transcriptional heterogeneity in cancer-associated …37 [email protected] (yz). 38 conflicts of...

Title: Transcriptional heterogeneity in cancer-associated regulatory T cells is predictive 1

of survival. 2

Authors: Nicholas Borcherding1,2,3*, Kawther K. Ahmed1,11*, Andrew P. Voigt3, Ajaykumar 3

Vishwakarma1,2, Ryan Kolb1,12, Paige Kluz1,4, Gaurav Pandey1, Katherine N. Gibson-Corley1, 4

Julia Klesney-Tait5, Yuwen Zhu6, Jinglu Lu7, Jinsong Lu7, Xian Huang8,9, Jinke Cheng9, Song 5

Guo Zheng10, Xuefeng Wu8,9, Yousef Zakharia5, Weizhou Zhang1,2,3,4,12 6

Author Affiliations: 7

1 Department of Pathology 8

2 Cancer Biology Graduate Program 9

3 Medical Scientist Training Program 10

4 Free Radical and Radiation Biology Program 11

5 Department of Internal Medicine 12

Iowa City, IA 52242 13

6 Department of Surgery, University of Colorado Anschutz Medical Campus, Aurora, CO 80045 14

7 Department of Breast Surgery, Renji Hospital, Shanghai Jiao Tong University School of 15

Medicine, Shanghai, Shanghai, 200127, China; 16

8 Shanghai Institute of Immunology, Department of Immunology and Microbiology, Key 17

Laboratory of Cell Differentiation and Apoptosis of Chinese Ministry of Education; 18

9 Shanghai Key Laboratory for Tumor Microenvironment and Inflammation, Key Laboratory of 19

Cell Differentiation and Apoptosis of Chinese Ministry of Education, Department of Biochemistry 20

and Molecular Cell Biology, Shanghai Jiao Tong University School of Medicine, Shanghai 21

200025, China 22

10 Department of Medicine, Penn State College of Medicine and Milton S. Hershey Medical 23

Center, Hershey, PA 17033 24

11 Current address: College of Pharmacy, University of Baghdad, Department of Pharmaceutics, 25

Baghdad, Iraq. 26

.CC-BY-NC 4.0 International licenseis made available under aThe copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It. https://doi.org/10.1101/478628doi: bioRxiv preprint

https://doi.org/10.1101/478628

http://creativecommons.org/licenses/by-nc/4.0/

12 Current Address: Department of Pathology, Immunology and Laboratory Medicine, University 27

of Florida, Gainesville, FL 32610 28

* Authors contributed equally to this work 29

30

Running Title: Heterogeneity and functional implications of cancer-associated regulatory T 31

cells. 32

Financial Support: W.Z. was supported by NIH grants CA200673, and CA203834. Y.Z. was 33

supported by funding from the Rock ‘n’ Ride. N.B. was supported by F30 CA29655. This work is 34

also supported by Holden Comprehensive Cancer Center support grant P30 CA086862. 35

Correspondences: [email protected] (WZ); [email protected] (XW); Yousef-36

[email protected] (YZ). 37

Conflicts of Interest: All the authors claim no conflict of interest. 38


https://doi.org/10.1101/478628


Regulatory T cells (Tregs) are a population of T cells that exert a suppressive effect on a variety 39

of immune cells and non-immune cells1. Similar to the role in peripheral tolerance, the 40

suppressive effects of Tregs are detrimental to anti-tumor immunity1. The prognostic value of 41

FOXP3, the master regulator of Tregs, is highly variable2–6. Recent investigations have identified 42

common expression patterns for tumor-infiltrating Treg (TI-Treg)7–9; however, the heterogeneity 43

of TI-Tregs at the functional levels is largely unknown across different cancer types. We 44

performed single-cell RNA sequencing (SCRS) on immune cells derived from renal clear cell 45

carcinoma (ccRCC) patients, isolating 160 peripheral-blood (PB) Tregs and 574 TI-Tregs. We 46

also analyzed a published SCRS datasets containing 634 TI-Tregs from human hepatocellular 47

carcinomas (HCC). We identified distinct transcriptional cell fates in both ccRCC and HCC, with 48

the suppressive subsets of TI-Tregs expressing a common gene signature including CD177 as 49

the most upregulated gene. We further demonstrate CD177+ Tregs have preferential 50

suppressive effects across human and mouse. Gene signatures derived from CD177+ Tregs 51

had superior predictability for patient survival in several cancer types, outperforming other 52

reported Treg signatures. Further investigation into the development and regulation of TI-Treg 53

heterogeneity will be vital to the application of tumor immunotherapies with potential minimal 54

side effects. 55

56

In order to investigate the heterogeneity and dynamics of TI-Tregs, we performed SCRS on 57

peripheral and TI immune cells from three ccRCC patients. ccRCC tumors are responsive to 58

immune checkpoint blockade despite low mutational loads, indicative of microenvironmental 59

involvement10. A total of 25,672 immune cells were sequenced, with 160 PB and 574 TI-Tregs 60

isolated using the expression of FOXP3 and CD25 (IL2RA) (Figure 1a-b). To identify 61

differentially-expressed genes and markers of TI-Tregs, we compared TI- versus PB-Tregs 62

expression (Figure 1c). A complete list is available in Supplemental Table 1. The only two genes 63


https://doi.org/10.1101/478628


with no expression in any PB Tregs were NR4A1 (51.4%) and CD177 (20.2%). A summary is 64

shown for the top eight upregulated (Figure 1d) or downregulated (Figure 1e) in TI-Tregs shown 65

as log2-fold change (LFC). 66

67

We also analyzed a recently-published single-cell profile of T cells in HCC 9, comprised of flow-68

sorted T cells including Tregs from PB, normal liver parenchyma, a transitional zone near the 69

tumor, and TI-Tregs. After clustering (Supplemental Figure 1a-b), we performed the same 70

differential gene analysis comparing TI-Tregs versus PB-Tregs (Supplemental Figure 1c). We 71

found a total of 273 differential genes in our ccRCC-infiltrating Tregs, 467 differential genes in 72

the HCC-infiltrating Tregs, and 143 shared differential genes (Figure 1f, Supplemental Figure 73

1c). Of note, we corroborated previous reports of increased CCR8, LAYN, and MAGEH1 74

expression in TI-Tregs (Figure 1f-g)7–9. The gene with the highest expression in ccRCC and 75

HCC TI-Tregs was CD177. A similar pattern of upregulated genes, including CD177, was seen 76

in pooled Tregs in breast, colorectal, and lung cancers (Supplemental Figure 1d-e). 77

78

Recent studies started to imply functional heterogeneity of FOXP3+ Tregs in PB11 as well as 79

colorectal cancer12 and glioma13, based on genomic studies with pooled Tregs. Using SCRS, we 80

were able to investigate the dynamic transcriptomic processes of individual Tregs. Using the 81

Monocle 2 algorithm14, we constructed a manifold using all the Tregs in ccRCC (Figure 2a) and 82

HCC (Supplemental Figure 2a). This technique orders the cells by expression pattern to 83

represent distinct cellular fates, with the ordinal construction creating a pseudo-time variable 84

that allows us to investigate changes in gene expression during the infiltration process. We 85

observed a bifurcated architecture of TI-Tregs in ccRCC (Figure 2a) and HCC (Supplemental 86

Figure 2a), implying two distinct cell fates. We also found a number of genes with significant 87

roles in the manifold ordering (Figure 2b), with complete results available in Supplemental Table 88


https://doi.org/10.1101/478628


2. A similar profile of genes was involved in manifold construction of the HCC Tregs with 89

reduced separation between TI versus PB Tregs in several genes including CTLA4 and CCR8 90

(Supplemental Figure 2b). In contrast, other genes like CCL20 and CD177 maintained distinct 91

separation in expression from TI versus PB Tregs (Supplemental Figure 2b). 92

93

Next, to understand the branching structure of the manifold, we performed differential gene 94

expression analysis. Markers of immune regulation had increased expression in TI-Tregs of the 95

first cell fate (CF1) compared to the Tregs of the second cell fate (CF2, Figure 2c, lower two 96

clusters). In contrast, the CF2 had a maintenance of ribosomal gene expression (Figure 2c). We 97

observed three trends in gene expression between the TI-Tregs cell fates: non-specifically 98

increased, like CCR8 and CTLA4; increased in CF1, like CD177 and TNFRSF4; and increased 99

in CF2, like CXCR4 and EGR1 (Figure 2d). To investigate the potential functional difference 100

between the cell fates, we performed gene signature analysis (Figure 2e). TI-Tregs had a 101

reduction in the naïve T cell signature and an increase in T cell exhaustion signature compared 102

to PB Tregs in both ccRCC (Figure 2e, upper row) and HCC (Supplemental Figure 2d). In 103

contrast, the cytotoxicity and cell cycle signatures were significantly higher in CF1 Tregs (Figure 104

2e, lower row). We assigned cell phases to single cells, finding an increased percentage of G2M 105

phase Tregs in in CF1 (Figure 2f). This increase in G2M was not demonstrated in the CF1 Tregs 106

in the HCC dataset, indicative of cancer-specific properties of TI-Tregs (Supplemental Figure 107

2d-e). We also observed a clonal expansion in TI-Tregs across all ccRCC patients 108

(Supplemental Figure 3a-c), in line with previous observations8,9,15–17. In addition, we noticed 109

that the clonal expansion of the T cell receptor repertoire was enriched in the Tregs of CF1 110

(Supplemental Figure 3d). 111

112

The increased expression of CD177 in both ccRCC and HCC Tregs compared to PB Tregs 113


https://doi.org/10.1101/478628


(Figure 1g) and the specificity towards the suppressive CF1 (Figure 3a-b) led us to further 114

determine the functional impact of CD177+ Tregs. CD177 is a glycosylphosphatidylinositol-115

linked surface protein that is expressed on neutrophils18 and has been used as a biomarker for 116

myeloproliferative diseases19. Our interest in CD177 is long standing and we have previously 117

shown that CD177 plays a role in neutrophil viability20. We performed flow cytometry on tumor-118

infiltrating T lymphocytes (TITL) and demonstrated that a majority of CD177+ cells were Tregs 119

(Supplementary Figure 4a-b). There were 22.4% CD177+ Tregs among TI-Tregs in breast 120

cancer and 16.8% in renal cancer (Figure 3c). There were negligible percentages of 121

conventional CD4+ T cells (Tconv) expressing CD177 in breast and renal cancers (Figure 3c). 122

Additionally, we developed an immunohistochemical (IHC) protocol (Supplemental Figure 4c-d), 123

finding CD177+ lymphocytic infiltration was enriched in the colon, lung, liver, and prostate. In 124

addition to the single-color IHC, we further characterized the lymphocytic infiltration in breast 125

tumors using dual-IHC staining, identifying CD177 and FOXP3 double-positive cells 126

(Supplemental Figure 4e). Further analysis revealed that CD177+ Tregs had larger pools of cells 127

expressing PD-1, CTLA-4 and CCR8 when compared to CD177- Tregs, suggesting an active 128

and suppressive phenotype (Figure 3d). CD27, an activation marker, exhibited a similar pattern 129

for CD177+ Treg or Tconv cells (52.7% vs 59.5%), but was lost in CD177- Tregs (Figure 3d). We 130

were also able to isolate a fraction of CD4+ conventional T cells (Tconv) with CD177 expression 131

failed to show a clear differential pattern compared to CD177- Tconv cells (Figure 3d), indicative 132

of a Treg-specific phenotype related to CD177 expression. 133

134

In order to confirm the immunosuppressive property of CD177+ Tregs, we performed in vitro 135

proliferation and suppression assays from patient-derived TITL cells (Figure 3e, Supplemental 136

Figure 5a). Results showed that CD177+ Tregs had a 4-fold greater suppressive effect on CD4+ 137

effector T cells when plated at a 2:1 ratio compared to CD177- Tregs (73% versus 16%) (Figure 138


https://doi.org/10.1101/478628


3E, Supplemental Figure 5A). The scarcity of CD177+ tumor infiltrating Tregs made it 139

challenging to repeat these experiments. However, given that the experiments were carried out 140

using Tregs pooled from three individual breast cancer tissues, results can be considered 141

biologically representative. 142

143

We used Cd177-KO mice and their WT littermate controls to verify the suppressive phenotype 144

of CD177+ Tregs. After titrating injections to identify optimal cell numbers to facilitate immune-145

based tumor control of Py8119 cells, a breast cancer model, we found an optimal dose of 500 146

cells. We observed a clear phenotype in regard to tumor growth in Cd177-KO mice versus WT, 147

where KO mice rejected tumor inoculation (5 of 10 total) and showed significantly slower tumor 148

growth (Figure 3f). The same increase in tumor rejection (4 of 7) and slower tumor growth was 149

seen in the syngeneic graft of MC38 colorectal cells in Cd177-KO versus WT controls (Figure 150

3g). To confirm that the observed phenotype is, in part, due to the suppressive action of CD177+ 151

Tregs in WT mice, we performed an in vitro immunosuppression assay using Tregs sorted from 152

splenocytes of tumor-bearing Cd177-KO and WT mice. Tregs isolated from WT mice were more 153

suppressive of CD4+ and CD8+ T cell proliferation compared with Cd177-KO Tregs 154

(Supplemental Figure 5b). 155

156

As our data suggests the transcriptional and functional difference in a suppressive subset of 157

tumor-infiltrating Tregs across several cancers, we hypothesized that gene signature 158

development from SCRS data would provide improved prognostic ability. We performed feature 159

selection to identify sets of genes most associated with overall survival for: 143 differentially-160

expressed genes of TI-Treg, 222 genes differentially-expressed in CF1, and 86 differentially-161

expressed genes in CF2 using the ccRCC dataset from the Cancer Genome Atlas (TCGA, 162

Figure 4a). Using 50% of the ccRCC samples as a training set, we trained supervised support 163

vector machines to discriminate between survival outcomes. Applying these signatures to the 164


https://doi.org/10.1101/478628


remaining 50% of samples, we showed the TI (Figure 4b) and CF1 signatures (Figure 4c) 165

discriminated prognosis in ccRCC, but not the CF2-based signature (Figure 4d). Despite both 166

the TI and CF1 signature significantly predicting poor overall survival in roughly 20% ccRCC 167

patients, CF1 signature has a superior ability to discriminate prognosis (hazard ratio of 3.22 168

versus 1.91) (Figure 4b-c). 169

170

The superior discrimination of CF1 Treg signature compared to the TI-Treg signature was 171

observed across a number of cancers. We applied the ccRCC-signatures across the 24 largest 172

TCGA datasets, finding that the TI-Treg signature significantly separated prognostic groups in 7 173

cancer datasets, with hazard ratio range limited to 1.39 to 3.37 (Figure 4e). In contrast, the CF1 174

Treg signature discriminated prognostic groups in 8 cancer types, with a larger hazard ratio 175

range of 1.4 to 5.5 (Figure 4f). The greater number and range of significant predictions by 176

cancer was also seen when comparing disease-specific survival in CF1 Treg versus TI-Treg 177

signature (Supplemental Figure 6a-b). CF2 Treg signature failed to discriminate groups based 178

on overall survival in any TCGA dataset (Supplemental Figure 6c). Both the CF1 Treg and the 179

TI-Treg signature separated prognostic groups in immune-checkpoint-inhibitor responsive 180

melanoma (SKCM) and lung adenocarcinoma (LUAD) (Figure 4e-f). The CF1 Treg signature 181

separated prognostic groups in sarcoma (SARC), low-grade glioma (LGG) and colon 182

adenocarcinoma (COAD). Both the CF1 Treg and TI-Treg signature had larger hazard ratios 183

than previously identified tumor Treg markers, like FOXP3, and ratio-based signatures 184

CCR8:FOXP3 or CCR8:CD3G (Figure 4g)7,8,21. It should be noted that feature selection of the 185

CF1 did not include CD177 into the gene signature, likely due to the expression of CD177 in 186

epithelial tissue (Figure 4h, Supplementary Figure 4). 187

188

In conclusion, although infiltration of FoxP3+ Tregs are thought to suppress antitumor immune 189


https://doi.org/10.1101/478628


response broadly, our work demonstrates transcriptional and functional heterogeneity with TI-190

Tregs. Using SCRS of 734 Tregs, we showed two distinct transcriptional fates of TI-Tregs 191

(Figure 2a). Although a transcriptional analysis of pooled TI-Treg was previously found 192

incidentally,8 we herein identified CD177 as a marker of a suppressive subset of Tregs with 193

superior ability to inhibit effector T cell populations and tumor growth in both breast and 194

colorectal models. Comparing prognostic ability for Treg gene signatures, we demonstrated the 195

superior outcome prediction for the suppressive CF1 signature in demarcating prognostic 196

groups in ccRCC and across the TCGA compared to the TI-Treg gene signature or CF2 Treg 197

signature (Figure 4). Our study displays the potential of using SCRS to focus investigation of 198

specific immune subsets and potentially identify new therapies or prognostic markers. Further 199

understanding of the development and regulation of this suppressive subset of TI-Tregs may be 200

crucial in the development of cancer immunotherapies with minimal autoimmune side effects. 201

202

Authors' Contributions 203

Conception and design: W.Z., N.B. K.K.A, X.W., Y.Z. 204

Development of methodology: N.B., K.K.A., P.K., K.G.C., A.V. 205

Acquisition of data: K.K.A., N.B., R.K., A.V., P.K., G.P., K.G.C., J.K.T., Y.W.Z., J.L., X.H., X.W., 206

W.Z. 207

Analysis and interpretation of data: N.B., K.K.A., R.K., A.V., X.W., Y.W.Z., W.Z. 208

Writing, review, and/or revision of the manuscript: N.B., K.K.A, R.K., Y.W.Z, S.G.Z, Y.Z., W.Z. 209

Study supervision: X.W., Y.Z., W.Z. 210

211

Acknowledgments 212

We thank Breast Molecular Epidemiologic Resource (BMER led by Dr. Sonia Sugg) and Tissue 213

Procurement Core at the University of Iowa/Carver College of Medicine for providing breast 214


https://doi.org/10.1101/478628


cancer tissues; TCGA for providing breast cancer, melanoma and other cancer RNA-seq 215

datasets and clinical parameters; Comparative Pathology Laboratory from Department of 216

Pathology University of Iowa/Carver College of Medicine for developing CD177 IHC protocol. 217

218


https://doi.org/10.1101/478628


References 219

1. Schmidt, A. M. et al. Regulatory T Cells Require TCR Signaling for Their Suppressive 220 Function. J. Immunol. (2015). doi:10.4049/jimmunol.1402384 221

2. Fontenot, J. D., Gavin, M. A. & Rudensky, A. Y. Foxp3 programs the development and 222 function of CD4+CD25+ regulatory T cells. Nat. Immunol. 4, 330–336 (2003). 223

3. Hori, S., Nomura, T. & Sakaguchi, S. Control of Regulatory T Cell Development by the 224 Transcription Factor. Science (80-. ). 299, 1057 LP-1061 (2003). 225

4. Liu, S. et al. Prognostic significance of FOXP3+ tumor-infiltrating lymphocytes in breast 226 cancer depends on estrogen receptor and human epidermal growth factor receptor-2 227 expression status and concurrent cytotoxic T-cell infiltration. Breast Cancer Res. (2014). 228 doi:10.1186/s13058-014-0432-8 229

5. Shang, B., Liu, Y., Jiang, S. & Liu, Y. Prognostic value of tumor-infiltrating FoxP3+ 230 regulatory T cells in cancers: a systematic review and meta-analysis. Sci. Rep. (2015). 231 doi:10.1038/srep15179 232

6. Sasaki, a et al. Prognostic value of tumor-infiltrating FOXP3+ regulatory T cells in 233 patients with hepatocellular carcinoma. Eur. J. Surg. Oncol. (2008). 234 doi:10.1016/j.ejso.2007.08.008 235

7. De Simone, M. et al. Transcriptional Landscape of Human Tissue Lymphocytes Unveils 236 Uniqueness of Tumor-Infiltrating T Regulatory Cells. Immunity (2016). 237 doi:10.1016/j.immuni.2016.10.021 238

8. Plitas, G. et al. Regulatory T Cells Exhibit Distinct Features in Human Breast Cancer. 239 Immunity (2016). doi:10.1016/j.immuni.2016.10.032 240

9. Zheng, C. et al. Landscape of Infiltrating T Cells in Liver Cancer Revealed by Single-Cell 241 Sequencing. Cell (2017). doi:10.1016/j.cell.2017.05.035 242

10. Miao, D. et al. Genomic correlates of response to immune checkpoint therapies in clear 243 cell renal cell carcinoma. Science (80-. ). 359, 801–806 (2018). 244

11. Miyara, M. et al. Functional Delineation and Differentiation Dynamics of Human CD4+T 245 Cells Expressing the FoxP3 Transcription Factor. Immunity (2009). 246 doi:10.1016/j.immuni.2009.03.019 247

12. Saito, T. et al. Two FOXP3+CD4+ T cell subpopulations distinctly control the prognosis of 248 colorectal cancers. Nat. Med. (2016). doi:10.1038/nm.4086 249

13. Lowther, D. E. et al. PD-1 marks dysfunctional regulatory T cells in malignant gliomas. 250 JCI Insight (2016). doi:10.1172/jci.insight.85935 251

14. Qiu, X. et al. Reversed graph embedding resolves complex single-cell trajectories. Nat. 252 Methods 14, 979–982 (2017). 253

15. Sainz-Perez, A., Lim, A., Lemercier, B. & Leclerc, C. The T-cell receptor repertoire of 254 tumor-infiltrating regulatory T lymphocytes is skewed toward public sequences. Cancer 255 Res. (2012). doi:10.1158/0008-5472.CAN-12-0277 256


https://doi.org/10.1101/478628


16. Savas, P. et al. Single-cell profiling of breast cancer T cells reveals a tissue-resident 257 memory subset associated with improved prognosis. Nat. Med. (2018). 258 doi:10.1038/s41591-018-0078-7 259

17. Guo, X. et al. Global characterization of T cells in non-small-cell lung cancer by single-cell 260 sequencing. Nat. Med. (2018). doi:10.1038/s41591-018-0045-3 261

18. Kissel, K., Santoso, S., Hofmann, C., Stroncek, D. & Bux, J. Molecular basis of the 262 neutrophil glycoprotein NB1 (CD177) involved in the pathogenesis of immune 263 neutropenias and transfusion reactions. Eur. J. Immunol. (2001). doi:10.1002/1521-264 4141(200105)31:5<1301::AID-IMMU1301>3.0.CO;2-J 265

19. Stroncek, D. F. Neutrophil-specific antigen HNA-2a, NB1 glycoprotein, and CD177. 266 Current Opinion in Hematology (2007). doi:10.1097/MOH.0b013e3282efed9e 267

20. Xie, Q. et al. Characterization of a novel mouse model with genetic deletion of CD177. 268 Protein Cell (2015). doi:10.1007/s13238-014-0109-1 269

21. Wolf, D. et al. The expression of the regulatory T cell-specific forkhead box transcription 270 factor FoxP3 is associated with poor prognosis in ovarian cancer. Clin. Cancer Res. 271 (2005). doi:10.1158/1078-0432.CCR-05-1244 272

273


https://doi.org/10.1101/478628


Figure Legends: 274

Figure 1. TI-Tregs display distinct expression program compared to PB controls in 275

ccRCC SCRS dataset. 276

(a). tSNE projection of immune cells from three ccRCC patients with normal PB cells (n=13433) 277

and TI cells (n=12,239). Treg population (blue) was isolated and separated as TI (orange) 278

versus PB Tregs (grey). (b). tSNE projection with highlighted expression of Treg markers, 279

FOXP3 and IL2RA (CD25). (c). Differential gene expression analysis using the log2-fold 280

change expression versus the difference in the percent of cell expressing the gene comparing 281

TI versus PB Tregs (∆ Percentage Difference). Genes labeled have log2-fold change > 1, ∆ 282

Percentage Difference > 20% and adjusted P-value from Wilcoxon rank sum test < 0.05. (d). 283

Top eight upregulated genes by log2-fold change in TI-Tregs with adjusted P-value < 0.05. (e). 284

Top eight downregulated genes by log2-fold change in TI-Tregs with adjusted P-value < 0.05. 285

(f). Comparison of differential genes in TI-Tregs in ccRCC (orange) and HCC (green) compared 286

to PB Tregs. Significant genes were defined as log2-fold change > 1 or < -1 with adjusted P-287

values < 0.05. (g). Relative mRNA level of Treg markers in PB (grey) and TI-Tregs in ccRCC 288

(top) and HCC (bottom). 289

290

Figure 2. Bifurcation in the transcriptional state of TI-Tregs reveal a more suppressive 291

cell fate. 292

(a). Trajectory manifold of Tregs from the renal clear cell carcinoma using the Monocle 2 293

algorithm, solid and dotted line represent distinct cell trajectories/fates defined by SCRS 294

expression profiles. (b). Pseudo-time projections of transcriptional changes in immune genes 295

based on the manifold. Significance based on differential testing by site of origin which was also 296

used to generate pseudo-time and adjusted for multiple comparisons. (c). Expression heatmap 297

of significant (Q < 1e-6) genes based on branch expression analysis comparing the two TI cell 298

fates and were used in the ordering of the pseudo-time variable. (d). Cell trajectory projections 299


https://doi.org/10.1101/478628


of transcriptional changes in immune genes based on the manifold. Significance based on 300

differential testing between the first and second cell fates of TI-Tregs. ̅ denotes the scaled 301

mean mRNA levels of each pole of the manifold. (e). SCRS gene signature analysis of the 302

poles of the trajectory manifold. P-value based on one-way ANOVA with individual comparisons 303

corrected for multiple hypothesis testing using the Tukey HSD method. * P < 0.05, ** P <0.01, 304

*** P <0.001, **** P < 0.0001. (f). Results of the cell cycle regression analysis of single cells for 305

each cell fate using the Seurat R package. 306

307

308

Figure 3. CD177 is a marker a suppressive subpopulation of TI-Tregs. 309

(a). Trajectory manifold of Tregs from the ccRCC Tregs with the number of CD177+ and CD177- 310

Tregs for each respective cell fate. Significance based on χ2 testing comparing the three poles 311

of the manifold. (b). Proportional distribution of CD177+ Tregs by cell fate across the manifold. 312

(c). Percentages of CD177+ cells relative to different CD4+ TILs in breast cancer (n=13-18) and 313

renal cancer (n=8) showing Data are presented as mean ± SD with median values presented in 314

the figure. (d). Representative flow cytometry data for select marker expression by CD177+ or 315

CD177- Tregs (CD3+CD4+CD25+CD127low/FoxP3+) or Tconv (D3+CD4+CD25-/FoxP3-) isolated 316

from breast cancers. n=2-3. (e). CD177+ TI-Tregs are suppressive and inhibit CD8 T cell 317

proliferation. Total, CD177+, or CD177- TI-Tregs were purified from fresh human breast cancer 318

specimens (combined from 3 patients) using flow cytometry and co-cultured with naïve CD8 T 319

cells from PBMC for ex vivo suppression assay at the indicated ratios. (f). Py8199 tumor growth 320

is significantly reduced in Cd177-KO mice compared to WT, P < 0.0001 (two-way ANOVA) in 321

mice challenged with 5x102 cells per inoculation, n=10 bilateral tumors. Numbers in parenthesis 322

equates to the number of mice which developed palpable tumors/total mice inoculated. Data are 323

presented as mean ± SEM. (g). MC38 tumor growth is significantly reduced in Cd177-KO mice 324


https://doi.org/10.1101/478628


compared to WT, P = 0.0005 (two-way ANOVA) in mice challenged with 5x104 cells per 325

inoculation. Numbers in parenthesis equates to the number of mice which developed palpable 326

tumors/total mice inoculated. Data are presented as mean ± SEM. 327

328

Figure 4. Improved prognostic prediction associated with signature from suppressive 329

Treg cell fate. 330

(a). Schematic of signature development using feature selection from: 1) 143 differential genes 331

of TI-Treg in ccRCC and HCC, 2) 86 genes differentially expressed in CF2, and 3) 222 genes 332

differentially expressed in CF1 using the ccRCC (n=538) dataset from the TCGA. (b-d). Kaplan-333

Meier curves for overall survival in ccRCC using the TITL gene signature (b) , in ccRCC using 334

the CF1 Treg gene signature (c), and the CF2 Treg signature (d). P-value based on log-rank 335

test and hazard ratio (HR) based on Cox proportional hazard regression. (e-f). Overall survival 336

prediction with Cox proportional hazard ratio and -log10(P-value) based on log-rank testing 337

across the 24 largest TCGA datasets using the TI-Treg signature (e) and the CF1 signature (f). 338

(g). Prognostic prediction for SCRS Treg signatures compared to other proposed signatures for 339

TI-Tregs. Hazard ratios, 95% confidence intervals, and P-values derived from Cox proportional 340

hazard regression modeling. (h). Relative mRNA violin plots of the CF1 signature based on the 341

transcriptional trajectory state across the 160 PB-Tregs and 574 TI-Tregs. 342

343

344

345

346


https://doi.org/10.1101/478628


Peripheral Blood Treg

01

2

3

4

56

7

8

910

11

12

13

14

1516

17

18

1920

21

22

23

24−25

0

25

−25 0 25

Com

pone

nt 2

Component 1

−25

0

25

FOXP3

Com

pone

nt 2

−25 0 25

IL2RA (CD25)

0123

log mR

NA

HSPA1B

NR4A1

CCR8

CDKN1A

LAYN

RHOB

SERTAD1

ATF3

CXCR6

NDFIP2

EGR1

CISH

STIP1

NR4A3

CD177

ABI3

CLEC7A

−2

0

2

4

6

-50 -25 0 25 50Δ Percentage Difference

Log-

Fold

Cha

nge

−25 0 25

Component 1

Regulatory T cells

Upregulated in Peripheral Tregs

Upregulated in Tumor-Infiltrating Tregs

a

b

c

Log-Fold Change

d

e

130

324

143

Hepatocellular Carcinoma

Renal Clear Cell Carcinoma

f

Tumor-infiltrating Treg

g

CDHR3PI16

RP11−342D11.3KLF3

VSIG1MGAT4A

CRIP2C16orf74

−3 −2 −1 0

RHOBAC145110.1

GEMEGR1CD177

HSPA1ANR4A1

HSPA1B

0 2 4 6

p=2.4e-5

p=6.7e-11p=1.32e-5p=1.15e-35

p=3.41e-35p=5.80e-25

p=3.5e-5

p=5.5e-14

p=2.03e-15p=1.56e-13p=7.24e-16p=9.21e-31p=3.86e-7p=5.3e-32p=2.0e-10

p=9.5e-8

TNFRSF4 TNFRSF9NR4A1 TNFRSF18

LAYN

MAGEH1

CTLA4 ICOSCCR8 CD177

Rel

ativ

e m

RN

A L

evel

ccRC

CH

CCTNFRSF4 TNFRSF9NR4A1 TNFRSF18

LAYN

MAGEH1

CTLA4 ICOSCCR8 CD177

Figure 1 .CC-BY-NC 4.0 International licenseis made available under aThe copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It. https://doi.org/10.1101/478628doi: bioRxiv preprint

https://doi.org/10.1101/478628


1

−4

−2

0

2

−5 0 5 10

Component 1

Com

pone

nt 2

Peripheral Treg

Tumor-Infiltrating Treg

Figure 2

a

b

0 5 10 15 20Pseudo-time

Cell Fate #1

Cell Fate #2

−2 −1−3 0 1 2 3

Ribsomal Genes

JUNBCXCR4CD69EGR1JUNFOSBHSP Genes

GZMHNKG7GZMB

CCL4

LGALS1CD74TNFRSF18TNFRSF4IL1R2CD177GAPDHENTPD1CCL20TNFCCL22DUSP4TNFRSF9LAYNCD27TNFRSF1BMAGEH1

RelativeExpression

Cell Fate #1 Cell Fate #2c

CXCR4

EGR1

−5 10

Component 1

CCR8

CTLA4−4

−2

0

2

CD177

TNFRSF4

−4

−2

0

2Com

pone

nt 2

0 5 −5 100 5

d

RelativeExpression

Low High

TNFRSF18 TNFRSF4 TNFRSF9

LAYN MAGEH1 TIGIT

FOS FOXP3 IKZF2

CD177 CTLA4 DUSP4

CCL20 CCL4 CCR8

0.1

1.0

10.0

1

10

1

1

10

0.1

1.0

10.0

0.11.0

10.0

1

10

1

10

1

10

1

10

0.1

10.0

0.01

1.00

1

10

100

0.1

1.0

1

10

Pseudo−time

q=4.83e-33 q=6.78e-25

q=1.79e-150 q=1.28e-55q=3.58e-1540 5 10 150 5 10 15 20 0 5 10 15 20 20

Rela

tive m

RN

A E

xpre

ssio

n

q=1.15e-55 q=2.62e-39

q=7.66e-63

q=2.55e-44

q=1.09e-6

q=1.40e-25

−5 100 5

q=3.50e-10

q=2.57e-26

q=1.60e-13

q=4.69e-34

q=0.020

q=0.025

q=8.12e-41

q=1.74e-173 q=2.62e-39

q=4.25e-65

0

25

50

75

100

Perc

ent

of

Sam

ple

f

G1 Phase S Phase

G2/M Phase

Fate #2Fate #1PBMC

186 4984

151 3674

102 2329

(45%)(42%)(45%)

(21%)(23%)(16%)

(33%)(34%)(40%)

X=0.79

X=0.35

X=-1.13

X=0.79

X=0.33

X=-1.12

X=1.03

X=-0.061

X=-0.97

X=1.04

X=-0.080

X=-0.96

X=0.032

X=0.98

X=-1.02

X=0.26

X=0.84

X=-1.02

e

Cytotoxicity Cell Cycle

Naive Exhaustion

0.0

0.5

1.0

0.0

0.5

1.0

PB #1 #2

Sig

natu

re S

core

p<2e-16 p<2e-16

p=9.5e-3p<9.6e-9

**** * **

***

****

PB #1 #2


https://doi.org/10.1101/478628


PD-1 CTLA-4 CCR8 CD27

52.7%

0%

59.5%

0.54%

75%

40%

3.45%

0.35%

33.3%

0%

1.7%

0.04%

27.5%

6.2%

9.5%

6.28%

150K

100K

50K

0K

150K

100K

50K

0K

150K

100K

50K

0K

150K

100K

50K

0K

CD

177+

CD

177-

CD

177+

CD

177-

Treg

sTc

onv

Cou

nts

CFSE

Teff/Treg 8:1

Teff/Treg 4:1

Teff/Treg 2:1

No Treg

CD177+ CD177- Total

25 27 31 33 35 380

500

1000

1500

Days Post-Injection

Tum

or V

olum

e (m

m3 )

P<0.0001

Cd177-WT (9/10)Cd177-KO (5/10)

Figure 3

d e f

CD4+

T CellT Conv Treg

% C

D17

7+ T

Cel

l

n.s.

Breast CancerRenal Cancer

0

20

40

60P<0.002

0

20

40

60

n.s.2.4%

0.31%

22.4%

P=0.052

P<0.0001

% C

D17

7+ T

Cel

l

4.5%

0.72%

16.8%

0 5 10 15 20 25 30 35

Cd177-WT (6/6)Cd177-KO (3/7)

0

500

1000

1500

P=0.0005

Days Post-Injection

Tum

or V

olum

e (m

m3 )

g

5x102 Py8119 cells

5x104 MC38 cells

−4

−2

0

2

−5 0 5 10Component 1

Com

pone

nt 2

CD177-CD177+

n=6n=102

n=110n=329

n=0n=187

P < 0.00001

Cell Fate #1

Cell Fate #2

0.00

0.25

0.50

0.75

1.00Fate #1 Fate #2 PBMC

Pro

porti

on o

f Cel

ls

74.9% 94.4% 100%

25.1% 5.6%

CD177-CD177+

a b c


https://doi.org/10.1101/478628


0 1000 2000 3000 4000

p=0.0051HR=1.910.00

0.25

0.50

0.75

1.00

Good (n=216)

Bad (n=51)

Tumor-Infiltrating Signature

p<0.0001HR=3.22

Good (n=215)

Bad (n=52)

0 1000 2000 3000 40000.00

0.25

0.50

0.75

1.00Cell Fate #1 Signature

p=0.16HR=1.52

Good (n=241)

Bad (n=26)

0 1000 2000 3000 40000.00

0.25

0.50

0.75

1.00Cell Fate #2 Signature

Ove

rall

Sur

viva

l

Days

KIRCLIHC KIRP

LAML

LUADUCECSKCM

0.0

0.5

1.0

1.5

2.0

−1.0 −0.5 0.0 0.5 1.0log(Cox Hazard Ratio)

KIRC

LGGKIRPSARC

LIHC

COADLUADSKCM

0

2

4

6

0.0 0.5 1.0 1.5

-log1

0(P

-val

ue)

-log1

0(P

-val

ue)

log(Cox Hazard Ratio)

Significant

Tumor-Infiltrating Signature

Ove

rall

Sur

viva

lO

vera

ll S

urvi

val

Cell Fate #1 Signature

Significant

b

c

d

e f

g

a

Cell Fate #1

Cell Fate #2

Tumor-InfiltratingTregs

ccRCC

HCC

Testing Set 50% of samples

Training Set 50% of samples

Support Vector MachineTCGA ccRCC

N=538

Feature Selectionto 7- 8 genes

143 genes

86 genes

222 genes

1

2

3Apply to Rest

of TCGA

Poo

r Pro

gnos

is

Poo

r Pro

gnos

is

Figure 4

CCR8|FOXP3

CCR8

CCR8|CD3G

FOXP3

TI Treg

Cell Fate #1

Cox Hazard Ratio

0.00133

0.225

0.021

0.430

0.00592

1.07e-7PrognosisP-value Poor

1 2 3 4 5

BANF1 CCL22 ITM2A

NDUFC2 PTTG1 RHOB SHMT2

Rel

ativ

e m

RN

A

Cell Fate #1 Signature

CF #1 Treg

Pre TI-Treg

CF #2 Treg

h


https://doi.org/10.1101/478628


transcriptional heterogeneity in cancer-associated …37 [email protected] (yz). 38 conflicts of...

Documents