

d) TCF1+ Tregs TH1 TFH e) RORgT <BV421> BCL6 <BV421> KI67 <BV650> f)
5 4

Total # of SMARTAs (Log10)

10 0.64 0.32 0 8.44 5
6.66 18.4 10
SMARTA population (Log10)

4 10
10

4 TRAMPC1-GP Sacrifice
4
TDLN
10
10 3
10
2.5 mil Flow cytometry
3
10
Analysis 3
TDLN

3 4 weeks 10
10
3
10
Molecule

0 0
2
10
2 15.0 84.0 17.6 73.9 8.92 66.1 D-1 10k SMARTAs 2
10 10
0 103 104 0 103 104 0 103 104
0 1 2 3 4 5 TDLN Tumor
Weeks post tumor inoculation TCF1 <PE>

g) h) KI67 <BV650> SLAMF6 <APC> SLAMF1 <BV605> i) TUMOR

Naive TDLN TUMOR 5 100 5 100
10 1.30 62.3 1.34 88.6 7.43 35.6 10 26.9 0.38

FOXP3 <PE-e610>
10
4
10
4 80 80

% of SMARTAs

% of SMARTAs
4 4
10 10
60 60
TDLN

3 3
3 10 10 3
10 40 10 40
Molecule

0 0 0 20 0 20
0 36.4 0 10.1 1.98 55.0 17.3 55.4
0 0
0 10
3
10
4 0 103 104 105 0 103 104 105 0 103 104 105 0 103 104 105

Tr F
s
1
Th H
17

17
F

1
PD1 <UV737> TCF1 <PE> TCF1 <PE>

eg

eg
Th
TC

Th
TC
TF

Th
Tr
j) D8 post LCMV Arm. infection - Gated on respective population k) Naive TBET+ TCF1- TBET- TCF1+
Spleen Liver Blood
6
10
Endogenous GP66+

46.4 21.5 72.1 9.61 36.7 16.0

Total # of GP66+ (Log10)
Endogenous GP66+

5
3 3 3
10
10 10 10
4
10
3
10
2
10
2
10
2 10
2
4.18 28.0 11.1 7.19 3.85 43.4 10
102 103 0 103 104 103
Li n

0
r
N
N
ve
e

aL

103 104 103 104 103 104

iL
le

0 0 0
GZMB <AF700> SLAMF1 <BV605> BCL6 <BV421>
Sp

56.3 18.0 69.0 13.8 49.2 20.4 5

10
Total # of SMARTAs (Log10)

SMARTAs

3 3 3 4
SMARTAs

10 10 10 10
TBET <Pecy7>

3
10
2 2 2
10 10 10 2
10
6.08 19.6 6.60 10.6 3.85 26.5 1
10 102 103 0 103 104 0 103
0 103 104 0 103 104 0 103 104
Li n

GZMB <AF700> SLAMF1 <BV605> BCL6 <BV421>

r
N
N
ve
e

aL
iL
le

TCF1 <PE>
Sp

l) n) Gated on CD4 T cells - TDLN Untx aPDL1 o) Untx aPDL1 Untx aPDL1
40
WT aPDL1 0.0023
Untx aPDL1 0.8413 0.6905 MC38
Sacrifice 105 10
0
6
Tumor Size (mm2)

23 days 10
4
10
4 30 10
CD44+ PD1+ (Log10)
GP66+ (%CD4s)
GP66 <APC>

GP66+ (Log10)

0.33 0.47
14 days
TDLN

20 0.0012
2.5 mil 10
3
10
3
4 -1 5
10 10 10
TRAMPC1-GP 10
0 0

0
103 10
-2 0 5 10 15 20 10
4
m) 0 10 3 10 4 0 10 3 10 4
Days post tumor inoculation TUMOR
CD44 <UV737>
120 Ctrl aPDL1 4 4 50 0.3095 50 0.0556 50 0.3829 100 0.1282 100 0.9452
Tbet+ (% of CD44+ PD1+)

10 10
FOXP3+ (% of CD44+ PD1+)

0.59 0.59 6.12 2.10

Tbet+ (% of CD44+ PD1+)
FOXP3+ (% of GP66+)
Tbet+ (% of GP66+)
Tumor Size (mm2)

90 TRAMPC1-GP 40 40 40 80 80
3 3
10 10
30 30 60 60
MC38

30
0.6450
TDLN

60
10
2
10
2 20 20 20 40 40
30 0 0
10 10 10 20 20
84.6 14.2 80.3 11.5
0 0 0 0 0 0
10 3 10 4 10 3 10 4
0 10 20 30 40 0 0

p)
FOXP3+ (% of CD44+ PD1+)

Days post tumor inoculation 0.1143 0.0857 0.0571

Tbet+ (% of CD44+ PD1+)

100 0.1667 100 0.6905 50 100 50

Tbet+ (% of CD44+ PD1+)

FOXP3+ (% of CD44+ PD1+)

Tbet+ (% of CD44+ PD1+)

10.7 4.78 16.9 9.01

80 80 40 80 40
RENCA-HA

3 3
10 10
60 60 30 60 30
TBET <Pecy7>
Tumor

40 40 20 40 20
2 2
10 10

0 0 20 20 10 20 10
20.2 64.3 18.1 55.9
0 0 0 0 0
0 10 3 10 4 0 10 3 10 4

FOXP3 <PE-e610>

Extended Data Fig. 4 | See next page for caption.

Extended Data Fig. 4 | TCF1+lin- CD4 T cells are stem-like cells that are Armstrong infection. Naïve (CD44- PD1-) CD4 T cells are plotted as a reference
actively restrained in the tumor response. a) Experimental design to for each marker. Data are representative of 2 independent experiments (n = 8
characterize SMARTA differentiation kinetics. b-e) Total numbers and recipient mice). l) Experimental design to test how PDL1 therapy affects stem-
representative phenotype of recovered SMARTAs in TDLNs and tumor 1- to like CD4 differentiation in TRAMPC1-GP refractory tumor model. m) Tumor
5-weeks after transfer in TRAMPC1-GP mice. Summary graphs show median or kinetics as shown by tumor diameter shown as mean ± s.d. for Untx and aPDL-1
mean ± s.d. Data are representative of 1-2 independent experiments for each treated mice and analyzed by two-sided unpaired Mann–Whitney U test (n = 5–7
timepoint (n = 4–6 mice per group for each timepoint). f) Experimental design mice per group). n) Phenotype of GP66 + CD4 T cells in TDLNs (top) and tumor
and total number of recovered SMARTAs in TDLNs and tumor 4-weeks post (bottom) in Untx or aPDL-1 treated mice 14-days after treatment in TRAMPC1-
transfer. g) Representative histogram of PD-1 expression on recovered GP bearing mice. Mean ± s.d. represented and analyzed by two-sided unpaired
SMARTAs in TDLNs and tumor as compared to naïve CD4 T cells. h-i) Phenotypic Mann–Whitney U test (n = 5–7 mice per group). o) Tumor kinetics shown as
analysis of SMARTAs in TDLNs and tumor 4-weeks post transfer (n = 9). Medians mean ± s.e.m. and phenotype of PD1 + CD4 T cells for Untx and PDL-1 treated
are shown for each summary plot. j) Representative phenotype of endogenous mice in MC38 responsive tumor model. Median shown for phenotype
GP66 + CD4 T cells (top) or transferred SMARTAs (bottom) in the indicated summaries. Statistical comparisons were analyzed by two-sided unpaired
tissue 8 days after LCMV Armstrong infection. Summary plots show median Mann–Whitney U test (n = 7 mice per group). p) Phenotype of bulk activated
of the total number of GP66+ or SMARTAs in each tissue. k) Representative CD4 T cells for untreated and PDL-1 treated mice in RENCA-HA responsive
histograms for Th1 (TBET + TCF1-) and Tfh (TBET- TCF1 + CXCR5 +) populations tumor model. Mean ± s.d. are represented and analyzed by two-sided unpaired
for the endogenous GP66+ or SMARTAs in the spleen 8 days after LCMV Mann–Whitney U test (n = 4 mice per group).
Article
a) Gated on GP66+ - TDLN - D5 post depletion 0.013
g) WT

SLAMF1+TBET+ (% of GP66+)
Untx Treg dep. 0.002 0.027
4 4 105 <0.001 100 <0.001 or B16F10-GP DT

Tbet+ GP66+ (Log10)

10 1.94 3.23 10 35.5 20.5
FOXP3-DTR 500K
104 80
TBET <Pecy7>

3 3
10 10
103 60 8 days 5 days

2 2
102 40
10 10

0 0
101 20
24.5 70.3 5.91 38.2 200K SMARTAs
100 0
3 4 3 4
0 10 10 0 10 10

x
5
17
x
5
D1
17
25

nt
D
<0.0001 0.0015
106

1
nt
D

D
50

U
D

D
U
1.35 0.23 22.9 0.81

Total SMARTAs (Log10)

Tbet+ (% of SMARTAs)
<0.001 0.065
105 100

PSGL1+TBET+ (% of GP66+)
40

TBET <Pecy7>
4.85 12.3 51.1 27.0 <0.001 105

KI67+ GP66+ (Log10)

3 3
10 10
KI67 <BV650>

4 4
80 30
104
10 10

104
60 20
103
3 3

103
10 10
0 0
40 10
102 95.9 2.48 75.4 0.92
0 0
20 0 102
18.3 64.6 9.59 12.3
0.051 0 10
3
10
4
0 10
3
10
4
WT Treg dep WT Treg dep
3 4 3 4
101 0 FOXP3 <PE-e610>
0 10 10 0 10 10

x
5
17

x
5
17
TCF1 <PE>

nt
D

nt
D
D

D
U

U
b) Gated on CD44+ PD1+ T cells - TDLN All activated clusters
h) 100 i)
Untx Treg dep. >0.999 TCF1+
Tbet+ Endo CD44+ PD1+ (Log10)

<0.001
4 4 106 106 <0.001 <0.001 Treg
10 0.58 0.97 10 33.3 16.4
75 Pdcd1 Cxcr3 12

FOXP3+ GFP+ (Log10)

Frequency
5 Th1
TBET <Pecy7>

10
105 TFH 9
104
3 3
10 10
Cycling 50
6
104 Naive
103 3
2 2
25
103
10 10
0.301
102

UMAP 2
0
24.5 73.9
0
9.67 40.6
0.361
102 101 0
3 4 3 4 UMAP 1
0 10 10 0 10 10
Untx Treg
x
5
11
17

TCF1 <PE>
x

5
17
nt
nt

D
D

dep.
D

D
D

U
U

c) Gated on CD4 T cells - Tumor - D5 post depletion

Untx Treg dep. 10
5 0.0054 <0.0001 j) Tcf7 Foxp3 Tbx21 Bcl6 Mki67 Ccr7
100
4 4 4
10 4.07 17.9 10 79.2 1.65 10
Tbet+ (% of GP66+)

80
GP66+ (Log10)

3
10
10
3
10
3
2 60
GP66+

10
1 40
2 2 10
10 10
0 20
10
0
21.1 56.9
0
19.1 0
1 - TCF1+
-1
10 0
0 10
3
10
4
0 10
3
10
4
Socs3 Slamf6 Gpr183 P2rx7 Lef1
Untx D5 Untx D5
Blood Tbet+ (% of CD44+ PD1+)

100 <0.0001 100 <0.001 0.001

Tbet+ (% of CD44+ PD1+)

4 4
10 9.40 2.78 10 75.2 2.26
80 80
3 3
10 10
60 60
TBET <Pecy7>
PD1+

2 2
40 40
10 10

0 0
20 20 2 - Tregs
33.3 54.5 17.6 4.94
0 0
Il2ra Ccr8 Klrg1 Tnfrsf9 Ctla4
3 4 3 4
0 10 10 0 10 10
x
5
15
17
x

FOXP3 <PE-e610>
nt
nt

D
D

D
D
U
U

d) WT WT
or DT or DT
B16F10-GP MC38
FOXP3-DTR FOXP3-DTR
250K 250K
3 - TH1
Sacrifice at D5
7 days 5 days 10 days and D11 Ifngr1 Cxcr6 Id2 Nkg7 Ifng

e) Gated on CD44+ PD1+ CD4 T cells - TDLN - B16F10-GP

Untx Treg dep. <0.0001 <0.0001 0.0002
100
Tbet+ (% of CD44+ PD1+)

100 0.0038 100

Tbet+ (% of CD44+ PD1+)

4 5.71 1.78 4 33.6 8.19

10 10
80 4 - TFH
Tbet+ (% of GP66+)
TBET <Pecy7>

80 80
60 Cd200 Tnfsf8 Bcl2a1b Il21 Il4
10
3
10
3
60 60
40
40 40
0 0 20 20 20
40.1 52.4 40.0 18.2
0 0 0
0 10
3
10
4
0 10
3
10
4
WT D1 D3 D5 WT D5 WT D5
FOXP3 <PE-e610>
TUMOR

f) Gated on CD44+ PD1+ CD4 T cells - TDLN - MC38 k) Mouse Precursor CD4 signature l) Stem-like TCF1+ CD4 cluster - m) Untx mice - All TCF1+ TCRs
Analysis between groups
Untx Treg dep. 0.1795 0.0569 Untx Treg dep.
0.003 0.0005
100 106 0.00002
Tbet+ (% of CD44+ PD1+)

4 1.56 38.1 0.23 32.5 2.57

10 10
4
<2e-16
TBET <Pecy7>

80 <2e-16
105
Tbet+ (Log10)

10
3
10
3
60
104
40
0 0 20 103
60.2 38.0 56.7 8.29
3 4 3 4
0 102
0 10 10 0 10 10
WT D5 D11 WT D5 D11
FOXP3 <PE-e610>

Untx - Top 10 TCRs in Treg cluster

4 4 4 3 2 2 2 2 2 2
Number of cells with clonotype

TCF1+ Tregs TFH Prolif

Extended Data Fig. 5 | See next page for caption.

Extended Data Fig. 5 | Phenotypic and transcriptional profiling of antigen Mann-Whitney U test. h) ScRNAseq analysis of bulk activated (CD44 + PD1 +)
specific CD4 T cells generated after Treg depletion in cancer. a-c) Phenotypic CD4 T cells sorted from TDLNs of Untx or Treg depleted TRAMPC1-GP mice
analysis of GP66+ and bulk PD1 + CD4 T cells in TDLNs, blood and tumor 5-days 5-days after depletion (n = 4-12 pooled mice per group). Naïve (CD44- PD1-) CD4
after Treg depletion. Median or mean ± s.e.m are shown and analyzed by two- T cells were included as a control. UMAP projection and cluster distribution of
sided unpaired Mann Whitney U or Kruskal-Wallis test with Dunn’s multiple- activated CD4 T cells in TDLNs for both groups. i-j) Normalized expression of
comparison tests when appropriate. Each individual timepoint represents transcription factors and selected genes defining each cluster. k) VISION GSEA
an independent experiment (n = 2–17 mice per group for each time point). analysis using a signature from precursor TCF1+ BCL6low CD4 T cells from
d) Schematic of experimental designs for B16F10-GP or MC38 tumors. chronic LCMV infection for Untx and Treg depleted mice. Mean is represented
e-f) Representative TBET by FOXP3 staining of bulk activated CD4 T cells in in each violin and were analyzed by one-way ANOVA with Tukey’s multiple
B16F10-GP (e) or MC38 (f) TDLNs 5-days after Treg depletion. Data represents comparison tests. l) Transcriptional comparison of stem-like CD4 T cell cluster
3-4 independent experiments (n = 4–5 mice per group for each experiment). between groups. Volcano plots show fold change versus -log(p-value) for each
Medians are shown and were analyzed by two sided unpaired Mann-Whitney gene. m) UMAP projection of the distribution of all the TCR clonotypes in the
U test or Kruskal-Wallis test with Dunn’s multiple-comparison tests when TCF1+ stem-like cluster in untreated mice (top). Cluster distribution of the
appropriate. g) Schematic of experimental design and phenotypic analysis 10 most dominant TCR clonotypes belonging to the Treg cluster in Untx mice.
of transferred SMARTAs in B16F10-GP TDLNs 5-days after Treg depletion. Data The number of cells sharing the respective TCR clonotype is indicated below
represents 2 independent experiments (n = 5 mice per group). Medians are (bottom).
represented in each summary plot and were analyzed by two-sided unpaired
Article
a) Gated on SMARTAs - TDLN c) Gated on CD44+ PD1+ CD4 T cells - TRAMPC1-GP
TDLN Stem CD4s TDLN Stem CD4s TDLN Tregs TDLN TUMOR TDLN TUMOR
WT WT WT DTR WT DTR 0.0004 4 <0.0001 4 <0.0001
100 10 10

TCF1+ lin- (% of SMARTAs)

0.075 0.211

Geometric Mean CTLA4

3.02 70.7 23.1 41.4 28.8 4.56 7.13 47.3 4.04 76.6

Geometric Mean CTLA4

4 4 4
10 10 10

80
4 4
CD226 <FITC>

10 10

CTLA4 <PE>
10
3
10
3
10
3
60 3 3
10
3
10
3
10 10
40
0 0 0
0 0
20
4.44 21.8 21.3 14.2 65.9 0.72 47.6 4.54 14.3 8.58
2 2
10 10
0 10
3
10
4
0 10
3
10
4
0 10
3
10
4
0 0 10
3
10
4
10
5
0 10
3
10
4
10
5
Non Tregs Non Tregs
TCF1 <PE> FOXP3 <PE-e610> Tregs Tregs
0.0188

TCF1+BCL6+ (% of SMARTAs+)
0 6.59 3.07 20.7 9.40 0.38
50
>0.999 0.010 d) e) Untx aCTLA4 4F10
aCTLA4 (4F10 clone) aCTLA4 (9H10 clone)
BCL6 <BV421>

10
4
10
4
10
4
40 120 aCTLA4 9H10
WT
30
Sacrifice

Tumor Size (mm2)

10
3
10
3
10
3 >0.9999
20 20 days D34 80
aCTLA4

0.0003
0.0084
0 0 0
10 14 days
6.59 86.8 28.9 47.3 89.1 1.13
40
0 10
3
10
4
10
5
0 10
3
10
4
10
5
0 10
3
10
4
10
5 0 2.5 mil
TCF1 <PE> TRAMPC1-GP
5 >0.999
10 0
0.006 0.041

Tbet+ SMARTAs (Log10)

4 0.36 4.97 4 25.7 17.2 4 41.7 2.88 0 10 20 30 40
10 10 10
4
10 Days post tumor inoculation
3
3 3 3 10 f) Gated on CD4 T cells - TDLN
10 10 10

2
10 Untx aCTLA4 (4F10 Blocking) aCTLA4 (9H10 Depleting) 0.1518 0.3264
0 0 0
1 0.44 0.30 0.24 105 <0.001 0.078
10 106
5 5 5
10 10 10

CD44+ PD1+ (Log10)

6.75 87.9 16.8 40.4 51.6 3.84
0 104

GP66 <APC>
10

GP66+ (Log10)
3 4 3 4 3 4 4 4 4

TDLN
0 10 10 0 10 10 0 10 10 10 10 10

TCF1 <PE>
>0.9999 103 105
PSGL1+ TBET+ (% of SMARTAs)

3 3 3
50 0.003 0.042 10 10 10
3.91 1.42 7.96 34.9 0.96 43.6
102
4 4 4
10 10 10 0 0 0
40

3 3 3 30 101 104
10 10 10 0 10 3 10 4 105 0 10 3 10 4 10 5 0 10 3 10 4 10 5

20 CD44 <UV737>
0 0 0
10
77.4 17.2 34.6 22.5 3.36 52.0 g) Gated on CD44+ GP66+ CD4 T cells - TDLN
0
0.018 0.018
3 4 5 3 4 5 3 4 5
0 10 10 10 0 10 10 10 0 10 10 10
Untx aCTLA4 (4F10 Blocking) aCTLA4 (9H10 Depleting)
PSGL1 <Percp-e710> >0.999 0.019 >0.999 0.105
0.5303 4 3.95 0.83 4 4.99 0 4 24.2 0.81 50 50

FOXP3+ (% of GP66+)
10 10 10
100 0.008 0.446

Tbet+ (% of GP66+)
4 2.13 3.20 4 10.7 32.1 4 12.7 31.9 10.8 9.07 4.84 40 40
Ki67+ (% of SMARTAs)

TBET <Pecy7>

10 10 10

80
TDLN

3 3 3
10 10 10 30 30
60
TBET <Pecy7>

3 3 3
10 10 10
20 20
40 0 0 0

0 0 0
10 10
85.0 10.2 85.9 9.07 71.0 4.03
20
39.1 55.6 14.8 42.4 13.9 41.5 0 10 3 10 4 10 5 0 10 3 10 4 10 5 0 10 3 10 4 10 5 0 0
0 10
3
10
4
10
5
0 10
3
10
4
10
5
0 10
3
10
4
10
5 0 FOXP3 <PE-e610>
KI67 <BV650>

TCF1+BCL6- (% of GP66+)
100 >0.999 100 0.1780
1.46 28.3 1.35 34.2 0.58 20.6

BCL6+ (% of GP66+)
b) Gated on SMARTAs - TUMOR 4 4 4 80 0.018 0.036
80
BCL6 <BV421>

10 10 10
TDLN

TDLN Stem CD4s TDLN Stem CD4s TDLN Tregs 0.3477 60 60

3 3 3
WT WT WT DTR WT DTR 4 0.006 0.572 10 10 10
10 40 40
0 0.61 0.43 3
SMARTAs (Log10)

10 0 0 0
20 20
CD44 <UV737>

4 4 4
10 10 10 10.4 59.9 10.4 54.1 11.9 67.0
2 0
10 0 10 3 10 4 10 5 0 10 3 10 4 10 5 0 10 3 10 4 10 5
0
3 3 3
1 TCF1 <AF488>
10
10 10 10

0
0 0 0 10 h) Gated on CD44+ PD1+ CD4 T cells - TUMOR

FOXP3+ (% of CD44+ PD1+)

0.001 0.0005
-1

TBET+ (% of CD44+ PD1+)

10 4 3.90 1.61 4 6.72 3.86 27.6 5.59
0 10
3
10
4
0 10
3
10
4
0 10
3
10
4 10 10 10
4
50 >0.999 0.027 100 0.896 0.019
TBET <Pecy7>

CD45.1 <UV395> >0.999 40 80

3 3
100 0.006 0.035
Tumor

3
10 10 10
30 60
0 0 87.5 3.12 0 61.1
Tbet+ (% of SMARTAs)

4 4 4
10 10 10
80 20 40
TBET <Pecy7>

0 0 0

60 10 20
3 3 3 22.7 71.8 18.4 71.0 22.6 44.2
10 10 10

0 10 3 10 4 10 5 0 10 3 10 4 10 5 0 10 3 10 4 10 5
0 0
40
0 0 0
FOXP3 <PE-e610>
20
0 0 9.38 0 11.1 27.8

0 10
3
10
4
10
5
0 10
3
10
4
10
5
0 10
3
10
4
10
5
0 i) j) Gated on CD44+ PD1+ CD4 T cells -
FOXP3 <PE-e610>
15 days after tumor inoculation - TUMOR
aCTLA4 (9H10) Untx aCTLA4 (9H10 Depleting)
Balb/c
Sacrifice 6.64 9.48 24.9 16.8
4 4
D15
TBET <Pecy7>
Kidney Tumor

10 10
7 days
3 3
10 10

RENCA-HA-Luc
0 0
100K/mouse
24.9 59.0 28.3 30.1

0 10 3 10 4 10 5 0 10 3 10 4 10 5

FOXP3 <PE-e610>
Ki67+ (% of FOXP3- CD4 T cells)

100 0.0010 0.0173 0.0010 0.0040

5000 50 100
FOXP3+ (% of CD44+ PD1+)
Tbet+ (% of CD44+ PD1+)
Geometric Mean CTLA4

80 4000 40 80

60 3000 30 60

40 2000 20 40

20 1000 10 20

0 0 0 0

Extended Data Fig. 6 | See next page for caption.

Extended Data Fig. 6 | Stem-like CD4 T cells are the source of Th1 cells in mean tumor diameter ± s.d. in untreated (n = 9), aCTLA4 4F10 (n = 9), or aCTLA4
response to Treg depletion and aCTLA4 therapy. a-b) Phenotypic analysis 9H10 (n = 9) mice and analyzed by Kruskal-Wallis test with Dunn’s multiple-
of recovered stem and Treg SMARTAs in TDLNs (a) and tumors (b) 7 days after comparison tests. f-g) Representative f) GP66 staining and g) phenotype of
transfer in WT or Treg depleted (DTR) recipients for each condition. Medians tumor specific CD4 T cells in TDLNs in each treatment group. h) Phenotype
are represented in each summary graph and analyzed by Kruskal-Wallis of bulk PD1 + CD4 T cells for each treatment group in tumors. Medians are
test with Dunn’s multiple-comparison tests. Data are representative of represented in each summary graph and were analyzed by Kruskal-Wallis test
2 independent experiments (n = 13 for stem-like and n = 5 for Tregs). with Dunn’s multiple-comparisons tests (n = 9 for each group). i) Experimental
c) Representative FOXP3 and CTLA4 staining with accompanying CTLA4 design to test the effects of aCTLA4 therapy in orthotopic RENCA-HA-
geometric MFI in TDLNs and tumors in TRAMPC1-GP bearing mice. Medians are Luciferase mice. j) Phenotype analysis of bulk PD1 + CD4 T cells in the tumor
shown and analyzed by two sided unpaired Mann-Whitney U test (n = 10 mice). for untreated or 9H10 CTLA4 treated orthotopic RENCA-HA-Luc mice. Medians
d) Experimental design to test how Tregs suppress stem-like differentiation are shown and were analyzed by two sided unpaired Mann-Whitney U test (n = 9
through CTLA4 in TRAMPC1-GP bearing mice. e) Tumor kinetics shown as for Untx and n = 5 for 9H10 aCTLA4 group).
Article
a) TRAMPC1-GP b) B16F10-GP c) MC38
0.0022 0.8808 WT WT + aCD8 0.0279 WT Treg dep. 0.0286
Treg dep +aCD8 Untreated 200 80
60 1.5 Treg dep. Treg dep + aCD8 100 0.1419 Treg dep + aCD8 100
100

Tbet+ (% of CD44+ PD1+)

Tbet+ (% of CD44+ PD1+)

Tbet+ (% of CD44+ PD1+)
Total CD4 depletion >0.9999

Tumor Size (mm2 )

80 80

Tumor Size (mm 2)

Tumor Size (mm2)
80 150 60

Tumor weight (g)

40 1.0
60 60 60
100 40
40 40 40
20 0.5
50 20
20 20 20

0 0 0.0 0 0 0 0
0 10 20 30 40 50 0 3 6 9 12 0 5 10 15 20
Days post tumor inoculation Days post tumor inoculation Days post tumor inoculation

d) Gated on GP33+ CD8 T cells - Draining LN - D5 post Treg depletion e) Gated on SPAS1+ CD8 T cells - Draining LN - D5 post Treg depletion
>0.999
Untx Treg dep. CD4 dep. WT Treg dep. CD4 dep.

Total GP33+/g of tumor (Log10)

5 5 0.081 0.265
10
0.094 0.94 1.40 10 0.041 0.16 0.062
4

Total SPAS1+ (Log10)

4

SPAS1 <BV421>
4
10
4
10 104 10 10
4
10
4
10
4
10
GP33 <APC>

3
10 3
3 3 2 3 10
103 3 3
10 10 10 10 10 10

1 2
10 10
0 0 0
0 0 0 0
1
10 10
0 15 30 45 60
0 103 104 105 0 103 104 105 0 103 104 105
Days post tumor inoculation
3 4 5 3 4 5 3 4 5 0
10
0 10 10 10 0 10 10 10 0 10 10 10

CD44 <UV737> CD44 <UV737>

>0.999 100 0.888

GZMB+ (% of SPAS1+)
100 20

% dead of each population

4 4.13 9.22 59.1 14.3 4.61 5.34 Untx 4 0 8.96 4 40.1 17.6 4 17.8 12.6
0.003 0.053

IFNgR1 (% of GP33)
10 0.007 0.005 10 10 10
Treg dep 80

GZMB <AF700>
IFNgR1 <BV605>

80 15
3 3
3
10 10 60
10 60 10
3
10
3
10
3

10
40 40
2 0
10 0
5 0 0 0
20
0 20
11.7 75.0 13.4 13.2 11.2 78.9 2.99 88.1 11.9 30.3 11.5 58.1
0 0 0
10 3 10 4 10 3 10 4 10 3 10 4 3 4 103 104 3 4
0 0 0
CD4 CD8 0 10 10 0 0 10 10

p
x >0.999

D 5
de
5
nt
100

CX3CR1+ (% of SPAS1+)
C D1
TCF1 <PE> D
U

4
5 5 5
0 5.97 31.4 5.90 12.6 6.50
10 10 10
0.005 0.038

CX3CR1 <PE-Dazzle>
f) Gated on GP33+ CD8 T cells -Tumor - D5 post Treg depletion
4 4 4
80
WT Treg dep. CD4 dep. 10 10 10
60
4
18.4 24.3 10
18.7 10
3
10
3
10
3
40
Total GP33+ (Log10)

3
4 4 4
10 10 10
10
GP33 <APC>

0 0 0 20
2
10
3
10
3
10
3 10 4.48 89.6 35.1 27.6 19.5 61.4
0
1 0 10
3
10
4
0 10
3
10
4
0 10
3
10
4

10 TCF1 <PE>
0 0 0

0
10
0 15 30 45 60 j) k)
0 10
3
10
4
10
5
0 10
3
10
4
10
5
0 10
3
10
4
10
5
Diptheria Toxin (DT) aCD4 (GK1.5) Naive LN-stem Effectors
Days post tumor inoculation
CD44 <UV737>
0.514 Pdcd1
4 4
104 100
WT or 12
10 47.9 1.54 10 70.0 4.93 60.1 2.35 <0.001 0.039 2.5 mil
FOXP3-DTR
TIM3+ (% of GP33+)

Sorted CD44+PD1+ 9
80 TRAMPC1-GP
TIM3 <BV421>

CD8 T cells
6
3 3
103 D5 post depletion

UMAP 2
10 10
60 4.5 weeks
3
40
0
20
0 0
UMAP 1
32.1 18.5 18.4 6.75 23.3 14.3
3 4 3 4 3 4
0 l) LN-stem cluster m)
0 10 10 0 10 10 0 10 10
p

Slamf6 Wnt10a Untx Treg dep.

15
17
T

Tcf7 Tox
de
5

TCF1 <PE>
W

D
D
D
4
GZMB+ GP33+ /g of tumor (Log10)

D
C

4 4 4
10 10 10
0.298
105 <0.001 0.013
15.4 1.10 36.9 5.14 11.8 4.96

104
GZMB <AF700>

3 3 3
10 10 10

103

0 0 0 102
Effector cluster
101
62.6 20.9 37.3 20.6 55.6 27.5
3 4 3 4 3 4
100 Mki67 Lgals1 Pclaf Gzmb CD4 dep.
0 10 10 0 10 10 0 10 10
p
T

de
5

TCF1 <PE>
W

Cluster distribution
4
D
C

WT WT by treatment
or or
g) FOXP3-DTR
B16F10-GP DT aCD4
FOXP3-DTR MC38 DT condition
250K 250K
Sacrifice at D5
7 days 5 days 10 days and D11
n) Treg dep. vs. Untx Treg dep. vs. CD4 dep. o) DEG Analysis between groups
LN-stem Effectors
h) Gated on CD44+ PD1+ CD8 T cells - Draining LN- D5 post depletion Untx Treg CD4 Treg CD4
100 <0.001 0.022 dep. dep. dep. dep.
WT Treg dep. CD4 dep.
0.075
80 Id2
TIM3+ (% of CD44+ PD1+)

2.91 1.94 48.9 21.4 9.84 21.0

B16-GP TDLN
TIM3 <BV421>

4 4 4
10 10 10
60
Tbx21

3 3 3
Eomes
10 10 10
40
Gzmb
0 0 0
20 Tgfbr1
4.85 90.3 7.06 22.6 5.03 64.1
3 4 3 4 3 4
0 Ifngr1
0 10 10 0 10 10 0 10 10
p

TCF1 <PE>
de
T
1
3
C D5
W
D
D

Average Expression Percent Expressed

4
D

i) WT Treg dep. 0.004 0.047 6 <0.001 0.509

100 10
TIM3+ (% of CD44+ PD1+)

5
8.85 4.10 59.8 13.9
-1 1 25 100
TCF1- TIM3+ (Log10)

10

80
TIM3 <BV421>

4 5
4
10
10
10
60
4 CD4 dep. vs. Untx Treg dep. vs. CD4 dep. p) IFN signaling signature
3
10
3
10
10
40 LN-stem clusters Effector clusters
3 0.999
20 10
<0.001 <2e-16 <2e-16
MC38 TDLN

0 0

6.84 80.2 3.10 23.2

0 2
3 4 3 4
10
0 10 10 0 10 10

0.261 0.004 5 <0.001 0.489 1 1

50 10
Enrichment score
GZMB+ (% of CD44+ PD1+)

TCF1- GZMB+ (Log10)

6.31 7.25 32.1 7.46

40
GZMB <AF700>

4
3 3 10
10 10
30

20 3
0 0 10
10 0 0
11.0 75.4 29.0 31.5
0 2
3 4 3 4 10
0 10 10 0 10 10
WT D5 D11 WT D5 D11 Untx Treg CD4 Treg CD4
TCF1 <PE> dep. dep.
dep. dep.

Extended Data Fig. 7 | See next page for caption.

Extended Data Fig. 7 | CD4 T cell help is required for effector CD8 Medians are represented in each summary plot and were analyzed by Kruskal-
differentiation in TDLNs. a-c) Th1 differentiation and tumor growth kinetics Wallis test with Dunn’s multiple-comparison tests. j) Experimental design.
shown as mean ± s.d. for a) TRAMPC1-GP (Untx n = 8, aCD8 n = 6), b) B16F10-GP ScRNAseq of sorted naïve and activated (CD44 + PD1 +) CD8 T cells from TDLNs
(n = 3 mice per group) and c) MC38 tumors (n = 4 mice per group) for Untx mice, of Untx, Treg depleted, and total CD4 depleted TRAMPC1-GP bearing mice
Treg depleted mice, total CD4-depleted mice (GK1.5 clone), and combination 5-days after depletion (n = 4–12 pooled mice per group). k-m) Cluster distributions
of Treg and CD8 T cell (Lyt3.2 clone) depleted mice. Medians shown in summary and UMAP projections of normalized expression of genes defining the naïve,
plots. Statistical comparisons were performed using two-sided unpaired LN-stem, and effector CD8 clusters in all conditions. n) Transcriptional
Mann-Whitney U test or Kruskal-Wallis test with Dunn’s multiple-comparison comparisons of LN-stem and effector CD8 T cells between groups. Volcano
tests when appropriate. d-e) Phenotypic analysis of d) GP33+ and e) SPAS1 + CD8 plots show fold change versus -log(p-value) for each gene within each respective
T cells in TRAMPC1-GP TDLNs 5-days after Treg or total CD4 T cell depletions. comparison. o) Transcriptional comparison of selected genes across LN-stem
f) Phenotype of GP33 + CD8 T cells in tumor 5-days after Treg or total CD4 T cell (green) and effector (red) clusters between Untx, Treg depleted, and CD4
depletion. Every time point represents an independent experiment (n = 5–17 depleted groups. The color and size of the circles represent the normalized
mice per group). Medians or mean ± s.e.m are represented in each summary expression and proportion of cells expressing that gene, respectively.
plot and were analyzed by Kruskal-Wallis test with Dunn’s multiple-comparison p) VISION GSEA using an IFN signaling signature. Signature enrichment score is
tests. g) Schematic of experimental designs. h-i) Phenotypic analysis of bulk represented as violin plots for LN-stem (green) and effector (red) clusters in the
activated CD8 T cells in h) B16F10-GP i) or MC38 TDLNs after Treg or total CD4 respective groups. Mean is represented in each violin and were analyzed by
depletion. Summary plots show total number or frequency of activated CD8 one-way ANOVA with Tukey’s multiple comparison tests (n = 4–12 pooled mice
T cells expressing the indicated marker at the respective time points per group).
(n = 3–7 mice per group per timepoint for each independent experiment).
Article
a) Gated on CD44+ SMARTAs- TDLN Gated on CD44+ SMARTAs- TDLN
WT WT + aPDL1 TetON WT WT + aPDL1 TetON
WT SMARTAs TetON TBET SMARTAs TetOn TBET SMARTAs WT SMARTAs TetON TBET SMARTAs TetOn TBET SMARTAs
TetON TetON + aPDL1 No DOX
No DOX No DOX TetON TetON + aPDL1 No DOX
<0.001
100 <0.001 50
3.67 5.13 17.7 0.93 1.66 0.097 5
0.32 0.64 5
2.80 7.32 5
0.18 0.18 0.0096
<0.001 0.014
10 10 10

IFNg+ (% of SMARTAs)
Tbet+ (% of SMARTAs)
80 0.003 0.002 40 0.031
TBET <Pecy7>

4 4 4
3 3 3
10 10 10 0.047 0.003

IFNg <APC>
10 10 10

60 30
3 3 3
10 10 10
2 2 2
40 20
10 10 10

10
0 0 0
20 0 0 0

81.2 10.0 77.5 3.79 92.5 5.75 59.6 39.5 58.7 31.2 60.3 39.4

0 10
3
10
4
0 10
3
10
4
0 10
3
10
4 0 0 10
3
10
4
0 10
3
10
4
0 10
3
10
4 0
FOXP3 <PE-e610> IL2 <PE>

b) Gated on Endogenous GP66+ CD4 T cells - TDLN

WT WT + aPDL1
Gated on Endogenous GP66+ CD4 T cells - TDLN
WT WT + aPDL1
WT SMARTAs TetON TBET SMARTAs TetOn TBET SMARTAs TetON TetON + aPDL1 WT SMARTAs TetON TBET SMARTAs TetOn TBET SMARTAs TetON TetON + aPDL1
No DOX No DOX
<0.001 <0.001
4 0.004 0.013
3.96 3.96 17.0 6.70 4.58 1.53 10 5
0.100 0.066 5
0.19 0.29 5
0.19 0.017
104
10 10 10

IFNg+ CD44+PD1+ (Log10)

0.032 0.018

Tbet+ GP66+ (Log10)

TBET <Pecy7>

4 4 4

IFNg <APC>
3 3 3
10 10 10 3 10 10 10
10
103
3 3 3
10 10 10
2
10 102
2 2 2
10 10 10

0 0 0 0 0 0

78.2 13.9 52.1 24.2 74.8 19.1 99.2 0.66 99.0 0.52 99.1 0.70
1
0 10
3
10
4
0 10
3
10
4
0 10
3
10
4
10
0 10
3
10
4
0 10
3
10
4
0 10
3
10
4 101
FOXP3 <PE-e610> IL2 <PE>

c) Gated on Endogenous CD4 T cells - TDLN d) TUMOR

WT SMARTAs WT SMARTAs TetON TBET SMARTAs TetON TBET SMARTAs WT WT + aPDL1 TetON TetON + aPDL1
+ aPDL1 + aPDL1 <0.001 0.001 <0.0001 0.001

GP66+ Endo CD4 T cells (Log10)

0.005 0.012 0.028

TBETT# CD44PD1/g of tumor

10
5
0.13 10
5
0.16 10
5
0.29 10
5
0.43 105 50 106 105

TBETT# CD44PD1 TDLN

<0.001 0.017 0.003

Tbet+ (% of GP66+)
GP66 <APC>

4 4 4 4 40 104
10 10 10 10
105
104
3 3 3 3
30 103
10 10 10 10
104
20 102
103
103
0 0 0 0
10 101
3 4 5 3 4 5 3 4 5 3 4 5
102 102 100
0 10 10 10 0 10 10 10 0 10 10 10 0 10 10 10
0
CD44 <UV737>

e) Gated on CD44+PD1+ CD8 T cells - TDLN f) Gated on Endogenous CD44+PD1+ CD8 T cells - BLOOD
WT SMARTAs WT SMARTAs TetON TBET SMARTAs TetON TBET SMARTAs
+ aPDL1 + aPDL1
<0.0001 <0.0001 <0.0001 0.0003

CD44+ PD1+ (% of CD8 T cells)

TIM3+ CD8 T cells (% of total)

2.63 10.5 7.75 10.2 2.92 2.75 16.3 2.78 106 0.0246 105 0.0542 0.0192 100
20
0.0163 0.0010 0.0033 0.0329
CD44+ PD1+ (Log10)

TIM3# 44PD1+ CD8s

4 4 4 4
10 10 10 10 0.0040 0.0285 0.0251
104 10-1
TIM3 <BV421>

15
105
3 3 3 3
10 10 10 10
103 10 10-2

104
0 0 0 0 102 5 10-3
2.92 83.9 4.05 78.0 2.31 92.0 10.9 70.0

0 103 104 0 103 104 0 103 104 0 103 104 103 101 0 10-4
TCF1 <PE>
g) Gated on Endogenous CD44+PD1+ CD8 T cells - TUMOR 0.0050 0.0050 0.0685 0.0655
0.0193 0.0228 0.0149 0.0249

TIM3# GP33+/g of tumor (Log10)

24.8 4.59 23.3 2.50 32.8 5.81 31.1 1.78 105 105 104 104
GZMB+ PD1+ CD8 T cells

4 4 4 4
0.0020
10 10 10 10 0.0052

per g of tumor (Log10)

per g of tumor (Log10)
CD44+PD1+ CD8 T cells
per g of tumor (Log10)

104 103

GZMB+ GP33+
104 103
GZMB <AF700>

3 3 3 3
10 10 10 10
103 102
103 102
102 101
0 0 0 0
102 101
101 100
44.0 26.6 53.3 20.8 35.5 25.9 53.3 13.8

0 10
3
10
4
0 10
3
10
4
0 10
3
10
4
0 10
3
10
4
101 100 100 10-1
TCF1 <PE>

WT WT + aPDL1 TetON TetON + aPDL1 0.0002

h) 200 200 200 200 3 0.0007
0.0161
WT
Tumor Size (mm2 )

Tumor Size (mm2)

Tumor weight (g)

Tumor Size (mm2)

Tumor Size (mm2)

150 150 150 150

2 WT + aPDL1
100 100 100 100 TetON
1
50 50 50 50 TetON + aPDL1

0 0 0 0 0
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
Days post tumor inoculation Days post tumor inoculation Days post tumor inoculation Days post tumor inoculation

i) 200k j) Gated on CD44+ SMARTAs- TDLN

WT SMARTAs WT SMARTAs TetON TBET SMARTAs
or <0.001 0.017 0.412
100 20 50
rtTA TetON TBET SMARTAs 4 1.10 0 4 50.0 0
IFNg+ (% of SMARTAs)
Tbet+ (% of SMARTAs)

IL2+ (% of SMARTAs)

10 10

80 40
TBET <Pecy7>

Doxycycline Sacrifice 15
4 days D14 10
3
10
3
60 30
10
40 20
2 2
10 10
5
500K 0 0 20 10
B16F10-GP 89.6 9.34 49.4 0.59
0 0 0
3 4 3 4
0 10 10 0 10 10
WT TetON WT TetON WT TetON
FOXP3 <PE-e610>

k) Gated on Endogenus CD44+ PD1+ CD4 T cells - TUMOR

WT SMARTAs TetON TBET SMARTAs
<0.001 0.002 <0.001 0.531
FOXP3+ (% of CD44+ PD1+)
TBETT# CD44PD1/g of tumor

4 4
Tbet+ (% of CD44+ PD1+)

4 5.27 37.5 4 39.2 7.09 50 10 100 10

10 10
FOXP3#/g of tumor
TBET <Pecy7>

40 80
3 3
3 3 10 10
10 10 30 60

20 2 40 2
10
2
10
2 10 10
0 0 10 20
35.3 22.2 32.6 21.1
0 1 0 1
3 4 3 4
10 10
0 10 10 0 10 10
WT TetON WT TetON WT TetON WT TetON
FOXP3 <PE-e610>

Extended Data Fig. 8 | See next page for caption.

Extended Data Fig. 8 | Localized Th1 response is sufficient to induce a-g) Median are represented in each summary plot and were analyzed by
endogenous stem-like to Th1 differentiation and improve response to Kruskal Wallis test with Dunn’s multiple-comparison tests (n = 4–6 mice per
immunotherapy in the presence of Tregs. a) Phenotype and cytokine group for each experiment). h) Individual tumor growth kinetics of TRAMPC1-
production after GP61-77 stimulation of transferred SMARTAs across all groups GP mice after SMARTA transfer and tumor weights at endpoint for each group.
in TDLNs. Summary plots show the frequency of SMARTAs expressing TBET Summary tumor weights represented as grams for each group. Medians are
and producing IFNg. b) Phenotype and cytokine production after GP61-77 shown and were analyzed by Kruskal Wallis test with Dunn’s multiple-
stimulation of endogenous GP66 + CD4 T cells in TDLNs for each group. comparison tests (n = 4–6 mice per group for each independent experiment).
Summary plots show the total number of GP66 + CD4 T cells expressing TBET i) Experimental design. j) Phenotype and cytokine production after GP61-77
and producing IFNg. c-d) Representative GP66 staining in TDLNs for each stimulation of transferred WT or TetON TBET SMARTAs in TDLNs. k) Phenotype
group and summary of the phenotype of endogenous bulk and GP66 + CD4 of endogenous PD1 + CD4 T cells in the tumor. Summaries show the frequency
T cell expressing Tbet in TDLNs and tumors. e-h) Phenotype of endogenous and number of Th1 and Tregs infiltrating the tumor. Data represents two
bulk activated (CD44 + PD1 +) or GP33 + CD8 T cells in e) TDLNs, f) blood and g) independent experiments and were analyzed by two-sided unpaired Mann
tumor for each group. Data represents 3 independent experiments. Whitney U test (n = 5–7 mice per group for each experiment).
Article
a) b) Untx e)
160 WT or
WT or aIFNg DT Treg dep. + aIFNg
FOXP3-DTR aIL-12 DT aIFNg

0.074 <0.001
FOXP3-DTR

Tumor Size (mm )

2
Treg dep.

<0.001
120 Sacrifice
25 days D12
80
18 days 250K B16-GP D6 D7 D8 D10
2.5 mil 40
TRAMPC1-GP
0
0
5 10 15 20 25 30 35 40 45
Days post tumor inoculation
c) Untx Treg dep. Treg dep. +aIFNg f) Gated on CD44+ PD1+ CD4 T cells - TDLNs- 5days after treatment
0.600 <0.001
Untx Treg dep. Treg dep. + aIL12 Treg dep. + aIFNg
5 <0.001 <0.001 0.212
0.002 0.003 50 <0.001
5 <0.001 <0.0001
10 100 10 100

Tbet+ (% of CD44+ PD1+)

<0.001 <0.001 4.85 5.98 40.8 9.16 45.3 9.58 26.9 8.03 <0.0001
Total GP66+ (Log10)

FOXP3+ (% of GP66+)
Tbet+ GP66+ (Log10)
Tbet+ (% of GP66+) 80 40 80 <0.0001

TBET <Pecy7>
4
10 3 3 3 3
10 10 10 10
60 30 60
4 3
10 10
40 20 40
2 2 2 2
2 10 10 10 10

20 10 10 0 0 0 0 20
18.8 70.3 38.9 11.2 38.0 7.20 35.3 29.7
3 0 1 0 0
10 10
0 10 3 10 4 0 10 3 10 4 0 10 3 10 4 0 10 3 10 4
Untx D5 D5 Untx D5 D5 Untx D5 D5 Untx D5 D5 FOXP3 <PE-e610>

d) Gated on GP33+ CD8 T cells - TDLNs g) Gated on GP33+ CD8 T cells - TDLNs - 5 days after treaments
Untx Treg dep. Treg dep. + aIFNg <0.001 Untx Treg dep. Treg dep. + aIL12 Treg dep. + aIFNg
100 <0.001<0.001 100 0.004
6.67 23.9 82.9 5.98 29.8 11.3 4.81 0.96 70.0 8.37 63.4 10.7 15.3 7.34 0.006

TIM3+ (% of GP33+)
Tim3+ (% of GP33)
4 4 4 80 4 4 4 4 80 <0.001

TIM3 <BV421>
TIM3 <BV421>

10 10 10 10 10 10 10

60 60
3 3 3 3 3 3 3
10 10 10 10 10 10 10
40 40
0 0 0
20 0 0 0 0 20
2.80 66.7 6.05 5.05 24.7 34.3 3.85 90.4 11.3 10.3 11.2 14.7 11.9 65.5
0 0
0 103 104 0 103 104 0 103 104 0 10 3 10 4 0 10 3 10 4 0 10 3 10 4 0 10 3 10 4
0.002
100
106 <0.001 0.039 0.022
GZMB+ GP33+ (Log10)

2.80 10.3 74.6 1.84 31.1 5.63 3.57 19.0 54.5 5.32 44.8 9.47 10.2 10.2

GZMB+ (% of GP33+)
4 4 4 4 4 4 4
<0.001
GZMB <AF700>

10 10 10 10

GZMB <AF700>
10 10 10

105 80

10
3
10
3
10
3 104 10
3
10
3
10
3
10
3 60

103 40

102
0 0 0
0 0 0 0
20
9.03 77.8 10.2 13.4 26.2 37.1 3.57 73.8 25.5 14.6 22.7 22.9 19.2 60.5
101 0
0 10
3
10
4
0 10
3
10
4
0 10
3
10
4 0 10 3 10 4 0 10 3 10 4 0 10 3 10 4 0 10 3 10 4

100 <0.001 <0.001 <0.001

5 4.81 5 67.0 5 5 19.8
4 4 4 10 32.7 10 14.3 10 57.5 17.5 10 37.3 100
IFNgR1+ (% of GP33)

10
18.9 10 70.5 10
44.7 <0.001 0.011 0.003
KI67 <BV650>

80

Ki67+ (% of GP33+)
IFNgR1 <BV605>

10
4
10
4
10
4
10
4
80
10
3
10
3
10
3
60
3 3 3 3 60
10 10 10 10 0.009
40
40
0 0 0
0 0 0 0
20
3.85 58.7 12.2 6.45 14.3 10.7 10.7 32.2 20
0 10
3
10
4
0 10
3
10
4
0 10
3
10
4 0 0 10 3 10 4 0 10 3 10 4 0 10 3 10 4 0 10 3 10 4
0
TCF1 <PE> TCF1 <PE>

h) Gated on CD44+ SMARTAs - TDLN i)

TetON TBET SMARTAs IFNg KO TetON TBET SMARTAs WT Untx no SMARTAs TetON +aPDL1 IFNg KO TetON + aPDL1
+ aPDL1 + aPDL1 0.013
100 240 240 240
24.0 28.0 21.8 40.6
200
IL2+ (% of SMARTAs)

200
Tumor Size (mm2)

200
Tumor Size (mm2)

Tumor Size (mm2)

10
4
10
4 80
TNFa <BV421>

160 160 160

60
3 3
120 120 120
10 10
40 80 80 80
40 40 40
0 0
20
0 0 0
43.8 4.21 33.5 4.14
0 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
0 10
3
10
4
0 10
3
10
4
Days post tumor inoculation Days post tumor inoculation Days post tumor inoculation
IL2 <PE>

j) Gated on Endogenous CD4 T cells - TDLN k) Gated on Endogenous CD8 T cells

TetON TBET SMARTAs IFNg KO TetON TBET SMARTAs TetON TBET SMARTAs IFNg KO TetON TBET SMARTAs
+ aPDL1 + aPDL1 + aPDL1 + aPDL1
5 0.345 0.181
IFNg# Endogenous CD4 T cells

5
0.15 0.016 5
0.11 0.017 10 18.9 14.7 8.45 12.8 40
10 10
CD8 T cells - TDLN

TIM3 <BV421>

4 4
TIM3+ (% of GP33+)

10 10
30
IFNg <APC>

4 4
10 10 4
10
GP33+

3 3
3 3
10 10 20
10 10
3
10
0 0 0 0 10
99.7 0.11 99.7 0.16 6.21 60.2 5.54 73.2
2
10 0
3 4 3 4 3 4 3 4
0 10 10 0 10 10 0 10 10 0 10 10

IL2 <PE> TCF1 <PE>

4 0.081
40 0.014 100 0.020 10
FOXP3+ KI67+ (% of PD1+)

FOXP3+ (% of CD44+ PD1+)

14.9 59.3 10.5 5.93 5.53

GP33+ / g of tumor (Log10)

5 5
10
4
10
4 64.9 l) 10 8.76 10 3.01
80 3
TBET <Pecy7>

GP33 <APC>

30 4 4 10
10 10
3 3
60
10 10
2
20 3 3
10
GP33+ CD8 T cells - TUMOR

40 10 10

1
10 10
0 0 20 0 0

25.8 48.9 29.1 59.4

0 0 0
3 4 3 4 3 4 3 4
10
0 10 10 0 10 10 0 10 10 0 10 10

FOXP3 <PE-e610> CD44 <UV737>

0.020
GZMB+ GP33+ /g of tumor (Log10)

15.3 6.81 5.00 3.24

40 0.052 10 4 0.178 68.2 3.48 38.6 8.57 10
3
4 4
10 10
Tbet+ GP66+ (Log10)
Tbet+ (% of GP66+)

4 4

30
10 10
2
TBET <Pecy7>

TIM3 <BV421>

10
10 3
3 3
10 10
20 3 3 1
10 10 10
10 2
10 0
0 0 0 0 10
57.9 20.0 67.1 24.7 16.0 12.3 20.0 32.9
0 10 1 -1
3 4 3 4 3 4 3 4
10
0 10 10 0 10 10 0 10 10 0 10 10

FOXP3 <PE-e610> TCF1 <PE>

Extended Data Fig. 9 | See next page for caption.

Extended Data Fig. 9 | Th1-derived IFNg is required for effective anti-tumor 5-days after each depletion or cytokine blocks. Medians are represented in
immunity. a) Experimental design for systemic IFNg block in TRAMPC1-GP each summary plot and were analyzed by One-way ANOVA test with Tukey’s
mice. b) TRAMPC1-GP tumor kinetics shown as tumor diameter. Data are multiple comparison tests between groups (n = 4–8 mice per group).
representative of 2 independent experiments. Mean ± s.e.m. shown and were h) Cytokine production after GP61-77 stimulation of transferred TetON or
analyzed by Kruskal Wallis test with Dunn’s multiple comparison tests between IFNg KO TetON SMARTAs in TDLNs of TRAMPC1-GP mice. i) Individual tumor
groups (Untx n = 11, Treg dep. n = 13, Treg dep + aIFNg n = 8). c) Total numbers growth kinetics of TRAMPC1-GP mice after SMARTA transfer for each group.
and phenotypic analysis of tumor specific GP66 + CD4 T cells in TDLNs 5-days j) Phenotype and cytokine production (GP61-77 peptide stimulation) of
after Treg depletion and IFNg block. d) Phenotypic analysis of GP33 + CD8 endogenous GP66+ or bulk PD1 + CD4 T cells in TDLNs of mice receiving WT
T cells in TDLNs in TRAMPC1-GP bearing mice after 5-days of Treg depletion or IFNg KO TetON TBET SMARTAS. k-l) Phenotypic analysis of GP33 + CD8
and IFNg block. Medians are represented in each summary plot and were T cells in TDLNs (k) and tumor (l) of mice receiving WT or IFNg KO TetON TBET
analyzed by One-way ANOVA test with Tukey’s multiple comparison tests SMARTAS. Median are represented in each summary plot and were analyzed
between groups (n = 6-13 mice per group). e) Experimental design for by two-sided unpaired Mann Whitney U test between groups. (n = 6–8 mice per
systemic IFNg and IL-12 block in B16-F10-GP mice. f-g) Phenotypic analysis of group).
CD44 + PD1 + CD4 or GP33 + CD8 T cells in TDLNs of B16F10-GP bearing mice
Article
a) P14 x iCAS9 b) VEX+ P14s - WT mice - no DOX c) VEX+ P14s VEX+ P14s VEX+ P14s
8 weeks VEX+ P14s - WT mice - DOX WT - no DOX WT - DOX DTR - no DOX
VEX+ P14s - DTR mice - no DOX 700 30

Geometric Mean IFNgR1

WT 0.046 4.50 0.042 9.19 16.8 10.4
P14 inducible 4 4 4

GZMB <AF700>
10 10 10

GZMB+ (% of P14s)
BM HSCs IFNgR1 KO
600 20
14 days 10
3
10
3
10
3

in culture
500 10
Sort VEX+ 0 0 0

HSCs 0.28 95.2 0.21 90.6 5.50 67.3

IFNgR1 sgRNA 3 400 0 10
3
10
4
10
5
0 10
3
10
4
10
5
0 10
3
10
4
10
5 0
0 10
TCF1 <PE>
IFNgR1 <BV605>

d) 104 40 e) 0.007 >0.999 f) 0.069

103 100
WT P14s KO P14s WT P14s KO P14s
GZMB+ (% of P14s)

PD1+ CD44+ P14s (%)

103 30 80
P14#/g of tumor

60
102

IFNgR1 MFI
20
40
101 10
20

100
3
10 3
102 0
0 0 10
0
IFNgR1 <BV605> WT KO WT KO PD1 <BV786> WT KO
P14s Endo GP33

g) WT P14s IFNgR1 KO P14s

<0.001 h) WT P14s IFNgR1 KO P14s
<0.001
100 100
5 5
10 18.6 61.5 10 4.52 37.2 20.7 19.2 2.22 4.44

80 80

TIM3+ (% of P14s)
4 4
KI67+ (% of P14s)

10 10

TIM3 <BV421>
KI67 <BV650>

4 4
10 10
60 60
3 3
10 10
3 3
10 10
40 40

0 0 20 0 0 20
4.54 15.3 3.35 54.9 4.32 55.8 3.33 90.0
0 0
0 10 3 10 4 0 10 3 10 4 0 10 3 10 4 0 10 3 10 4

TCF1 <PE> TCF1 <PE>

i) Endo GP33+ Endo GP33+ 0.388 6

0.928 j) 0.045 WT P14s KO P14s 0.514
100 10 6

P14 number / g of tumor (Log10)

10
4 51.8 8.10 10
4 49.5 6.99 10 4 53.0 9.07 4 48.0 7.43 100
10 10
GZMB+ (% of GP33)

GZMB+ (% of P14s)
80
GZMB <AF700>

5 80
10 5

GZMB <AF700>
GP33+ (Log10)

10
10
3
10
3 60 3 3 60
4 4 10 10
10 10
40 40
0 0 3 3
10 10 0 0
20 20
17.9 22.2 22.6 20.9 20.8 17.2 20.1 24.5
2 2 0
0 10 10
0 10 3 10 4 0 10 3 10 4 0 10
3
10
4
0 10
3
10
4

TCF1 <PE> TCF1 <PE>

k) Gated on PD1+ CD45RA- CD4 T cells

CD28 Geometric MFI GZMB Geometric MFI
TH1 low TH1 high 0.273 0.143
1 1 4 4
10 10 10 10
4 29.8 2.61 4 7.61 0.57 0 3 81.3
10 10
10 3
TCF1+ lin- (% of total)

10
0
FOXP3+ (% of total)

3
10

GZMB <AF700>
10 10
CD28 <Pecy7>

-1
FOXP3 <PE>

10 17.8
-2 -1 2
10
3
10
3
10 10 2 10
10
3 2
-3 10 10
10 -2 1
10 10
-4 15.7 0
2 2
10 0
0
10 10
-3 10
0 0 -5 10
10
29.4 38.2 59.2 32.7
-6 -4 2 -1
10 10 10 10
10 2 10 3 10 4 10 2 10 3 10 4 0 10
3
0 10
3
TH1 low TH1 high TH1 low TH1 high
TCF1 <AF488> TBET <BV605> TBET <BV605>
2 <0.001
10
1
3.06 1.56 31.1 10.3 10
0
10
Tbet+ (% of total)

3 3
TBET <BV605>

10 10
-1
10 n) Models of T cell differentiation in cancer
-2
10
-3 “Restricted” TDLN - Poor anti-tumor response
10
0 0
-4
10
-5 Limited CD4 help
61.2 34.1 36.7 21.9
10
-6
10
10 2 10 3 10 4 10 2 10 3 10 4
TH1 low TH1 high
TCF1 <AF488>
Activated
l) Gated on PD1+ CD45RA- CD8 T cells m) nTreg CD4s Stem-like iTreg CD4s TCF1+ CD8s
TH1 low TH1 high 0.337 TCF1+ lin- CD4s LN Stems
1 0
10 10 r = 0.7961
GZMB+ CD8 T cells (% total)

64.4 5.25 73.2 6.43 0 p=0.0017

10 -1
TCF1+ (% of total)

“Active” TDLN - Effective anti-tumor response

4 4
10
CD39 <BV421>

10 10
-1
10
-2 -2
10
3
10
3
10 10
IFNg
-3
10 -3
10
-4
0 0 10
27.7 7.06 12.6 7.79 Stem-like Th1 CD4s Activated Effector CD8s
-5 -4
2 3 4 2 3 4 10 10 TCF1+ lin- CD4s TCF1+ CD8s Cytotoxic
10-5 10-4 10-3 10-2 10-1
10 10 10 10 10 10
TH1 low TH1 high LN Stems
TCF1 <AF488> TBET+ CD4 T cells (% total)

Extended Data Fig. 10 | See next page for caption.

Extended Data Fig. 10 | IFNg is intrinsically required for tumor specific patients that received immunotherapy. Patients where stratified based on
CD8 LN-stem to effector differentiation. a) Schematic to make IFNgR1 KO the % of total TBET + PD1 + CD4 T cells in the resected primary tumors prior to
P14s using an HSC BM chimera lentiviral system. b-d) Phenotypic analysis of therapy. l) Representative phenotype of activated (PD1 + CD45RA-) CD8 T cells
transferred VEX+ P14s in WT mice with and without doxycycline administration in Th1 low and Th1 high patients. Medians are represented and analyzed by
and Treg depleted (DTR) mice without doxycycline treatment in TDLNs and two-sided unpaired Mann-Whitney U test (n = 47 kidney cancer patients).
tumor. Medians are represented in each summary plot (n = 4–5 mice per group). m) Spearman correlation between TBET + CD4 T cells and effector GZMB + CD8
e-f) Representative flow cytometry stain of IFNgR1 and PD1 expression in WT T cell populations as a percent of total cells in non-metastatic draining lymph
and KO P14s after transfer into FOXP3-DTR TRAMPC1-GP mice. Endogenous nodes from kidney cancer patients (n = 12). n) Proposed models of T cell
GP33 + CD8 T cells in each respective mouse are included as a comparison. differentiation states in TDLNs. Restricted TDLN state. Tregs actively
Medians are represented in each summary plot from one representative suppress stem-like CD4 T cells, preventing their differentiation into Th1 cells
experiment and analyzed by two-sided unpaired Mann Whitney U test between and promoting iTreg differentiation. Minimal CD4 T cell help is provided in this
WT and KO (n = 7 mice). g-h) Phenotype of transferred WT and IFNgR1 KO P14s state and CD8 T cells are maintained in an activated stem-like state in TDLNs.
in TDLNs 5-days after Treg depletion. i) Total number and phenotypic analysis Anti-tumor response is not optimal, and tumor outcompetes the T cell response.
of endogenous GP33 + CD8 T cell numbers in TDLNs in mice with transferred Active TDLN state. In the absence of Treg suppression stem-like CD4 T cells
WT and KO P14s. j) P14 number and phenotype in the tumor 5-days after Treg undergo Th1 differentiation. Th1 CD4s secrete IFNg which promotes stem-like
depletion. Data are representative of 2 independent experiments (n = 5–6 mice to effector CD8 T cell differentiation. Anti-tumor effector response is optimal,
per group for each experiment). Median are represented in each summary and tumor is controlled. Stem-like CD4 T cell differentiation can be targeted
plot and were analyzed by two-sided unpaired Mann-Whitney U test. and stimulated to generate Th1 cells in the presence of Tregs. Stem to Th1 CD4
k) Representative phenotype of activated (PD1 + CD45RA-) CD4 T cells in differentiation is then sufficient to switch between active and restricted states
primary tumors upon surgical resection in a cohort of 47 kidney cancer and rescue response to immunotherapy.