Downloaded Using TCGA

Home , Pazopanib

bioRxiv preprint doi: https://doi.org/10.1101/2020.05.01.071613; this version posted May 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1 Pazopanib induces dramatic but transient contraction of myeloid suppression 2 compartment in favor of adaptive immunity 3 4 Darawan Rinchai1α, Elena Verzoni2α, Veronica Huber3, Agata Cova3, Paola Squarcina3, Loris 5 De Cecco4, Filippo de Braud2, Raffaele Ratta2, Matteo Dugo4, Luca Lalli3, Viviana Vallacchi3, 6 Monica Rodolfo3, Jessica Roelands1, Chiara Castelli3, Damien Chaussabel5, Giuseppe 7 Procopio2, Davide Bedognetti1,6Ω, Licia Rivoltini3Ω 8 9 α/ Ω Contributed equally 10 11 1Cancer Research Department, Sidra Medicine, Doha, Qatar 12 2Hopital Foch, Medical Oncology Department, Suresnes, France 13 3Unit of Immunotherapy of Human Tumors, Fondazione IRCCS Istituto Nazionale dei Tumori, 14 Milan, Italy 15 4Platform of Integrated Biology, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy 16 5Immunology Department, Sidra Medicine, Doha, Qatar 17 6Dipartimento di Medicina Interna e Specialità Mediche, Università degli Studi di Genova, 18 Genova, Italy 19 20 21 Corresponding authors 22 Licia Rivoltini, MD 23 Unit of Immunotherapy of Human Tumors 24 Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy 25 [email protected] 26 27 Davide Bedognetti, MD, PhD 28 Cancer Research Department 29 Sidra Medicine, Doha, Qatar 30 [email protected] 31

40 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.01.071613; this version posted May 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1 Abstract

2 Anti-angiogenic tyrosine-kinase inhibitors (TKIs) and immune checkpoint blockade (ICB) 3 constitute the backbone of metastatic renal cell carcinoma (mRCC) treatment. The 4 development of the optimal combinatorial or sequential approach is hindered by the lack of 5 comprehensive data regarding TKI-induced immunomodulation and its kinetics. Through the 6 use of orthogonal transcriptomic and phenotyping platforms combined with functional analytic 7 pipelines, we demonstrated that the anti-angiogenic TKI pazopanib induces a dramatic and 8 coherent reshaping of systemic immunity in mRCC patients, downsizing the myeloid-derived 9 suppressor cell (MDSC) compartment in favor of a strong enhancement of cytotoxic T and 10 Natural Killer (NK) cell effector functions. The intratumoral expression level of a MDSC 11 signature here generated was strongly associated with poor prognosis in mRCC patients. The 12 marked but transient nature of this immunomodulation, peaking at the 3rd month of treatment, 13 provides the rationale for the use of TKIs as a preconditioning strategy to improve the efficacy 14 of ICB. 15 16 Keywords: (3-10) 17 Renal cell cancer, tyrosine kinase inhibitors, immunotherapy, immunosuppression, myeloid- 18 derived suppressor cells, PBMCs, Pazopanib, blood transcriptomic profile, transcriptional 19 modular repertoire analysis, antiangiogenics 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52

1 Abbreviation List 2 3 DC Dendritic cell 4 iDC Immature dendritic cell 5 mDC Myeloid dendritic cells 6 pDC Plasmacytoid dendritic cells 7 ICB Immune checkpoint blockade 8 IFN Interferon 9 IPA Ingenuity Pathway Analysis 10 MDSC Myeloid derived suppressor cells 11 G-MDSC Granulocytic Myeloid derived suppressor cells 12 M-MDSC Monocytic Myeloid derived suppressor cells 13 mRCC Metastatic renal cell carcinoma 14 NK Natural killer cell 15 PCA Principal component analysis 16 PBMCs Peripheral blood mononuclear cells 17 Tcm Central memory T cell 18 Tem Effector memory T cells 19 Tfh T follicular helper cells 20 Th2 cells T helper 2 cells 21 Th1 cells T helper 1 cells 22 Th17 cells T helper 17 cells 23 Tgd T gamma delta cells 24 Treg Regulatory T cell 25 TKI Tyrosine-kinase inhibitor 26 VEGF Vascular endothelial growth factor 27

1 Introduction

2 The off-target activity of cancer treatments on systemic immunity is a well-known process,

3 deemed to play an active role in disease control [1]. In the course of conventional therapies

4 including chemotherapy or radiotherapy, a reshaping of spontaneous immune response

5 toward a boost of tumor-specific T cells might be required to achieve significant clinical benefit

6 [2]. Such a reinvigoration of adaptive T cell immunity stems from a network of effects triggered

7 by tumor cell death, amplified by additional mechanisms involving multiple immune pathways

8 depending on the drug’s mechanism of action [3].

9 The immunomodulatory properties of cancer therapies are recently gaining attention in

10 view of potential combinations with immunotherapy such as immune checkpoint blockade

11 (ICB). The nowadays consolidated evidence that ICB mediates effective tumor control only in

12 a minority of patients and in selected malignancies, points to the use of drug combinations as

13 a strategy to increment ICB effectiveness [4]. Promising results showing increased efficacy of

14 PD-1 blockers combined with chemotherapy in non-small cell lung cancer (NSCLC) [5] and

15 breast cancer [6, 7], along with multiple clinical trials ongoing in different tumor types, suggest

16 that combinatorial approaches hold promise to become the potential gold standard of treatment

17 in different settings.

18 However, the effects of cancer therapies on the multiple components of tumor immunity

19 might be complex and variegated and need to be carefully considered when combination

20 strategies based on desired synergies are designed. Antineoplastic treatments might directly

21 potentiate tumor immunogenicity by broadening antigenic repertoire or favoring antigen

22 presentation that boost T cell priming [5]; or they can act indirectly by reducing myeloid-derived

23 suppressor cell (MDSC)-mediated immunosuppression as a beneficial outcome of their

24 common myelotoxicity [8]. Conversely, according to in vitro and/or in vivo experimental studies,

25 anti-proliferative therapeutic strategies, particularly those based on the inhibition of multiple

26 tyrosine kinases, might affect T cell proliferation and function as well, thus potentially interfering

27 with the protective activity of adaptive immunity [9–12]. Hence, gaining detailed information on

28 the type and kinetics of the immunomodulating properties of anticancer drugs would be 4

1 essential to maximize clinical efficacy when diverse therapeutic strategies are combined with

2 immunotherapy.

3 Given the complexity of the human immune system and its dynamic nature,

4 immunomonitoring approaches are moving toward multiplex and “omics” strategies, with first

5 results emerging in autoimmunity and viral infections [13, 14]. Transcriptomic analysis of

6 peripheral blood [15, 16] for instance, has been extensively used to dissect mechanisms of

7 action of vaccination against infectious diseases [17], to elucidate pathogenic mechanisms of

8 different immunologic disorders [18, 19], and to identify perturbations associated with different

9 viral [20], parasitic [21], or bacterial infections [18, 22]. However, such an approach remains

10 relatively unexplored in the context of cancer therapy, including immunotherapy [23].

11 Pioneering studies in cancer patients treated with interleukin (IL)-2 have contributed to the

12 characterization of systemic changes induced by this cytokine [24–26]. More recently,

13 peripheral blood transcriptomic analysis has been used to identify signatures associated with

14 responsiveness to anti-CTLA4 [27], and to describe changes differentially associated with

15 CTLA4 and combined CTLA4/PD-1 blockade [23, 28]. To the best of our knowledge, no

16 peripheral blood transcriptomic studies have been performed so far in solid cancer patients

17 receiving tyrosine kinase inhibitors (TKI) or in renal cell carcinoma (RCC) patients treated with

18 any other drugs. According to first-line randomized 3 metastatic RCC trials using PD-1

19 blockades, up to 50-60% of patients do not respond to such treatment [29–31]. Recently, a

20 retrospective analysis of the CheckMate 025 phase III trial demonstrated a remarkable

21 increased overall survival (HR, 0.60; 95% CI,.0.42–0.84) in patients previously treated with

22 first-line pazopanib [32].

23 In the present work, we applied an integrative analysis encompassing transcriptional

24 profiling (leucocyte subtype abundance estimation, functional characterization by pathway

25 analysis and modular repertoire analysis) and multiplex flow cytometry, to comprehensively

26 capture the immunomodulation mediated by anti-angiogenic treatment in mRCC patients.

27 RCC was specifically chosen for its potent immunosuppressive properties linked to

28 hypoxia/VHL-related oncogenic pathways that lead to the secretion of proangiogenic factors 5

1 known to mediate the blunting of adaptive antitumor immunity [33]. Anti-angiogenic drugs

2 interfering with one or multiple factors of the VEGF, PDGF, HGF and ANG2 cascade, are

3 endowed with the intrinsic ability to overcome VEGF-mediated immunosuppression [34, 35].

4 Here, we investigated the entity and the kinetics of immunomodulation mediated in mRCC

5 patients by pazopanib, a TKI anti-angiogenic drug included in the standard care of this disease

6 [36–38].

7 By performing a matched analysis of transcriptional and phenotypic profiling in

8 peripheral blood mononuclear cells (PBMCs) obtained before and at different time points

9 during therapy, we demonstrated, for the first time, that pazopanib mediates a potent but

10 transient reprogramming of systemic immunity, resulting in an enhanced T and NK cytotoxic

11 response accompanied by an attenuation of the myeloid suppressor compartment. Our study

12 shows that monitoring systemic immunity by transcriptomics may help designing drug-tailored

13 combination strategies aimed at maximizing clinical efficacy thanks to the timely engagement

14 of a full-fledged immune response.

16 Results

17 Integrative transcriptional analysis reveals the immune modulatory properties of

18 pazopanib.

19 The study has been conducted on clear cell mRCC patients treated with first line pazopanib,

20 whose PBMCs were obtained from blood withdrawn at baseline, 3- and 6-months during

21 treatment. Transcriptional profiling was analyzed using integrative and complementary

22 pipelines.

23 We first wanted to explore the molecular heterogeneity of the sample set through

24 principal component analysis (PCA) based on the whole transcriptomic profile (12,913 genes)

25 (Figure 1A). The first 3 major PCs accounted for 20.7% (PC1), 10.7% (PC2), and 8.0% (PC3)

26 of the variability observed for these conditions. These three-dimensional plots showed the

27 distribution of individual patient samples for each time point with no outlier sample Figure 1A.

28 In general, a certain degree of separation according to different time points was observed. 6

1 We then performed differential expression analysis between post-treatment versus

2 pre-treatment samples. Two hundred and thirteen transcripts were significantly different

3 between 3 months post-treatment (Post 3) and pre-treatment (Pre), 47 transcripts between 6

4 months and 3 months post-treatment (Post 6 vs Post 3), and 98 transcripts between Post 6

5 and Pre. Volcano plots showing log2 fold-change (log2FC) and p-value (paired t-test) on

6 differentially expressed genes are shown in Figure 1B, and Supplementary Table 1.

7 Strikingly, among the top 40 genes ranked according to the log2FC, the large majority was

8 associated with cytotoxic functions and interferon signaling (Table 1). Representative

9 transcripts related to T and NK cells cytotoxic functions and T cells activation (e.g., CD8A,

10 CD160, GZMB, GZMH, KLRD1, KLRB1, NRC3, LAG3, IL12RB1, KIR2DL1, NCR1, NKG7,

11 PRF1 and GNLY) are represented in Figure 1C. The over-expression of such transcripts was

12 attenuated at 6 months after treatment. The common transcripts significantly up-regulated at

13 both Post 3 and Post 6 (N=16) as compared to pre-treatment include CD8A, CTSW, NCAM1

14 (CD56), NCR3, NKG7, consistent with a persistence but attenuated NK and T cell response

15 [39] (Figure 1B, and Supplementary Table 1).

17 Functional interpretation of transcriptomic changes induced by pazopanib

18 The top ten differentially modulated canonical pathways in post-treatment vs pre-treatment

19 samples are shown in Figure 2A-C. Genes of the top 3 pathways in each comparison are

20 shown in Figure 2D. The graphical representation of the top pathway at each time point

21 comparison is shown in Supplementary Figure 1. The majority of the top canonical pathways

22 modulated by pazopanib (7/10 in both Post 3 and Post 6 comparison) were associated with

23 immune functions (Figure 2A-D). The perturbations induced at the 3rd month of treatment are

24 consistent with the triggering of NK/cytotoxic signaling, the positive modulation of the crosstalk

25 between dendritic cells and NK cells, the regulation of IL-2, T cell receptor (TCR) signaling,

26 and IL-8 signaling. After 3 further months of treatment, an attenuation of the immune

27 modulatory effect induced by pazopanib was observed. This was substantiated by the

28 downregulation of transcripts associated with T helper (Th)-1 and Th2 functional orientation 7

1 when comparing Post 6 vs Post 3 samples. The activation of NK-related pathways was still

2 sustained at the 6th montht of treatment, although attenuated.

4 Pazopanib-induced molecular fingerprints

5 We applied modular repertoire analysis to further dissect the immune-modulatory properties

6 of pazopanib. The percentage of responsive transcripts constitutive of a given module was

7 determined at each time point (see methods for details). The group comparison analysis

8 confirmed that module perturbations peaked at 3 months of treatment and decreased at 6

9 months. These perturbations include upregulation of modules M3.6 and M8.46 defining

10 cytotoxic/NK cells, M4.11 (plasma cells), and M8.89 (immune response). Moreover, the

11 responsiveness of M4.14 (monocytes) was decreased (Figure 3). However, mapping

12 perturbations of the modular repertoire for a group of subjects does not account for the

13 heterogeneity observed at the individual level. We therefore performed deeper individual-level

14 analysis. This approach demonstrated that pazopanib administration was associated with the

15 decrease of M9.34 (immunosuppressive) in the majority of patients. The most coherent

16 changes were represented by modulations of cytotoxic/NK cells modules (M3.6 and M8.46)

17 while the majority of the other modules demonstrated a considerable heterogeneity.

18 Interestingly, a rapid increase of Interferon (IFN) modules (M1.2 and M3.4) was observed

19 exclusively in patients displaying up-regulation of cytotoxic/NK cells modules (Figure 4,

20 Supplementary Table 2)

22 Modulation of leukocyte functional orientation induced by pazopanib as derived by

23 transcriptomic data

24 To estimate the changes in leukocyte populations, we used single sample Gene Set

25 Enrichment Analysis (ssGSEA). Comparison of the enrichment scores of post-treatment and

26 their baseline pre-treatment, showed that NKCD56dim, NKCD56bright, T gamma-delta (Tgd),

27 NKT, cytotoxic cells and CD8 T cells increased coherently at 3 months of treatment and

28 subsequently slightly decreased without reaching baseline levels (Figure 5). Conversely, 8

1 regulatory T cells (Tregs) signature was decreased and a similar trend was observed for

2 MDSCs (Figure 5C). These results suggest that pazopanib induces synergistic immune

3 modulations by enhancing protective immunity and reducing suppressive mechanisms.

5 Flow cytometry analysis confirms the positive immune modulation associated with

6 pazopanib administration

7 Transcriptome profiling in bulk cell populations provided a high-level and unbiased perspective

8 on the changes taking place following initiation of therapy. It is ideally complemented by flow

9 cytometry analyses which provide a targeted but highly granular view of changes taking place

10 at the cellular and protein levels.

11 Multicolor flow cytometry analysis of PBMCs was performed concomitantly in biological

12 replicates of the same blood samples submitted to transcriptional profiling, plus an additional

13 patient for whom RNA was not available. A broad panel of markers encompassing innate and

14 adaptive immune cell subsets of the lymphoid and myeloid repertoire was studied and

15 modulation in on-treatment with respect to baseline samples was assessed. The analysis

16 showed that pazopanib administration was associated with a remarkable increase of activated

17 and cytolytic effectors including the subset of activated T cells (CD3+PD-1+), reported to

18 contain tumor-specific T cells [40], activated NK cells (CD3-CD16+CD56+PD-1+) and cytotoxic

19 NK cells (CD3-CD16+CD56dim) (Figure 6) [41]. Of note, this evidence is in line with the data

20 that emerged from the transcriptional profiling, depicting an overall boost of genes involved in

21 TCR signaling, cytotoxic cell populations and NK activity. Again, similarly to findings obtained

22 via transcriptomic analyses, the detected changes over baseline were more evident at 3

23 months of therapy and tended to a plateau or decreased at 6 months. The boost of T and NK

24 cell activation was paralleled by a significant decrease in the frequency of different myeloid

25 cell subsets, including CD14+ monocytes, MONO-MDSCs (CD14+HLA-DRneg) [42],

26 inflammatory monocytes (CD14+PDL-1+) [43], and PMN-MDSCs (CD15+) [42](Figure 6A).

27 Regulatory T cells (CD4+CD25highFoxp3+) also displayed a remarkable reduction during

28 treatment. Down-modulation of all these cell subsets was mostly detectable at 3 months during 9

1 treatment with respect to baseline, with a stabilization of monocytic MDSCs or/and a rebound

2 for total and inflammatory monocytes in cell frequencies at 6 months (Figure 6). Taken

3 together, the kinetics of immune modulation as detected by flow cytometry are in line with the

4 data emerging from the transcriptional profiling data and confirm the transient nature of

5 immunomodulation mediated by pazopanib.

7 Intratumoral estimates of MDSC are associated with poor prognosis in kidney cancer

8 One of the more remarkable findings obtained through combined transcriptomic and flow-

9 cytometry based immune monitoring is the decrease in MDSC populations and associated

10 signatures.

11 To explore the relevance of our observation, and as no data exists regarding the

12 prognostic role of MDSCs in kidney cancer, we assessed the expression of the three MDSC

13 signatures in The Cancer Genome Atlas (TCGA) clear cell RCC cohort (KIRC, N=517, Figure

14 7). The MDSC_INT signatures was strongly associated with decreased overall survival

15 (MDSC_INT High vs LowMed, HR =2.057 (95% CI = 1.52-2.79, Figure 7A). In particular,

16 MDSC High group had poor prognosis, while MDSC Low and Med group (Supplementary

17 Figure 2) have similar favorable prognosis. No such differences were observed using the

18 other two MDSC signatures MDSC_Angel and G-MDSC, suggesting that MDSC_INT, which

19 was developed experimentally based on extracellular-driven monocyte-MDSC differentiation,

20 might represent a novel prognostic biomarker in kidney cancer. Remarkably, MONO-MDSCs

21 were strongly suppressed after pazopanib treatment (Figure 6). MDSC_INT correlates with

22 Stage and Grade, which are major prognostic factors in kidney cancer (Figure 7B). We then

23 assessed the relationship between MDSC_INT with the disposition of oncogenic pathways,

24 and found that MDSC_INT expression linearly correlates with many oncogenic processes

25 associated with cancer aggressiveness, including angiogenesis (R = 0.59, p < 2e-16), and

26 epithelial-to-mesenchymal transition (EMT) (R = 0.75, p < 2e-16, Figure 7C-D) although

27 there was no overlapping between MDSC_INT signature and angiogenesis or EMT

28 (Supplementary Figure 3). Despite the correlation with Stage and Grade, MDSC_INT 10

1 retained its prognostic value even when included in a Cox regression multivariable model

2 (Figure 7E).

3 Thus, the analysis of the tumor specimens from the TCGA cohort permitted to expand

4 our initial finding by providing indirect evidence about the role of MDSCs in mRCC

5 progression. Indeed, this in turn suggests that MDSC suppression by pazopanib may be one

6 of the means by which treatment could contribute to improved outcome.

8 Discussion

9 This is the first translational study that investigates longitudinally the immunomodulatory

10 effects of an anti-angiogenic therapy on PBMCs of mRCC patients through transcriptomic

11 analysis. Our results show that pazopanib triggers cytotoxic cells and IFN pathways and

12 relieves immunosuppression by reducing MDSCs. This invigoration of anti-tumor immunity

13 was mostly evident after 3 months of therapy and still detectable, although less accentuated,

14 at the 6th month of treatment. The transcriptional profiling of PBMCs clearly revealed

15 treatment-induced immunomodulation, detecting modifications validated by flow cytometry,

16 but also expanding them by revealing pathway networks and broader functional information.

17 Our data indicates that the analysis of transcriptional profiles of blood cells with the support of

18 appropriate deconvolution approaches represents a valid strategy for monitoring immune cell

19 behavior at a high throughput and reliable level.

20 To achieve the results reported here, one of the major challenges was the identification

21 of gene-signatures appropriate to capture the activity of circulating MDSCs. Indeed, monocytic

22 MDSCs are defined in flow cytometry only by the lack/low expression of HLA-DR in cells

23 expressing the monocytic marker CD14, alone or in combination with CD11b and CD33 [42],

24 but their genomic features have been poorly defined. To this aim, we used the dataset selected

25 by Angelova [44] and Fridlender [45] as reference transcriptional data for myeloid cells and

26 the data set obtained from human MDSCs generated in vitro according to a model developed

27 in our laboratory [46]. This MDSC model was produced by exposing blood CD14+ monocytes

28 to tumor extracellular vesicles, a process leading to cells highly overlapping for phenotype, 11

1 immunosuppressive function and transcriptional profiles with MDSCs isolated from blood of

2 melanoma patients [46–48]. The gene signature reflects most of the signaling pathways

3 expected for these cells and overlap with monocytes sorted from cancer patients [46]. Our

4 MDSC signature, applied to bulk tumors, was the only one with prognostic implications in

5 mRCC, confirmed in multivariate analyses, providing here an essential contribution for the

6 estimations of MDSCs in different tissues.

7 Pazopanib is a multikinase inhibitor, among the ones of standard of care for first-line

8 treatment of mRCC patients, especially if they are "low risk" according to Heng criteria. This

9 TKI targets the VEGF receptor, platelet-derived factor receptor (PDGFR), KIT (proto-

10 oncogene receptor tyrosine kinase), fibroblast growth factor receptors (FGFR) and RAF

11 kinases [36–38]. The broad array of targets largely justifies the multifaceted

12 immunomodulating activity we observed in our study, which involved different myeloid cell

13 subsets together with T cells, NK cells and Tregs. However, given the known inhibitory activity

14 of several antiangiogenics on lymphocytes when tested in vitro [49], it is tempting to speculate

15 that this general immunological reshaping mostly stems from blunting of the complex

16 immunosuppressive pathways mediated by MDSC and Tregs. Indeed, pazopanib might exert

17 off-target inhibitory effects on MDSCs by acting on the VEGFR down-stream signaling [50,

18 51], or by interfering with the KIT pathway, another key node in myeloid cell function favoring

19 DC reprogramming [8]. Hence, the boost of T and NK cell activation and cytolytic functions we

20 broadly observed by both transcriptomics and flow cytometry, likely results from reduced

21 intratumoral immunosuppression leading to an enhanced anti-tumor immune response, rather

22 than from a stimulatory activity of the drug on these immune cell subsets.

23 A recent study demonstrated that pazopanib induces DC maturation in vitro [52], while

24 novel VEGF-directed drugs, such as axitinib, enhances the expression of NKG2D ligand and

25 consequently potentiates NK-cell cytolytic activity [53].

26 Previous studies in humans have shown that the anti-VEGF monoclonal antibody

27 sunitinib reduces the abundance of monocytes [54] (assessed 4 and 6 weeks post-treatment),

28 monocytic MDSCs [55] and Tregs [56] (assessed at 4 weeks after treatment only) in the blood. 12

1 However, the persistence of Tregs suppression (assessed at 1.5 and 3 months post-

2 treatment) has been correlated with prolonged overall survival [57]. There are no studies that

3 have assessed immunologic perturbations in patients receiving pazopanib as first line

4 treatment. Only one study has assessed immunologic changes in mRCC patients treated with

5 pazopanib, but administered as third-line treatment, therefore studying patients with a

6 potentially heavily compromised immune system. In such work, Pal et al [58] observed that

7 post-treatment non-responder patients had lower levels of HGF, VEGF, IL-6, IL-8 and soluble

8 IL-2R, and increased numbers of monocytic MDSCs as compared with responders. Failure to

9 detect an overall decrease of MDSCs as compared to baseline might result from the fact that

10 analyses were performed at late time points (at 6 and 12 months after treatment initiation). In

11 fact, our study showed that the changes induced by pazopanib partially revert after 6 months.

12 The number of patients analyzed here was rather small but reflected the rarity of RCC

13 patients who could be prospectively enrolled for first-line pazopanib administration especially

14 in research hospitals with competitive clinical trials enrolling. Even so, dynamic changes were

15 extremely coherent across patients and confirmed using orthogonal immune monitoring

16 platforms and analyses. The data reported here provide a set of key information that might

17 have relevant implications for the design of combinatory treatment strategies in mRCC clinical

18 setting. First, pazopanib mediates a specific reshaping of tumor immunity that should favor a

19 prompter response to immunotherapy due to the decrease of immunosuppressive effectors

20 and the concomitant boost of PD-1+ T cells and NK cells. Secondly, this effect reaches its

21 peak at the 3rd month of treatment but tends to be attenuated at later time points likely due to

22 the homeostatic mechanisms that regulate systemic immunity and tumor-mediated

23 immunosuppression. These data indicate that a short-term pre-conditioning treatment with

24 pazopanib might create the optimal immune setting for ICB to potentiate antitumor immune

25 responses in vivo. In this view, the combination of pazopanib with PD-1 blockers, which

26 resulted so far in unmanageable toxicity during initial clinical testing [59], could be replaced

27 by intermittent schedules that might help maximizing treatment synergies. The concomitant

28 administration of PD-1/PD-L1 inhibitors with anti-angiogenic monoclonal antibodies 13

1 (bevacizumab, IMmotion 151 trial) [60] or TKI axitinib (MK426 [61] and JAVELIN RENAL101

2 [62] trials), is rapidly emerging as a strategy to increase overall survival in RCC patients, with

3 further ongoing studies evaluating combination with novel TKI lenvatinib (CLEAR trial),

4 cabozantinib (CheckMate 9ER), and tivozanib [63] but the optimal schedule of such

5 combinations still needs to be defined. In this context, our data suggest that more dynamic

6 and innovative approaches based on intermittent or alternate schedules could be also

7 explored to ameliorate the therapeutic index of combinatorial regimens in cancer [64, 65].

9 Conclusion

10 Transcriptional profiling of blood immune cells, particularly if combined with specific

11 deconvolution programs to finely dissect the behavior of diverse immunological components,

12 represents a reliable tool to readily capture the immunomodulating properties of anticancer

13 therapies for designing scientifically-sound combination therapies. Thanks to this analysis we

14 found that pazopanib has a strong immune modulatory effect peaking at the 3rd month of

15 treatment and consists in relieved immunosuppression with enhanced cytotoxic T and NK

16 mechanisms and IFN pathways. Taken together with the detrimental role of MDSCs in mRCC

17 demonstrated in the TCGA cohort, these results provide a strong rationale for the use of TKIs

18 as preconditioning strategy to improve immunological and clinical efficacy of immunotherapy

19 in cancer patients.

21 Materials and Methods

22 Patients and study description

23 From January 2016 to June 2016 nine patients (8 males, 1 female) with metastatic clear cell

24 RCC were treated with first line pazopanib as per clinical practice in Istituto Nazionale dei

25 Tumori, Milan, Italy. Safety assessment included physical examination and laboratory tests

26 every month. All patients had a good performance status (ECOG 0:8/9, ECOG 1: 1/9), a

27 median age of 65 years and prevalence of intermediate risk according to Heng score (5/9).

28 They received pazopanib at a standard dose of 800 mg orally once daily, continuously, for at 14

1 least 6 months. All patients signed an informed consent according to a protocol approved by

2 the INT Ethical Committee [INT146/14].

4 Blood collection

5 Blood samples (30 ml) were obtained from 9 patients at baseline (Pre), and at the 3rd and 6th

6 month during therapy (Post 3 and Post 6). For one single patient, samples were collected only

7 at baseline and at 3 months therapy. Blood was processed within 1 hour from withdrawal.

8 PBMCs were separated by Ficoll gradient (Leuco-sep tubes, ThermoFisher Scientific) and

9 viable cells stored in liquid nitrogen until use, or frozen in Qiazole (Qiagen) for RNA extraction

10 and gene expression profiling.

12 Transcriptomic analysis

13 Suitable material for transcriptional analysis was available from 8 patients. RNA was extracted

14 using miRNAeasy kit (Qiagen). After quality check and quantification by 2100 Bioanalyzer

15 system (Agilent) and Nanodrop ND-1000 spectrophotometer (ThermoFisher), respectively,

16 RNA expression was assessed using Illumina HT12v4 BeadChip. Illumina's BeadStudio

17 version 1.9.0 software was used to generate signal intensity values from the scans. Data were

18 further processed using the Bioconductor “Lumi” package. Following background correction

19 and quantile normalization, expressions were log2-transformed for further analysis. Raw

20 expression and normalized data matrix have been deposited at NCBI’s Gene Expression

21 Omnibus database (http://www.ncbi.nlm.nih.gov/geo/), with accession numbers GSE146163.

22 From a total of 47,323 probes arrayed on the Illumina HT12v4 beadchip, the probes targeting

23 multiple genes were collapsed (average expression intensity) and a final data matrix

24 containing 12,913 unique genes was generated. Data analyses were performed using R

25 (Version 1.0.44, RStudio Inc.) and Ingenuity Pathway Analysis (QIAGEN Bioinformatics). The

26 comparison between each group (Post 3 vs pre, Post 3 vs Post 6 and Post 6 vs Pre) was

27 performed using the paired t-test. For detection of differentially expressed genes we used a p

28 value cutoff of 5x10-3, and false discovery rate (FDR) was provided as descriptive statistic 15

1 (Supplementary Table 1), and not to dictate significance as the risk of type I error was

2 mitigated by the use of orthogonal platforms (i.e., flow cytometry) for validation purposes.

3 Hierarchical clustering was performed using the function “Heatmap” from the R package

4 “ComplexHeatmap” [66]. Euclidean distance and complete linkage methods were used by

5 default. Principle component analysis (PCA) was performed using the R function

6 “scatterplot3d” package. The first three principal components, PC1, PC2 and PC3, were

7 plotted against each other.

9 Pathway analysis. Gene ontology analyses were performed using Ingenuity Pathway Analysis

10 (QIAGEN Bioinformatics). A relaxed p value cut-off of 0.05 was used to select transcripts for

11 pathway analysis. The proportion of upregulated and downregulated transcripts was

12 represented. The z-score was used to indicate the direction of pathway deregulation.

13 Transcripts from the top three pathways in each comparison group were plotted in the

14 corresponding heatmaps.

16 Leucocyte subset estimations. To estimate the enrichment of various cell types, gene

17 expression deconvolution analyses were performed with ssGSEA [1, 2] implemented in the

18 “GSVA package” using cell-specific signatures (Supplementary Table 3): T cells, CD8 T

19 cells, cytotoxic T cells, T helper 1 cells (Th1 cells), central memory T cells (Tcm), effector

20 memory T cells (Tem), T helper cells, T follicular helper cells (Tfh), Th2 cells, Th17 cells,

21 gamma delta T cells (Tgd), natural killer cells (NK cells), NK CD56dim, NK CD56bright cells, B

22 cells, neutrophils, eosinophils, mast cells [67], regulatory T cells (Treg), NKT cells, monocytes,

23 dendritic cells (DC), immature DCs (iDC), plasmacytoid DCs (pDC), myeloid DCs (mDC) [44].

24 For MDSCs, we constructed a specific signature based on 25 genes highly correlated in the

25 present dataset, selected from the top 100 genes upregulated in extracellular vesicle (EV)-

26 MDSCs vs monocytes (MDSC_INT, Supplementary Figure 4) in our recent work [46].

27 Additional MDSC signatures include the one proposed by Angelova et al. [44] (MDSC_Angel),

28 based on markers selected according to the literature, and a granulocytic myeloid-derived 16

1 suppressor cell (G-MDSC) signature defined by comparing G-MDSC vs naïve neutrophils [45].

2 Forest plots were plotted by using mean enrichment scores (ESs) ratio between Post 3 vs

3 Pre, Post 6 vs Post 3 and Post 6 vs Pre. Differentially expressed ESs between pre-treatment

4 and post-treatment were calculated through paired t-test (p < 0.05).

6 Modular repertoire analysis. A set of 260 modules (co-expressed genes) was used for the

7 analysis of this data set. This fixed modular repertoire was a priori determined, being

8 constructed based on co-expression measured across 9 reference datasets encompassing a

9 wide range of diseases (infectious, autoimmune, inflammatory) [17, 18, 68]

10 (https://github.com/Drinchai/DC_Module_Generation2). This data-driven approach allowed

11 the capture of a broad repertoire of immune perturbations which were subsequently subjected

12 to functional interpretation. This collection of annotated modules was then used as a

13 framework for analysis and interpretation of our blood transcriptome dataset. The approach

14 used for the construction, annotation and reuse of modular blood transcriptome repertoires

15 was previously reported [14, 15, 17, 18, 68]. After normalization, raw expression intensity was

16 used for the module analysis. Briefly, data was transformed from gene level data into module

17 (M) level activity scores, both for group comparison (Post 3 vs Pre, Post 6 vs Post 3 and Post

18 6 vs Pre) and individual patients comparison at each time point. The modules defined by this

19 approach (M1–M9, a total of 260 modules) were used as a framework to analyze and interpret

20 this dataset. For group comparisons, the expression profile at each time point was calculated

21 as a FC relative to a mean expression of all samples within that time points. Then, paired t-

22 test was used to evaluate each time point comparison. If the FC between each group

23 comparison was greater than 1, and the p value < 0.05, the transcript was considered as

24 upregulated. If the FC between each group comparison was less than 1, and the p value <

25 0.05, it was considered as downregulated. Then the percentages of “module responsiveness”

26 were calculated for each module. For individual comparison, the expression profile for each

27 individual patient was calculated as a FC and difference relative to the expression of individual

28 samples at each time point. If the FC between each time point comparison was more than 1, 17

1 and difference more than 10, the transcript was considered as upregulated. If the FC between

2 each time point comparison was less than 1, and the difference less than 10, it was considered

3 as downregulated. For both, group and individual comparisons, the “module-level” data is

4 subsequently expressed as a percent value representing the proportion of differentially

5 regulated transcripts for a given module. A module was considered to be responsive when

6 more than 15% of the transcripts were down or upregulated.

8 Multiparameter flow cytometry

9 PBMC samples from 9 patients were thawed and tested simultaneously for all time points by

10 flow cytometry. Phenotypic profiling was performed after labeling PBMCs with monoclonal

11 fluorochrome-conjugated antibodies: CD14 FITC (Clone M5E2, BD Pharmingen), CD3 FITC

12 (Clone UCHT1, BD Biosciences) or KO525 (Clone UCHT1, Beckman Coulter), PD-1 APC

13 (Clone MIH4, BD Pharmingen) or PC7 (Clone PD1.3, Beckman Coulter), HLA-DR APC (Clone

14 G46-6, BD Pharmingen), CD15 PerCP-CY5.5 (Clone HI98, BD Pharmingen), PD-L1 PE

15 (Clone MIH1, BD Pharmingen), CD4 PE (Clone RPA-T4, BD Pharmingen), CD25 PerCP-

16 Cy5.5 (Clone M-A251, BD Pharmingen), CD56 ALEXA750 (Clone N901, Beckman Coulter),

17 CD16 BV650 (Clone 3G8, BD Biosciences), Live/Dead Fixable Violet (ThermoFisher), FOXP3

18 APC (Clone FJK-16s eBioscience) used after cell permeabilization with the kit Perm Buffer

19 (10X) and Fix/Perm Buffer (4X) (BioLegend), according to manufacturer’s instructions.

20 Samples were incubated with Fc blocking reagent (Miltenyi Biotec) for 10 minutes at room

21 temperature before the addition of monoclonal antibodies for 30 minutes at 4°C. Thereafter,

22 samples were washed, fixed acquired by Gallios Beckman Coulter FC 500 or BD FACSCalibur

23 (BD Biosciences) flow cytometers, and analyzed with Kaluza software (Beckman Coulter).

24 Gating strategies are depicted in Supplementary Figure 5. Distinct cell subsets were

25 quantified in terms of frequency rather than absolute numbers, since the latter are influenced

26 by sampling manipulation procedures that are unrelated to biological patterns. Pre-treatment

27 and post-treatment samples (Post 3 vs Pre, Post 6 vs Post 3 and Post 6 vs Pre) were

28 compared by using paired t- test. 18

1 TCGA transcriptomic Analysis

2 RNA-seq data from TCGA clear cell RCC (KIRC) cohort were downloaded using TCGA

3 Assembler (v2.0.3). Data normalization was performed within lanes, and between lanes using

4 R package EDASeq (v2.12.0) and quantile normalized using preprocessCore (v1.36.0). A

5 single primary tumor sample was included per patient using the TCGA Assembler

6 “ExtractTissueSpecificSamples” function. Previously flagged samples that did not pass assay-

7 specific QCs were excluded [69]. Data was log2 transformed with an (+1) offset. Enrichment

8 scores (ES) were calculated by ssGSEA on the log2 transformed, normalized gene-level data.

9 Gene sets to define ES of tumor-associated pathways (n=51) were used as described

10 previously [70]. The correlation between tumor-associated signatures was calculated using

11 Pearson test and plotted using “corrplot” (v0.84).

13 Survival Analysis

14 Clinical data from the TCGA-KIRC cohort was obtained from the TCGA Pan-Cancer Clinical

15 Data Resource [71]. Patients were divided in tertiles based on enrichment scores of MDSC

16 gene signatures (MDSC_INT, G-MDSC, and MDSC_Angel). Overall survival (OS) was used

17 to generate Kaplan-Meier curves using a modified version of the ggkm function [72]. Survival

18 data were censored after a follow-up period of 10 years. Hazards ratios (HR) between groups,

19 corresponding p values, and confidence intervals were calculated using cox proportional

20 hazard regression with R package survival (v2.41-3).

22 Competing interests

23 EV reports personal fees for advisory boards from Pfizer, Ipsen, Novartis outside of the

24 submitted work. No potential conflicts of interest were disclosed by the other authors.

25 26 Acknowledgements

27 This work was supported by the Associazione Italiana per la Ricerca sul Cancro (AIRC)

28 Special Program Innovative Tools for Cancer Risk Assessment and early Diagnosis 5X1000

1 (no. 12162 to LR) and by Fondi 5x1000 Ministero della Salute 2015 (D/17/1VH to VH) and by

2 the Italian Ministry of Health (RF-2016-02363001 to G.P.). We would like to thank the patients

3 and participating study team for making this study happen. DB, DR, and DC work has been

4 supported by Sidra Medicine, Member of Qatar Foundation.

6 References;

7 1. Romano A, Parrinello NL, Chiarenza A, Motta G, Tibullo D, Giallongo C, et al. Immune off- 8 target effects of Brentuximab Vedotin in relapsed/refractory Hodgkin Lymphoma. Br J 9 Haematol. 2019;185:468–79.

10 2. Weiss T, Weller M, Roth P. Immunological effects of chemotherapy and radiotherapy 11 against brain tumors. Expert Rev Anticancer Ther. 2016;16:1087–94.

12 3. Galluzzi L, Buqué A, Kepp O, Zitvogel L, Kroemer G. Immunogenic cell death in cancer and 13 infectious disease. Nat Rev Immunol. 2017;17:97–111.

14 4. Wei SC, Duffy CR, Allison JP. Fundamental Mechanisms of Immune Checkpoint Blockade 15 Therapy. Cancer Discov. 2018;8:1069–86.

16 5. Gandhi L, Rodríguez-Abreu D, Gadgeel S, Esteban E, Felip E, De Angelis F, et al. 17 Pembrolizumab plus Chemotherapy in Metastatic Non-Small-Cell Lung Cancer. N Engl J Med. 18 2018;378:2078–92.

19 6. Schmid P, Cortes J, Pusztai L, McArthur H, Kümmel S, Bergh J, et al. Pembrolizumab for 20 Early Triple-Negative Breast Cancer. New England Journal of Medicine. 2020;382:810–21.

21 7. Schmid P, Adams S, Rugo HS, Schneeweiss A, Barrios CH, Iwata H, et al. Atezolizumab 22 and Nab-Paclitaxel in Advanced Triple-Negative Breast Cancer. New England Journal of 23 Medicine. 2018;379:2108–21.

24 8. Ostrand-Rosenberg S, Fenselau C. Myeloid-Derived Suppressor Cells: Immune- 25 Suppressive Cells That Impair Antitumor Immunity and Are Sculpted by Their Environment. J 26 Immunol. 2018;200:422–31.

27 9. Yue P, Harper T, Bacot SM, Chowdhury M, Lee S, Akue A, et al. BRAF and MEK inhibitors 28 differentially affect nivolumab-induced T cell activation by modulating the TCR and AKT 29 signaling pathways. OncoImmunology. 2019;8:e1512456.

30 10. Bedognetti D, Roelands J, Decock J, Wang E, Hendrickx W. The MAPK hypothesis: 31 immune-regulatory effects of MAPK-pathway genetic dysregulations and implications for 32 breast cancer immunotherapy. Emerging Topics in Life Sciences. 2017;1:429–45.

33 11. Dushyanthen S, Teo ZL, Caramia F, Savas P, Mintoff CP, Virassamy B, et al. Agonist 34 immunotherapy restores T cell function following MEK inhibition improving efficacy in breast 35 cancer. Nat Commun. 2017;8:606.

36 12. Gu Y, Zhao W, Meng F, Qu B, Zhu X, Sun Y, et al. Sunitinib impairs the proliferation and 37 function of human peripheral T cell and prevents T-cell-mediated immune response in mice. 38 Clin Immunol. 2010;135:55–62.

1 13. Hasin Y, Seldin M, Lusis A. Multi-omics approaches to disease. Genome Biology. 2 2017;18:83.

3 14. Chaussabel D, Pulendran B. A vision and a prescription for big data-enabled medicine. 4 Nat Immunol. 2015;16:435–9.

5 15. Chaussabel D, Baldwin N. Democratizing Systems Immunology with Modular 6 Transcriptional Repertoires Analyses. Nat Rev Immunol. 2014;14:271–80.

7 16. Pascual V, Chaussabel D, Banchereau J. A genomic approach to human autoimmune 8 diseases. Annu Rev Immunol. 2010;28:535–71.

9 17. Obermoser G, Presnell S, Domico K, Xu H, Wang Y, Anguiano E, et al. Systems scale 10 interactive exploration reveals quantitative and qualitative differences in response to influenza 11 and pneumococcal vaccines. Immunity. 2013;38:831–44.

12 18. Singhania A, Graham CM, Gabryšová L, Moreira-Teixeira L, Stavropoulos E, Pitt JM, et 13 al. Transcriptional profiling unveils type I and II interferon networks in blood and tissues across 14 diseases. Nat Commun. 2019;10:2887.

15 19. HIPC-CHI Signatures Project Team, HIPC-I Consortium. Multicohort analysis reveals 16 baseline transcriptional predictors of influenza vaccination responses. Sci Immunol. 2017;2.

17 20. Zhai Y, Franco LM, Atmar RL, Quarles JM, Arden N, Bucasas KL, et al. Host 18 Transcriptional Response to Influenza and Other Acute Respiratory Viral Infections – A 19 Prospective Cohort Study. PLOS Pathogens. 2015;11:e1004869.

20 21. Rinchai D, Presnell S, Vidal M, Dutta S, Chauhan V, Cavanagh D, et al. Blood Interferon 21 Signatures Putatively Link Lack of Protection Conferred by the RTS,S Recombinant Malaria 22 Vaccine to an Antigen-specific IgE Response. F1000Res. 2015;4:919.

23 22. Mahajan P, Kuppermann N, Mejias A, Suarez N, Chaussabel D, Casper TC, et al. 24 Association of RNA Biosignatures With Bacterial Infections in Febrile Infants Aged 60 Days or 25 Younger. JAMA. 2016;316:846–57.

26 23. Butterfield LH, Disis ML, Fox BA, Kaufman DR, Khleif SN, Wang E, et al. SITC 2018 27 workshop report: Immuno-Oncology Biomarkers: State of the Art. J Immunother Cancer. 28 2018;6:138.

29 24. Bedognetti D, Wang E, Sertoli MR, Marincola FM. Gene expression profiling in vaccine 30 therapy and immunotherapy for cancer. Expert Rev Vaccines. 2010;9:555–65.

31 25. Panelli MC, Wang E, Phan G, Puhlmann M, Miller L, Ohnmacht GA, et al. Gene-expression 32 profiling of the response of peripheral blood mononuclear cells and melanoma metastases to 33 systemic IL-2 administration. Genome Biol. 2002;3:RESEARCH0035.

34 26. Weiss GR, Grosh WW, Chianese-Bullock KA, Zhao Y, Liu H, Slingluff CL, et al. Molecular 35 insights on the peripheral and intra-tumoral effects of systemic high dose rIL-2 (Aldesleukin) 36 administration for the treatment of metastatic melanoma. Clin Cancer Res. 2011;17:7440–50.

37 27. Friedlander P, Wassmann K, Christenfeld AM, Fisher D, Kyi C, Kirkwood JM, et al. Whole- 38 blood RNA transcript-based models can predict clinical response in two large independent 39 clinical studies of patients with advanced melanoma treated with the checkpoint inhibitor, 40 tremelimumab. J Immunother Cancer. 2017;5:67.

1 28. 34th Annual Meeting & Pre-Conference Programs of the Society for Immunotherapy of 2 Cancer (SITC 2019): part 1. J Immunother Cancer. 2019;7 Suppl 1. doi:10.1186/s40425-019- 3 0763-1.

4 29. Motzer RJ, Tannir NM, McDermott DF, Arén Frontera O, Melichar B, Choueiri TK, et al. 5 Nivolumab plus Ipilimumab versus Sunitinib in Advanced Renal-Cell Carcinoma. New England 6 Journal of Medicine. 2018;378:1277–90.

7 30. Motzer RJ, Rini BI, McDermott DF, Arén Frontera O, Hammers HJ, Carducci MA, et al. 8 Nivolumab plus ipilimumab versus sunitinib in first-line treatment for advanced renal cell 9 carcinoma: extended follow-up of efficacy and safety results from a randomised, controlled, 10 phase 3 trial. Lancet Oncol. 2019;20:1370–85.

11 31. Rini BI, Battle D, Figlin RA, George DJ, Hammers H, Hutson T, et al. The society for 12 immunotherapy of cancer consensus statement on immunotherapy for the treatment of 13 advanced renal cell carcinoma (RCC). J Immunother Cancer. 2019;7:354.

14 32. Escudier B, Sharma P, McDermott DF, George S, Hammers HJ, Srinivas S, et al. 15 CheckMate 025 Randomized Phase 3 Study: Outcomes by Key Baseline Factors and Prior 16 Therapy for Nivolumab Versus Everolimus in Advanced Renal Cell Carcinoma. Eur Urol. 17 2017;72:962–71.

18 33. Labrousse-Arias D, Martínez-Alonso E, Corral-Escariz M, Bienes-Martínez R, Berridy J, 19 Serrano-Oviedo L, et al. VHL promotes immune response against renal cell carcinoma via 20 NF-κB–dependent regulation of VCAM-1. J Cell Biol. 2017;216:835–47.

21 34. Jászai J, Schmidt MHH. Trends and Challenges in Tumor Anti-Angiogenic Therapies. 22 Cells. 2019;8. doi:10.3390/cells8091102.

23 35. Niu G, Chen X. Vascular Endothelial Growth Factor as an Anti-angiogenic Target for 24 Cancer Therapy. Curr Drug Targets. 2010;11:1000–17.

25 36. Gril B, Palmieri D, Qian Y, Smart D, Ileva L, Liewehr DJ, et al. Pazopanib reveals a role 26 for tumor cell B-Raf in the prevention of HER2+ breast cancer brain metastasis. Clin Cancer 27 Res. 2011;17:142–53.

28 37. Sternberg CN, Davis ID, Mardiak J, Szczylik C, Lee E, Wagstaff J, et al. Pazopanib in 29 locally advanced or metastatic renal cell carcinoma: results of a randomized phase III trial. J 30 Clin Oncol. 2010;28:1061–8.

31 38. van der Graaf WTA, Blay J-Y, Chawla SP, Kim D-W, Bui-Nguyen B, Casali PG, et al. 32 Pazopanib for metastatic soft-tissue sarcoma (PALETTE): a randomised, double-blind, 33 placebo-controlled phase 3 trial. Lancet. 2012;379:1879–86.

34 39. Stoeckle C, Gouttefangeas C, Hammer M, Weber E, Melms A, Tolosa E. Cathepsin W 35 expressed exclusively in CD8+ T cells and NK cells, is secreted during target cell killing but is 36 not essential for cytotoxicity in human CTLs. Exp Hematol. 2009;37:266–75.

37 40. Gros A, Parkhurst MR, Tran E, Pasetto A, Robbins PF, Ilyas S, et al. Prospective 38 identification of neoantigen-specific lymphocytes in the peripheral blood of melanoma patients. 39 Nat Med. 2016;22:433–8.

40 41. Mariotti FR, Petrini S, Ingegnere T, Tumino N, Besi F, Scordamaglia F, et al. PD-1 in 41 human NK cells: evidence of cytoplasmic mRNA and protein expression. Oncoimmunology. 42 2019;8:1557030.

1 42. Bronte V, Brandau S, Chen S-H, Colombo MP, Frey AB, Greten TF, et al. 2 Recommendations for myeloid-derived suppressor cell nomenclature and characterization 3 standards. Nat Commun. 2016;7:12150.

4 43. Pan T, Zhou T, Li L, Liu Z, Chen Y, Mao E, et al. Monocyte programmed death ligand-1 5 expression is an early marker for predicting infectious complications in acute pancreatitis. Crit 6 Care. 2017;21:186.

7 44. Angelova M, Charoentong P, Hackl H, Fischer ML, Snajder R, Krogsdam AM, et al. 8 Characterization of the immunophenotypes and antigenomes of colorectal cancers reveals 9 distinct tumor escape mechanisms and novel targets for immunotherapy. Genome Biol. 10 2015;16. doi:10.1186/s13059-015-0620-6.

11 45. Fridlender ZG, Sun J, Mishalian I, Singhal S, Cheng G, Kapoor V, et al. Transcriptomic 12 Analysis Comparing Tumor-Associated Neutrophils with Granulocytic Myeloid-Derived 13 Suppressor Cells and Normal Neutrophils. PLoS One. 2012;7. 14 doi:10.1371/journal.pone.0031524.

15 46. Huber V, Vallacchi V, Fleming V, Hu X, Cova A, Dugo M, et al. Tumor-derived microRNAs 16 induce myeloid suppressor cells and predict immunotherapy resistance in melanoma. J Clin 17 Invest. 2018.

18 47. Valenti R, Huber V, Filipazzi P, Pilla L, Sovena G, Villa A, et al. Human tumor-released 19 microvesicles promote the differentiation of myeloid cells with transforming growth factor-beta- 20 mediated suppressive activity on T lymphocytes. Cancer Res. 2006;66:9290–8.

21 48. Filipazzi P, Valenti R, Huber V, Pilla L, Canese P, Iero M, et al. Identification of a new 22 subset of myeloid suppressor cells in peripheral blood of melanoma patients with modulation 23 by a granulocyte-macrophage colony-stimulation factor-based antitumor vaccine. J Clin 24 Oncol. 2007;25:2546–53.

25 49. Stehle F, Schulz K, Fahldieck C, Kalich J, Lichtenfels R, Riemann D, et al. Reduced 26 immunosuppressive properties of axitinib in comparison with other tyrosine kinase inhibitors. 27 J Biol Chem. 2013;288:16334–47.

28 50. Kusmartsev S, Eruslanov E, Kübler H, Tseng T, Sakai Y, Su Z, et al. Oxidative stress 29 regulates expression of VEGFR1 in myeloid cells: link to tumor-induced immune suppression 30 in renal cell carcinoma. J Immunol. 2008;181:346–53.

31 51. Yang J, Yan J, Liu B. Targeting VEGF/VEGFR to Modulate Antitumor Immunity. Front 32 Immunol. 2018;9. doi:10.3389/fimmu.2018.00978.

33 52. Zizzari IG, Napoletano C, Botticelli A, Caponnetto S, Calabrò F, Gelibter A, et al. TK 34 Inhibitor Pazopanib Primes DCs by Downregulation of the β-Catenin Pathway. Cancer 35 Immunol Res. 2018;6:711–22.

36 53. Morelli MB, Amantini C, Santoni M, Soriani A, Nabissi M, Cardinali C, et al. Axitinib induces 37 DNA damage response leading to senescence, mitotic catastrophe, and increased NK cell 38 recognition in human renal carcinoma cells. Oncotarget. 2015;6:36245–59.

39 54. van Cruijsen H, van der Veldt AAM, Vroling L, Oosterhoff D, Broxterman HJ, Scheper RJ, 40 et al. Sunitinib-induced myeloid lineage redistribution in renal cell cancer patients: CD1c+ 41 dendritic cell frequency predicts progression-free survival. Clin Cancer Res. 2008;14:5884– 42 92.

1 55. Ko JS, Zea AH, Rini BI, Ireland JL, Elson P, Cohen P, et al. Sunitinib mediates reversal of 2 myeloid-derived suppressor cell accumulation in renal cell carcinoma patients. Clin Cancer 3 Res. 2009;15:2148–57.

4 56. Finke JH, Rini B, Ireland J, Rayman P, Richmond A, Golshayan A, et al. Sunitinib reverses 5 type-1 immune suppression and decreases T-regulatory cells in renal cell carcinoma patients. 6 Clin Cancer Res. 2008;14:6674–82.

7 57. Adotevi O, Pere H, Ravel P, Haicheur N, Badoual C, Merillon N, et al. A decrease of 8 regulatory T cells correlates with overall survival after sunitinib-based antiangiogenic therapy 9 in metastatic renal cancer patients. J Immunother. 2010;33:991–8.

10 58. Pal SK, Hossain DMS, Zhang Q, Frankel PH, Jones JO, Carmichael C, et al. Pazopanib 11 as third line therapy for metastatic renal cell carcinoma: clinical efficacy and temporal analysis 12 of cytokine profile. J Urol. 2015;193:1114–21.

13 59. Amin A, Plimack ER, Ernstoff MS, Lewis LD, Bauer TM, McDermott DF, et al. Safety and 14 efficacy of nivolumab in combination with sunitinib or pazopanib in advanced or metastatic 15 renal cell carcinoma: the CheckMate 016 study. J Immunother Cancer. 2018;6:109.

16 60. Rini BI, Powles T, Atkins MB, Escudier B, McDermott DF, Suarez C, et al. Atezolizumab 17 plus bevacizumab versus sunitinib in patients with previously untreated metastatic renal cell 18 carcinoma (IMmotion151): a multicentre, open-label, phase 3, randomised controlled trial. 19 Lancet. 2019;393:2404–15.

20 61. Rini BI, Plimack ER, Stus V, Gafanov R, Hawkins R, Nosov D, et al. Pembrolizumab plus 21 Axitinib versus Sunitinib for Advanced Renal-Cell Carcinoma. N Engl J Med. 2019;380:1116– 22 27.

23 62. Motzer RJ, Penkov K, Haanen J, Rini B, Albiges L, Campbell MT, et al. Avelumab plus 24 Axitinib versus Sunitinib for Advanced Renal-Cell Carcinoma. N Engl J Med. 2019;380:1103– 25 15.

26 63. Alonso-Gordoa T, García-Bermejo ML, Grande E, Garrido P, Carrato A, Molina-Cerrillo J. 27 Targeting Tyrosine kinases in Renal Cell Carcinoma: “New Bullets against Old Guys.” Int J 28 Mol Sci. 2019;20.

29 64. Schlemmer M, Spek A, Rodler S, Schott M, Casuscelli J, Staehler M. Sequential Treatment 30 Based on Sunitinib and Sorafenib in Patients with Metastatic Renal Cell Carcinoma. Cureus. 31 2019;11:e4244.

32 65. Fukumura D, Kloepper J, Amoozgar Z, Duda DG, Jain RK. Enhancing cancer 33 immunotherapy using antiangiogenics: opportunities and challenges. Nat Rev Clin Oncol. 34 2018;15:325–40.

35 66. Gu Z, Eils R, Schlesner M. Complex heatmaps reveal patterns and correlations in 36 multidimensional genomic data. Bioinformatics. 2016;32:2847–9.

37 67. Bindea G, Mlecnik B, Tosolini M, Kirilovsky A, Waldner M, Obenauf AC, et al. 38 Spatiotemporal dynamics of intratumoral immune cells reveal the immune landscape in human 39 cancer. Immunity. 2013;39:782–95.

40 68. Chaussabel D, Quinn C, Shen J, Patel P, Glaser C, Baldwin N, et al. A modular analysis 41 framework for blood genomics studies: application to systemic lupus erythematosus. 42 Immunity. 2008;29:150–64.

1 69. Thorsson V, Gibbs DL, Brown SD, Wolf D, Bortone DS, Ou Yang T-H, et al. The Immune 2 Landscape of Cancer. Immunity. 2018;48:812-830.e14.

3 70. Roelands J, Hendrickx W, Kuppen PJK, Mall R, Zoppoli G, Saad M, et al. Genomic 4 landscape of tumor-host interactions with differential prognostic and predictive connotations. 5 bioRxiv. 2019;:546069.

6 71. Liu J, Lichtenberg T, Hoadley KA, Poisson LM, Lazar AJ, Cherniack AD, et al. An 7 Integrated TCGA Pan-Cancer Clinical Data Resource to Drive High-Quality Survival Outcome 8 Analytics. Cell. 2018;173:400-416.e11.

9 72. Abhijit. Stat Bandit. https://statbandit.wordpress.com/author/aikiadg/. Accessed 18 Apr 10 2020.

1 Table 1: Top 40 of differentially expressed between post-treatment 3 months vs pre-treatment

Symbol Gene name P-value log2FC GZMB granzyme B 0.00169 0.726 KLRD1 killer cell lectin like receptor D1 0.00192 0.651 GNLY granulysin 0.00029 0.649 KLRF1 killer cell lectin like receptor F1 0.00308 0.648 PRF1 perforin 1 0.00461 0.646

CLIC3 chloride intracellular channel 3 0.00150 0.637 KIR2DL3 killer cell immunoglobulin like receptor, two Ig 0.00060 0.630 GZMH granzyme H 0.00394 0.616 NKG7 naturaldomains killer and cell long granule cytoplasmic protein tail 7 1 0.00137 0.601 KIR2DL4 killer cell immunoglobulin like receptor, two Ig 0.00194 0.570 CST7 cystatin F 0.00140 0.564 CTSW cathepsindomains and W long cytoplasmic tail 4 0.00072 0.556 KIR2DL1 killer cell immunoglobulin like receptor, two Ig 0.00097 0.525 KIR3DL1 killer cell immunoglobulin like receptor, three Ig 0.00077 0.507 CD160 CD160domains molecule and long cytoplasmic tail 1 0.00001 0.461 KIR3DL2 killerdomains cell andimmunoglobulin long cytoplasmic like receptor, tail 1 three Ig 0.00044 0.461 RNF165 ring finger protein 165 0.00081 0.458 TTC38 tetratricopeptidedomains and long repeat cytoplasmic domain tail 38 2 0.00058 0.447 KIR2DS5 killer cell immunoglobulin like receptor, two Ig 0.00126 0.431 LAG3 lymphocyte activating 3 0.00004 0.427 PYHIN1 pyrindomains and and HIN long domain cytoplasmic family member tail 3 1 0.00450 0.394 SPON2 spondin 2 0.00374 0.377 HOXC4 homeobox C4 0.00006 0.363 PTGDR prostaglandin D2 receptor 0.00059 0.362 KLRC1 killer cell lectin like receptor C1 0.00128 0.361 IL18RAP interleukin 18 receptor accessory protein 0.00240 0.358 AGAP1 ArfGAP with GTPase domain, ankyrin repeat and PH 0.00055 0.355 NCR3 natural cytotoxicity triggering receptor 3 0.00160 0.354 CBLB Cbldomain proto 1- oncogene B 0.00497 0.348 NCAM1 neural cell adhesion molecule 1 0.00007 0.339 TSEN54 tRNA splicing endonuclease subunit 54 0.00068 0.339 SDF2L1 stromal cell derived factor 2 like 1 0.00118 0.339 PDZD4 PDZ domain containing 4 0.00247 0.338 PRSS30P serine protease 30, pseudogene 0.00237 0.334 NMUR1 neuromedin U receptor 1 0.00020 0.329 IFITM1 interferon induced transmembrane protein 1 0.00223 0.328 FKBP11 FK506 binding protein 11 0.00128 0.319 KIR3DL3 killer cell immunoglobulin like receptor, three Ig 0.00152 0.318 HOPX HOP homeobox 0.00217 0.310 CD8A CD8adomains molecule and long cytoplasmic tail 3 0.00348 0.304 2

5 26

1 Figure legends

2 Figure 1: Transcriptional analysis of PBMCs from mRCC patients treated with

3 pazopanib. (A) Principle component analysis (PCA) of all patient samples color-coded by time

4 of treatment (left) and individual patient (right). (B) Volcano plots of differentially expressed

5 genes between pre- and post-treatment (Post 3 vs Pre, Post 6 vs Post 3 and Post 6 vs Pre);

6 the horizontal line represent cut off at p <0.005 and vertical lines represent fold-change > 1.2

7 (right) or < -1.2 (left). (C) Violin and line plots of selected genes. The paired t-test was used

8 for comparison of the expression levels of each gene between patient groups.

10 Figure 2: Impact of pazopanib treatment on gene expression. Top ten canonical pathways

11 ranking modulated by treatment identified using IPA analysis according to significance level

12 (paired t-test, p < 0.05). (A) post-treatment 3 months (Post 3) vs baseline (Pre), (B) Post

13 treatment 6 months (Post 6) vs Post 3, and (C) Post 6 vs Pre. (D) Top 3 pathways genes (as

14 listed in IPA software) in each timepoint comparison were presented by heatmap on the right

15 panels. The z-score was used to indicate the direction of the pathway activation.

17 Figure 3: Modular mapping of changes in blood transcriptome elicited by pazopanib in

18 mRCC patients. Changes in transcript abundance measured in PBMCs using whole-genome

19 arrays were mapped against a pre-constructed repertoire of co-expressed gene sets

20 (modules). The proportion of transcripts for which abundance was significantly changed in

21 comparison between samples collected at 3 months (Post 3) vs baseline (Pre), 6 months (Post

22 6) vs 3 months (Post 3) and 6 months (Post 6) vs baseline (Pre) in each module. When the

23 percentage of response exceeds 15 %, the module was considered as responsive to

24 treatment. Responsive modules are mapped on a grid, the proportion of significant transcripts

25 for each module is represented by a spot of color, with red representing increased abundance

26 and blue representing decreased abundance. The degree of intensity of the spots denotes the

27 percentage of transcripts in a given module showing significant difference in abundance in

28 comparison to the baseline. A legend is provided with functional interpretations indicated at 27

1 each position of the grid by a color code. Functional interpretations are indicated by the color-

2 coded grid at the bottom of the figure.

4 Figure 4: Mapping perturbations of the modular repertoire across individual samples.

5 Individual comparisons. Percentage of response of Individual patients at Post 3 vs Pre (blue),

6 Post 6 months vs Post 3 months (yellow), and Post 6 months vs Pre (orange). The expression

7 profile for each individual patient was calculated as a FC and difference relative to an

8 expression of individual samples in each time point. For determining post-treatment changes

9 for individual subjects, a cut-off is set against which individual genes constitutive of a module

10 are tested (|FC| > 1 and |difference| >10).

12 Figure 5: Cell-type specific analysis in pre-treatment and post-treatment samples. (A)

13 Forest plot of leukocyte enrichment score comparison between Post 3 vs Pre, Post 6 vs Post

14 3, and Post 6 vs Pre. (B) Heatmap analysis of fold-change of leukocyte enrichment score; the

15 fold change-scored values of representative fold-change between Post 3 vs Pre, Post 6 vs

16 Post 3, and Post 6 vs Pre are displayed in a heatmap. (C) Violin plots and line charts of

17 significant cell types. Red asterisks: * = p < 0.05, ** = p < 0.01, *** = p < 0.001; ns = p not

18 significant.

20 Figure 6: Flow cytometry analysis in samples with pre-treatment and post-treatment.

21 (A) Forest plot of the ratio of cell-type proportions between Post 3 vs Pre, Post 6 vs Post 3,

22 and Post 6 vs Pre analyzed by flow cytometry. (B) Heatmap analysis of fold-change of cell-

23 type proportions; the z-scored values of representative fold-change between Post 3 vs Pre,

24 Post 6 vs Post 3, and Post 6 vs Pre are displayed in a heatmap. (C) Violin plots and line chart

25 of significant cell types. Red asterisks: * = p < 0.05, ** = p < 0.01, *** = p < 0.001; ns = p not

26 significant.

1 Figure 7: Prognostic implications of MDSC gene signature in kidney renal clear cell

2 carcinoma (KIRC, n = 515). (A) Kaplan Meier curves showing overall survival (OS) of patients

3 within the highest tertile of MDSC_INT enrichment versus the two lower tertiles (LowMed).

4 Cox proportional hazards statistic are shown. (B) Boxplots of MDSC_INT enrichment scores

5 by AJCC pathologic stage (left) and histological grade (right). T-test: * = p < 0.05, ** = p <

6 0.01, *** = p < 0.001. C) Scatterplots showing the association between MDSC_INT scores

7 and the enrichment score (ES) of genes related to epithelial mesenchymal transition (upper),

8 and angiogenesis (lower). Regression line with corresponding Pearson’s correlation

9 coefficient (R) and p-value are shown. (D) Pearson correlation matrix of enrichment scores of

10 tumor-associated pathways. MDSC_INT signature is indicated with a red square. (E)

11 Univariate and multivariate overall survival Cox proportional hazards regression analysis

12 including MDSC_INT signature, pathologic stage (III&IV versus I&II), and histological grade

13 (G3&G4 versus G1&G2).

Volcano plot EnhancedVolcano

NS ● Log2 FC ● p−value ● p - value B. Post 3 vs Pre (213) Post 6 vs Post 3 (47) Post 6 vs Pre (98) 5

4 ●

P ● ● ●

● ● IFI6 RTF1● ●●● GBP1P1● ● ATP13A2

10 ●● ● ●B3GALNT2 3 ● ●●● ● ● ● g ● ● ● ●●●CLPTM1L● ● ●● ● ●● ●●●●ADK● ●● ●● ● ● ●● ● ● ●

o ACVR1 ● ● ● ● ● ● ● ● ● ● ● ● INSR● ● ● ●● ● CYTOR● ● ● ● ●● ●● ● ●● ● ● ● ● ● ●

L ● ● ● ●●● ● ● ● ● ● ●● ●●● ●● ● ●● ●● ●●● ● ● ● ●● ● ●● ● ● ● ● ●● ● ●● ● ●●● ● ● ● ●●● ● ● ●● ● ●●●● ●● ● ● ● ●● ● ●●● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●●● ● ● ●●● ● ●● ●● ● - ● ● ● ●●● ● 2 ●●● ● ●● ● ●● ● ● ● ●● ●● ●● ●● ●●● ● ●●●● ● ●●● ●●●●● ● ● ● ●● ●●●●● ● ● ● ● ● ● ● ● ●● ● ●● ●● ● ● ●● ●● ● ● ● ● ● ● ● ●● ●●● ●●●●● ● ●● ● ● ● ● ● ● ●●●●●●●●● ●●●● ●●●●●●● ● ●● ● ●●● ●●● ●●● ● ● ● ●● ●●●●● ●● ●●●●●●● ●● ● ●●● ●●●●● ●●● ●●●● ●● ● ● ●● ● ●●● ●● ● ● ● ● ●● ● ● ●● ●●●●●● ● ● ●● ●●●●●●●●● ●● ●●● ● ●●● ●● ●● ●● ●● ● ●●●●●● ● ● ● ●●●●●● ● ●● ● ● ●●●●●●●● ●● ●●●●● ●●●●●● ●●● ●●● ●● ● ● ●●●●● ●●● ●● ●● ●●●●● ● ● ● ● ●●●● ●●● ●●●● ●●●●● ● ● ● ●●●● ●●●●● ● ● ●●●●●●●●●● ● ●● ● ● ●●●●●●●●●●●●● ● ●●●●● ●●● ● ●● ● ● ● ●● ●●●●●●●● ●●●●● ●●●●●●●●●●●● ● ● ●● ● ● ● ●●●●●●● ●● ●●●●●●●●●●●● ● ● ●●●●●●●●●●●●●●● ● ●●●●●●●●● ●●●●●●●●● ● ● ● ● ●●●●●●●● ●●● ● ●● ●●●●●● ●●● ●● ● ● ●● ●●●●●●●●●●●●●● ● ●●●●●●●●●●●●●● ● ● ● ● ● ●●●●●●●●●●●● ●●●● ●●●●●●●●●●●●●●●●●●●●●● ● ●●● ●●●●●●●● ●●●●●● ● ●●●●●●●●●●● ● ● ● ●●●●●●● ●●●●●●●●●●●●●●● ●●●●●●●●●●●● ● ● ● ●●●●●●●●●●●●●●●●●● ●●●●●●●●● ●● ● ●● ●●● ●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●● ●●●●●● ● ● ● ●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●● ● ●● ●●●●●●● ●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●● ●● ● ●● ●●●●●●●●●●●●●●● ●●●●●●●●●●●● ●●● ● ●● ● ●● ●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●● ●● ● ● ●● ●●●●●●●●●●●●●●●●●●●● ● ●●●●●●●●●●●●●●●●●●● ● ● ●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●● ●● ● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●● ● ●● ● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●● ● ● ● ●● ●●●●●●●●●●●●●●●●●●●●●●● ● ●●●●●●●●●●●●●●●●● ●●●●●● ● ●● ●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●● ● ● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ● ●●● ●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●● ● ● ● ●●● ●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●● ●● ●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ●● ●●●● ●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●● ●● ● 1 ● ●● ●●● ●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●● ● ● ● ● ●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●● ●● ● ● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●● ●● ● ●● ●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ●● ●● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●● ●●● ● ● ● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ● ● ●●●●●●●● ●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ●● ● ●●● ●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●● ● ●● ●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●● ●● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●● ● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●● ●● ●● ●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●● ●●●● ●● ● ●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●● ● ●● ● ●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●● ● ● ●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ●● ●●●●●●●●●●●●●●●●●●●●●●●●● ● ● ●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●● 0 ●●●●●●●● −1.0 −0.5 0.0 0.5 1.0 Log2 fold change Total = 12913 variables C. 13 CD8A CD160 GZMB GZMH KLRD1 KLRB1 NCR3 14 ns *** * 13 ns * ns ns ns * ns 9.0 *** 11 ● ns ● 12 ● ns *** *** *** ● * *** *** 13 ● ● ● *** *** ● ● 12 ● 13 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 8.5 ● ● ● ● 12 ● ● ● ● 10 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 12 ●● ● ● ● ● ● ● ● 11 ● ● ● ● ●● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● 11 ● ● ● ● ● ● 12 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 8.0 ● ● 9 ● ● ● ● ● ● ● ● 11 ● ● ● ● ● ● ● 11 ● ● ● ● ● ● ● ● 10 ● ●● ● ● ● 10 ● ● 8 ● 11 ● ●● 7.5 10 10 ● ● ● 9 7 9 7.0 Expression value (log2) Expression value 9 9 10 10.5 12.0 ● ● ● ● ● ● ● 13 ● ● Pre ● Pre Pre Pre ● ● Pre Pre● Pre ● ● ● Post3 Post6 Post3 Post6 Post3 Post6 Post3● ●Post6 Post3 Post6 Post3 Post6 Post3 Post6 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 8.4 10.0 ● ● ● ● 11.5 ● ● 12.5 ● ● 11.5 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 12 ● ● ● ● ● ● ● ● ● ● ● ● ● ● 12 ● ● ● ● 9.5 ● ● ● ● 11.0 ● ● ● ● 11.0 ● ● ● ● ● ● ● ● ● ● ● 12.0 ● 8.0 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 9.0 ● ● ● 10.5 ● ● ● ● ● ● ● 10.5 ● ● ● 11 11 ● ● ● 8.5 ● 11.5 7.6 ● ● 10.0 ● ● ● ● ● 10.0 ● ●

● 8.0 ● ● ● ● ● ●

Expression value (log2) Expression value 9.5 11.0

Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 LAG3 IL12RB1 KIR2DL1 NCR1 NKG7 PRF1 GNLY 9.5 8.4 13 ns 11 7.4 ns 14 *** ns * ** * ns ns ns 13 ns ● ● ns ● ns ● *** *** *** ● ● *** ● * ● *** *** *** ● 9.0 ● 8.2 10 ● ● ● 12 ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● 7.2 ● ● ● ● 13 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 12 ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● 9 ● ● ● ● ● ● ● ● 8.0 ● ● ●● ● 8.5 ● ● ● ● ● ● ● ● ● ● ● ● 11 ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● 7.0 ● ● ● ● ● ● ● ● 12 8 ● 11 ●● ● ● 7.8 ● 8.0 ● ● ● ● ● ● ● ● 10 ● ● ● ● ● ● ● ● 7 ●● ● 6.8 ● 7.5 ● 7.6 11 ● 10 6 9 Expression value (log2) Expression value 6.6 9.2 13.5 ● ● ● ● ● ● ● ● 8.2 ● ● ● ● ● 12.5 Pre Pre Pre Pre Pre ● ● Pre Pre ● ● ● Post3● Post6 Post3 Post6 Post3 Post67.2 Post3 Post6 Post3 Post6 Post3 Post6● Post3 Post6● ● ● ● ● 12.0 ● ● ● ● ● ● ● ● ● ● ● ● ● 13.0 ● ● ● ● ● 8.8 8.1 ● ● ● ● ● ● ● ● ● ● ● ● ● ● 12.0 ● ● 9 ● ● ● ● ● ● 11.5 ● ● ● ● ● ● ● ● 7.1 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 8.0 ● ● 12.5 ● ● ● ● ● ● ● ● ● ● 8.4 ● ● ● ● ● ● ● 11.0 11.5 ● ● ● ● ● ● ● ● ● ● ● 7.9 7.0 ● ● ● ● ● 12.0 ● ● 8 ● ● ● ● 10.5 ● 8.0 ● ● 11.0 7.8 ● ● ● 6.9 ● 11.5 ● ● ● 10.0 ● ● ● ● 7.6 ● 7.7 ● ● ● ● ● ● 10.5 ● Expression value (log2) Expression value

Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Figure 1 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.01.071613; this version posted May 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A. Post 3 vs Pre D. Pre Post3 Post6 10.0 100

- TLR7.1

7.5 log(p TNFSF10 75 HLA.DRB3 HLA.DRB1.1

- IFNG 5.0 value) IL15 50 NCR3.2 KIR3DL1.2 KIR3DL2 2.5 KIR2DL5A

Percentage 25 Antigen presentation HLA.DRB4 KIR2DL5B.2 0 HLA.DQB1 0 HLA.DOA.2 HLA.DPB1 10.0 KLRD1.1 PSMB9 0.5 PSMB8 7.5 APH1B

0.4 Ratio HLA.DRB3.2 HLA.DRB1.2 MICB

value) 0.3 ACTA2 - 5.0 IL2RG 0.2 CD8A STAT4 log(p - 2.5 IL12A 0.1 CD40 TNFRSF4 TNFSF4 0 0.0 TLR7 2 A IFNGR2 - ACVR1.1 JAG2 HLA.DPA1 HLA.DMB

activation Crosstalk btw DC & NK JAK2

8 Signaling 8 IL10 -

Granzyme HLA.DOA.3 IL TCR TCR signaling HLA.DPB1.2 RxRa

PPAR Signaling PPAR IL24.1 / NK cell signaling NCR3.1 Regulation of IL

Th1/Th2 activation PRF1 CD247.2 Telomerase signaling IL2RB Crosstalk btw DC & NK

PPARa KLRD1 KLRD1.3 KIR2DL5A.1 KIR2DL4.1 KIR3DL1.1 KIR3DL3.1 KIR3DL2.1 B. Post 6 vs Post 3 HLA.DRB4.2 Time Pathway KIR2DL5B RAF1 Pre Crosstalk btw DC & NK PLCG2.1 Post3 Th2 signaling 4.0 VAV2 KLRC1 100 Post6 Th1|Th2 Activation 3.5 KLRB1 NK cell signaling KIR2DL5A.2

PatientID - Regulation of IL−2 3.0 log(p KIR2DL4 P1 KIR3DL1 75 Antigen presentation P2 2.5 KIR3DL3

- KIR2DL5B.1 P3 z scores 2.0 value) 4 FYN 50 P4 PRKCH P5 2 1.5 NCR3 KLRC3 Percentage 25 P6 0 NK cell signaling1.0 HCST P8 −2 RAC2 0.5 CD247 −4 KLRD1.2 0 0.0 RRAS2 SH2D1B 4.0 1.00 KLRC2 NCR1 3.5 KIR2DS3 KIR3DL2.2 3.0 0.75 CD300A Ratio MAPK1 2.5 KRAS value) - 2.0 PIK3R2 0.50 MAPK8 1.5 FYN.1

log(p NFATC3 - 1.0 0.25 CD247.1 RRAS2.1 0.5 NFKBIB MAPK1.1 0.0 0.00 KRAS.1 Regulation of IL−2 RAF1.1 MAP3K1 Th2 Th1

APC VAV2.1 TGFBR2 IKBKB

mannose PLCG2 Cell cycle - NFATC1 NOTCH1.1 EIF2 signaling EIF2 CML signaling

GDP IL18R1 RUNX3 Glioma Signaling Glioma IFNG.1 Th1/Th2 activation Role of PKR in IFN IL12RB1.1 TAP2 ACVR2A.1 C. Post 6 vs Pre IL24.2 HLA.DRB4.1 HLA.DQB1.1 Th1|Th2 Activation HLA.DOA 4.0 HLA.DOA.1 HLA.DPB1.3 100 3.5 HLA.DPB1.4

- HLA.DPA1.1 3.0 log(p HLA.DMB.1 75 HLA.DRB3.1 2.5 HLA.DRB1 - ACVR1.2 value) 2.0 PIK3R6 50 NOTCH1 1.5 ACVR1 Percentage PIK3R6.1 25 1.0 ACVR2A HLA.DOA.4 0.5 HLA.DPB1.1 Th2 0.0signaling IL24 0 HLA.DQB1.2 RUNX3.1 4.0 1.00 IL12RB1 3.5 IFNG.2 3.0 0.75 Ratio 2.5 P1.0 P2.0 P3.0 P4.0 P5.0 P6.0 P8.0 P1.3 P2.3 P3.3 P4.3 P5.3 P6.3 P8.3 P1.6 P2.6 P3.6 P4.6 P5.6 P6.6 P8.6 value)

- 2.0 0.50 Downregulated 1.5

log(p No change - 1.0 0.25 Upregulated 0.5 0.0 0.00 No overlap with dataset -log(p-value) Th2 Th1 APC

Positive z-score Role of CHK NK signaling Z-score = 0 OX40Signaling

Th1/Th2 activation Negative z-score Crosstalk DC & NK No activity pattern available Assembly RNA Poly III Poly RNA Assembly Ratio Oxidative Phosphorylation Oxidative

Figure 2 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.01.071613; this version posted May 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

M1 ●●●●●●●●●●●●●●●●●●●●●●●●Post●● 3 ●vs● Pre● %Response M2 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● 100 M3 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● M4 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● 50 M5 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● M6 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● 0 M7 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● −50 M8 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● M9 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● −100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 24 34 42 46 61 83 89

M1 ●●●●●●●●●●●●●●●●●●●●●●●Post● ●6 vs● Post●● 3● %Response M2 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● 100 M3 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● M4 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● 50 M5 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● M6 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● 0 M7 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● −50 M8 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● M9 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● −100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 24 34 42 46 61 83 89

M1 ●●●●●●●●●●●●●●●●●●●●●●●●Post●● 6 ●vs● Pre● %Response M2 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● 100 M3 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● M4 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● 50 M5 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● M6 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● 0 M7 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● −50 M8 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● M9 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●● −100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 24 34 42 46 61 83 89

% gene set p < 0.05

No module

Figure 3 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.01.071613; this version posted May 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Post3vsPre Post6vsPost3 Post6vsPre Time PatientID ●●●●●●●● ●●●●●●● ●●●●●●● M3.6.Cytotoxic/NK Cell ●●●●●●●● ●●●●●●● ●●●●●●● M8.46.Cytotoxic/NK Cell ●●●●●●●● ●●●●●●● ●●●●●●● M4.11.Plasma Cells ●●●●●●●● ●●●●●●● ●●●●●●● M4.15.T cells ●●●●●●●● ●●●●●●● ●●●●●●● M4.1.T cells ●●●●●●●● ●●●●●●● ●●●●●●● M4.10.B cells ●●●●●●●● ●●●●●●● ●●●●●●● M3.4.Interferon ●●●●●●●● ●●●●●●● ●●●●●●● M1.2.Interferon ●●●●●●●● ●●●●●●● ●●●●●●● M9.42.Cell Cycle ●●●●●●●● ●●●●●●● ●●●●●●● M5.9.Protein Synthesis ●●●●●●●● ●●●●●●● ●●●●●●● M4.3.Protein Synthesis ●●●●●●●● ●●●●●●● ●●●●●●● M5.12.Interferon ●●●●●●●● ●●●●●●● ●●●●●●● M8.89.Immune Responses ●●●●●●●● ●●●●●●● ●●●●●●● M2.3.Erythrocytes ●●●●●●●● ●●●●●●● ●●●●●●● M8.61.Immune Responses ●●●●●●●● ●●●●●●● ●●●●●●● M3.3.Cell Cycle ●●●●●●●● ●●●●●●● ●●●●●●● M5.1.Inflammation ●●●●●●●● ●●●●●●● ●●●●●●● M7.1.Inflammation ●●●●●●●● ●●●●●●● ●●●●●●● M4.5.Protein Synthesis ●●●●●●●● ●●●●●●● ●●●●●●● M5.7.Inflammation ●●●●●●●● ●●●●●●● ●●●●●●● M9.34.Immunosuppression ●●●●●●●● ●●●●●●● ●●●●●●● M6.18.Erythrocytes ●●●●●●●● ●●●●●●● ●●●●●●● M2.1.Cell Cycle ●●●●●●●● ●●●●●●● ●●●●●●● M4.2.Inflammation ●●●●●●●● ●●●●●●● ●●●●●●● M4.13.Inflammation ●●●●●●●● ●●●●●●● ●●●●●●● M4.6.Inflammation ●●●●●●●● ●●●●●●● ●●●●●●● M6.6.Apoptosis/Survival ●●●●●●●● ●●●●●●● ●●●●●●● M6.13.Cell Death ●●●●●●●● ●●●●●●● ●●●●●●● M7.16.DC/Apoptosis ●●●●●●●● ●●●●●●● ●●●●●●● M5.15.Neutrophils ●●●●●●●● ●●●●●●● ●●●●●●● M3.2.Inflammation ●●●●●●●● ●●●●●●● ●●●●●●● M4.14.Monocytes ●●●●●●●● ●●●●●●● ●●●●●●● M7.22.Inflammation ●●●●●●●● ●●●●●●● ●●●●●●● M8.83.Immune Responses ●●●●●●●● ●●●●●●● ●●●●●●● M1.1.Platelets P1 P2 P3 P4 P5 P6 P8 P9 P1 P2 P3 P4 P5 P6 P8 P1 P2 P3 P4 P5 P6 P8

Figure 4 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.01.071613; this version posted May 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A. Cell typePost 3 vs Pre p Cell typePost 6 vs Post 3 p Cell typePost 6 vs Pre p ** NKCD56dim 1 Th17 cells 1 * NKCD56dim 1 ** NKCD56bright 2 Tem 2 * NKCD56bright 2 *** Tgd 3 MDSC_RivINT 3 * NKT 3 ** NKT 4 Treg 4 * Tgd 4 ** Cytotoxic cells 5 Tcm 5 * Cytotoxic cells 5 * CD8 T cells 6 MDSC_Angel 6 Tem 6 T cells 7 mDC 7 CD8 T cells 7 Th1 cells 8 iDC 8 T cells 8 Tcm 9 TFH 9 Tcm 9 Tem 10 Monocyte 10 T helper cells 10 T helper cells 11 pDC 11 pDC 11 NK cells 12 T helper cells 12 mDC 12 pDC 13 T cells 13 NK cells 13 mDC 14 CD8 T cells 14 Th17 cells 14 TFH 15 B cells 15 TFH 15 Monocyte 16 NK cells 16 MDSC_Angel 16 MDSC_Angel 17 DC 17 Monocyte 17 Th2 cells 18 Cytotoxic cells 18 Th1 cells 18 DC 19 Th2 cells 19 MDSC_RivINT 19 MDSC_RivINT 20 G−MDSC 20 iDC 20 G−MDSC 21 NKT 21 Th2 cells 21 22 Th17 cells Th1 cells 22 DC 22 23 iDC NKCD56dim 23 Treg 23 B cells 24 * Tgd 24 G−MDSC 24 Post3vspre Post6vspost3Treg 25* post6vspre NKCD56bright 25 B cells 25 Time 0.925 1 1.05 1.1 1.15 0.925 0.97511.025 1.075 0.925PatientID1 1.05 1.1 Mean ratio Mean ratio Mean ratio Post3 / Pre CD8 T cells Post6 / Post3 Post6 / Pre B. Post 3 vs Pre Post 6 vs Post 3 Post 6 vs Pre NKCD56bright z score Time NKCD56dim 2 Post3vspre 1 Post6vspost3 NKT 0 post6vspre −1 PatientID Tgd −2 P1 P2 Cytotoxic cells P3 P4 Mast cells P5 P6 Treg INT P8 G−MDSC MDSC_Riv C. CD8 T cells CytotoxicCD8 T cells CytotoxicG−MDSC cells MastG−MDSC cells MDSC_AngelMast cells MDSC_RivMDSC_AngelINT MDSC_AngelMDSC_Riv ns 0.4 0.4 0.35 0.450.35 0.45 ns ns ns P1 P2 P3 P4 P5 P6 P8 P1 P2 P3 P4 P5 P6 P8 P1 P2 P3 P4 P5 P6 P8 ● *● ● ● ● ● ● ns● ● ● −0.25 ● ● −0.25 ● ● ● ● ● ns ns ns ● ns ● ns ns ● ns ● ● 0.33 ● ● ● ● 0.33 ● ● ● 0.2 0.2 ● ● * ● ● ● ● ● ● ● ● ● ● ● ● ● **● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.40 ● ● ● ● 0.2 0.40 ● ● ● 0.2 ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● −0.30 ● ● ● ● ● −0.30 ● ● ● ● 0.30 ● ● ● ● ● ● 0.30 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ●● ● ● ● ● 0.30 ● ● ● 0.30 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.1 ● 0.1 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● −0.35 ● ● ● ● ● ● ● ● 0.35 −0.35 ● ● ● ● 0.35 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.0 ● ● ● ● ● ● 0.0 ● ● ● ● ● ● ● ● ● ● ● ● ● 0.25 ● ● ● ● ● ● 0.25 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● 0.27 0.27 −0.40 −0.40 0.30 0.30 −0.2 0.0 0.0 0.20 0.20 −0.2 CD8 T cells CytotoxicCD8 T cellscells CytotoxicG−MDSC cells MastG−MDSC cells 0.24 MDSC_AngelMast cells 0.24 MDSC_RivMDSC_Angel MDSC_Riv ● ● ● ● ● ● ● ● ● ● −0.27 ● ● ● CD8 T ●cells Cytotoxic● cells● −0.27 G−MDSC ●Mast cells ● MDSC_Angel MDSC_Riv ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.32 ● ● ● ● ● 0.32 ● ● ● ● ● ● ● Pre PrePre PrePre ● ● ●PrePre PrePre● Pre Pre Pre 0.32 ● ● 0.4 ● 0.32 Post3 Post6 Post3Post3Post6● Post60.45 ● Post3Post3Post6Post6● 0.2 Post3Post3 Post6Post6 Post3Post3 Post6Post6 Post3Post3Post6Post6 0.16 Post3 Post6 ● ● 0.2 ● 0.16 0.35 ● ● ● −0.30 ● ● ● 0.31 ● ● ● 0.40 ● ● −0.30 −0.25 ● ● 0.31 ● ● ● ● ● 0.40 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.33 ● ● ● ● ● ● ● ● ● ●● ● ● 0.2 NKCD56bright NKCD56dim NKT Tgd Treg ● NKCD56bright● ● 0.30 NKCD56dim● NKT● ● Tgd Treg ● ● ● ● 0.30 ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.4 ● ● 0.4 ● ● ● ● ● ● ● ● 0.12 ● ● 0.30 ● ● 0.40 ● 0.2 ● ● ● ● 0.3 ● ● 0.3 ● 0.30 0.12 ● ● ● ● 0.1 ● ● ● ● ● 0.06 ● 0.1 ● −0.060.33 ● −0.30 ● ● ● ● ● ● 0.30 ● ● ● ● ● ● ● −0.33 ● ● ● ● ● ● ● ● 0.28 ● ● ● ●●● ● ● ● ● ● ● ● ● 0.28 ● ● ● ● ● ● ● ● Enrichment scores ● ● ● ● ● ●

Enrichment scores ● ● ● ● ● 0.2 0.2 ●● ● ●● ● ● ● 0.30 0.36 ● ● ● ● ● ● ● ● ● ● 0.4 ● ● 0.4 ● ● ● ● ● 0.36 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.29 ● ● 0.2 ● ● ● 0.2 ● ● ● ● ●● ● ● ● 0.29 ● ● ● ● ● ● ● ● ● ● ● ● 0.1 ● ● ● 0.04 ● ● ● ● ● 0.04 ● ● ● ● ● ● ● ● ● 0.2 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● −0.35 ● 0.2 ● ● 0.08 0.35 ● ● 0.26 ● ● ● ● ● ● ● ● ● −0.36 ● 0.08 ● ● ● ● ● ● ● 0.26 ● ● ● ● ● ● ● ● ● ● ● ● 0.0 ● ● −0.36 ● ● ● 0.0 ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.0 ● ● ● ● ● ● ● ● ● ● ● ● ● 0.25 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● 0.1 ● ● ● ● ● ● 0.1 ● ● ● ● ● ● 0.28 ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.28 ● ● ● ● ● ● ● ● ● ●● ● 0.2 ● ● ● ● ● 0.27 ● ●● 0.2 ● ● ● ● ● ● ● ● ● ● ● ● 0.02 ● ● ● ● ● ● 0.02 ● ● 0.0 ● 0.0 ● ● ● ● ● ● ● ● ● ● ● ● 0.24 ● ● ● ● ● ● ● ● ● ● ● ● 0.32 ● −0.40 0.24 ● ● ● ● ● ● ● ● ● ● ● ● 0.32 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.0 ● ● ● −0.39 ● ● 0.0 ● 0.30 ● 0.04 ● ● ● −0.39 ●● ●● ● ● 0.04 ● ● 0.0 ● 0.0 ● ● ● ● 0.0 ● ● ● ● ● ● ● ● ● ● ● ● −0.2 ● ● ● ● ● ● ● ● ● ● ● 0.20 0.0 0.000.0 ● 0.00 ● CD8 T cells Cytotoxic cells G−MDSC Mast cells MDSC_Angel● MDSC_Riv ● ● ● 0.24 ● ● ● ● ● ● −0.1 −0.1 ●● ●● −0.2 −0.2 ● ● ● ● ● ● Pre PrePre PrePre ● PrePre −0.27 PrePre Pre●Pre Pre −0.2 ●Post3 Post6 −0.2 Post3● Post6● Post3 Post6 Post3 Post6● Post3 Post6 Post3 Post6 Post3 Post6 Post3 Post6 Post3● Post6● Post3● Post6 Post3 Post6 −0.02 Post3 Post6 ● ● −0.02 0.32 ● ● ● ● Pre Pre Pre ● Pre● Pre Pre −0.2 ● −−0.20.2 −0.2 ● 0.32 Post3 Post6 Post3 Post6● Post3 Post6 Post3 Post6 Post3 Post6 Post3 Post6 ● ● 0.2 0.16 NKCD56bright NKCD56dimNKCD56bright NKCD56dimNKT● TgdNKT TregTgd Treg ● ● −0.30 0.31 ● ● 0.40 ● ● 0.2 ● ● ● ● ● ● ● 0.2 ● ● ● ● ● ● ● ● ● ● NKCD56bright● ● NKCD56dim● ● ● NKT ●● Tgd● ● Treg● 0.30 ● ● ● ● ● ● ● ● ● ● ● ● 0.04 ● ● ● 0.2 ● ● ● ● 0.04 ● ● 0.2 ● ● ● ● 0.1 0.4 ● ● ● 0.1 Pre Enrichment scores PrePre ● Pre● Pre 0.3 PrePre ● ● PrePre● ● ● Pre ● 0.12 Enrichment scores ● ● ● ● ● ● ● 0.060.30 Post3● Post6 Post3●Post3Post6Post6 Post3Post3● Post6Post6 0.1 ●● Post3●Post3 Post6●Post6 0.3 Post3Post3● Post6Post6 Post3 Post6 ● ● 0.3 ● ● ● −0.33 ● ● ● ● ● ● ● * ● ● ● ● *● ● 0.28 ● ● * *● ● ● Enrichment scores ● * ● ● ● ● 0.03 ● ● 0.2 ns ● 0.03 ● ● ● ns● ●● ● ● ns ● ns ● 0.4 ● ns 0.36 0.1 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.10.1 ● 0.2 ● ● ● ● ●● ● ● ● ● ● ●● 0.1 ● ● ● ● ● ● ** ● ● 0.29 ● ● ● ** ● ● ● ** ● ● ● ● ● 0.04 ● ***● ● ● ● ● ● ● ● 0.2 ● * ● ● ● ● ● ● 0.2 ● ● ● ● ● 0.0 ● 0.2 ● ● ● ● ● ● ● ● ● ● 0.08 0.0 0.26 ● ● ● ● ● ● ● ● ● ● ● −0.36 ● ● ● 0.02 ● ● ● ●● ● ● ● ● ● ● ● ● 0.02 ● ● ● ● ● ● 0.0 ● Post3 Post6● Pre ● ● ● ● Post3 Post6 Pre● time ● ● ● ● ● time ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● 0.1 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.28 ● ● ● ● ● ● ●● 0.2 ● ● ● ● ● ● ● ● ● 0.0 ● ● ● 0.02 ● ● ● 0.0 ● ● ● ● ● ● ● ● 0.0 ● ● ● ● ● ● ● ● ● ● ● ● 0.1 ● ● ● ● ● 0.1 ● ● ● 0.24 ● ● ● ● ● ● ● ● ● 0.0 ● ● ● ● 0.01 ● ● ● 0.32 ● ● ● ● 0.0 ● ● ● 0.01 ● ● ● ● ● 0.0 ● ● ● ● ● ● ● ●● ● ● ● −0.39 ● 0.04 −0.1 ●● 0.0 ● ● ● ● ● ● ● −0.1 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.00 ● 0.00.0 ● ● ● ● −0.1 0.0 0.00 ● −0.1 ● ● 0.00 ● ● ● ●● ● ● −0.1 ● −0.2 Pre● −0.2 Pre● Pre● Pre● Pre ● Pre ● ● ● −0.1 ● ● −0.2 −0.2 Post3 Post6 −0.1 Post3 Post6 Post3 Post6 Post3 Post6 −−0.020.01 Post3 Post6 Post3 Post6 −−0.20.1 −−0.01−0.20.1 NKCD56bright NKCD56dim NKT Tgd Treg Pre ● Pre ● 0.2 Pre ● Pre ● ●Pre Pre Pre ●Post3 Post6 Pre Post3● Post6 Pre Post3 Post6 Pre Post3● Post6 ● Post3 Post6 Post3 Post6 Post3 Post6 Post3 Post6 Post3● Post6 Post3 Post6● 0.2 ● 0.04 ● 0.1 Pre Pre Pre ● Pre Pre

Enrichment scores ● ● Post3 Post6 Post3 Post6 Post3 Post6 ● Post3 Post6● Post3 Post6 ● ● 0.3 ● ● ● ● ● ● ● Enrichment scores Enrichment ● ● ● ● ● 0.03 ● ● ● ● ● 0.1 ● ● ● ● ● ● ● ● 0.1 ● ● ● ● ● ● P1 P3 P5 P8 ● ● ● ● P1 P3 P5 P8 ● ● ● ● Patient 0.2 ● ● ● ● 0.0 ● ● Patient● ● ● ● ● ● ● ● ● ● ● P2 P4 ● P6 P9 0.02 ● ● ● P2 P4 P6 P9time Post3 Post6 Pre ● ● ● ● ● ● ● ● ● ● ● ● 0.0 ● ● ● ● ● ● 0.1 ● ● 0.0 ● 0.01 ● ● ● ● ● ● −0.1 ● ● ● ● ● ● ● ● ● −0.1 0.0 0.00 ● ● ● ● ● −0.1 ● ● ● −0.2 −0.1 −0.01

Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Time ● P1 ● P3 ● P5 ● P8 Patient ● P2 ● P4 ● P6 ● P9

Figure 5 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.01.071613; this version posted May 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A. Cell typePost 3 vs Pre p Cell typePost 6 vs Post 3 p Cell typePost 6 vs Pre p ActivatedActivated NK NK cell cell 1 ActivatedActivated NK NK cell cell 1 ActivatedActivated NK NK cell cell 2 * CytotoxicCytotoxic NK NK cell cell 1 InflammatoryCD14+PDL monocyte−1+ 1* CytotoxicCytotoxic NK NK cell cell 1 ActivatedActivated T Tcell cell 1 MONOCD14+HLA_DR-MDSC − 1 ActivatedActivated T cell T cell 1 ** PMNCD15-MDSC MDSC 0.8 CD14+Total monocyte 1 InflammatoryCD14+PDL monocyte−1+ 1 InflammatoryCD14+PDL −MDSC1+ 0.8 ActivatedActivated T cellT cell 1* CD14+Total monocytemonocyte 0.8 CD14+Total monocytemonocyte 0.7 ** CytotoxicCytotoxic NK NK cell cell 1 MONOCD14+HLA_DR-MDSC − 0.8 RegulatoryRegulatory T Tcell cell 0.6 RegulatoryRegulatory T cell T cell 0.9 RegulatoryRegulatory T cell T cell 0.6 MONOCD14+HLA_DR-MDSC − 0.6 ** PMNCD15-MDSC MDSC 0.7 PMNCD15-MDSC MDSC 0.5 Post3vspre Post6vspost3 post6vspre 0.5 1 1.5 Time Mean ratio 0.5 1 1.5 2 PatientID 0.5 1 1.5 2 Mean ratio Post3 / Pre Mean ratio Total monocyte Post6 / Pre B. Post3vspre Post6vspost3 Post6 / Post3post6vspre Time PatientID z scores Time CD14+HLA−DRneg 2 TotalCD14+ monocyte monocyte Post3vspre Activated NK cell Activated T cell CD14+HLA_DR− CD14+PDL1 −Post6vspost31+ RegulatoryCD14+HLA_DRMONO- MDSCT cell− 0 post6vspre −1 PatientID Regulatory50 T Tcell cell 100 20 ● 50 ● −2 ● P1● ●

● ●● CD15PMN -MDSC● ● CD15 MDSC P2● ● ● ● ● ● ● ● ● ● ● ● ● ● 50 ● ● ● ● ● Activated T cell P3 ● 25 ● ● ● Activated T cell● ● ● 25 ● ● ●● ● 10 ● ● ● ● ●● ●● ●● ● ● ● ● P4 ● ● ● ● ● ● ● ● ● ● ● ● ●● ● CD14+PDL−1+ ● ● ● Inflammatory● MDSC ● ● ● ● ● Activated● T cell ● P5 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0 ● ● ● ● 0 ● ● ● ●● Activated NK cell P6 0 ● ● 0 Activated NK cell CD14+PDLCytotoxic −NK NK1+ cell cell P7 −25 −50 P8 P1 P2 P3 P4 P5 P6 P7 P8 P1 P2 P3 P4 P5 P6 P7 P8 P1 P2 P3 P4 P5 P6 P7 P8

20 Pre ActivatedPre NKPre cell Pre Post3 Post6● Post3● Post6 ● Post3 Post6 80 Post3 Post6● Inflammatory 40 ● C. Activated NK cell Activated T cell CD14+HLA_DRMONO-MDSC− CD14+PDLmonocyte−1+ CD15PMN- MDSC 40CytotoxicCytotoxic NK NK cell cell Regulatory● T cell CD14+Total monocyte ● ● 6 Cytotoxic NK cell ● ns ns 15 ● 60 ns ns ns ● ● ns % Positive cells % Positive ● ● ● ● 30 ●● ns ns ns * 3 ● ● 60 * ns 100 ● ns ns ns ns ns 30 ● ns ● ns ● ns 50 ● ●●● 20 ● 50 ● ** 75 ● ● ●● ** * ● ● ● ● ** ● ● ● ● ● ● ● 4 ● ● ● ● ● 40 ● ● 10 ● ● ● ●● ● P1 P2 P3 P4 P5 P6 P7 P8 P1 P2 P3 P4 P5 P6 P7 P8 P1 P2 P3 P4 P5 P6 P7 P8 ● 2 ● 20 ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● 40 ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● 50 ● ● ● 20 ● ● ● ● ● ● ● ● ● ● ● 50 ● 25 ● ● ● ● ● ● ● ● ● ● ● ● ● 25 ● ● ● 10 ● ● ● ● ● ● ● ● ● ● ● ● ● ● 1 ● ● ● ●● ● ● ● ● ● 20 ● ● ● ● 52 ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● 10 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● 20 ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● 0 ● ● ● ● ● ● ● ● ● ● 2510 ● ● ●● ● ● 0 ● ● ● ●● 0 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● 0 0 0 ● 0 ● ●● −1 −50 0 0 −25 Pre Pre Pre Pre 20 Post3 Post6 Post3 Post6 Post3 Post6 Post3 Post6 ● ● ● 80 ● 40 ● ● ● 3 ● ● Pre Pre Pre Pre● Pre Pre Pre Pre Post3 Post6 40 Post3 Post6 ● Post3 Post6 Post3 Post6 4 ● Post3 Post6 Post3 Post6 Post3 Post6 Post3 Post6 % Positive cells % Positive 50 ● ● ● ● ● ● 15 CD15 MDSC Cytotoxic ● NK cell Regulatory T cell 60 Total monocyte 60 ● ● ● ● ● ● ● ● 30 ● ● ● 6 ● ● ● 40 ● 3 ● ● ● ●

% Positive cells % Positive 30 ● ● 2 ● ● ● ● ● ● ● ● ● 3 ● ● ● 60 ● ● ● ● ● ● ● ● ● ● 40 ●● ● ● ● ● 10 75 ● ● ● 30 ● ● 20 40 ● ● ● ● ● ● ● ● ● ● ● 4 ● ● ● 2 ● ● ● ● ● ● ● ● ● 2 20 ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● 40 1 ● ● ● ● ● % Positive cells Positive % ● ● ● ● ●● ● ● ● ● ● 20 ●● ● 20 ● ● ● ● ● ● ● 5 ● ● 50 ● ● ●● ● ● ● ● ● ● ● ● ● ● 10 ● ● ● ● ● ● ● ● ● ● ● ● ● 1 ● ● ● ● ● ● 1 ● ● ● 2 ● ● ● 20 ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 10 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● 10 ● ● ● ● ●● ● ● ● ● ● ● 20 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● 0 ● ● ● ● ● ● ● ●● 0 ● ● 25 ● 0 ●● ● 0 ● ● ● ● ● ● ● ● ● ● 0 ● ●● Pre Pre −1 Pre 0 Pre Post3 Post6 0 Post3 Post6 Post3 Post6 Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 ● ● 3 ● ● 4 ●

% Positive cells % Positive 50 ● Pre Post3 Post6 TimePre Post3 Post6● Pre Post3 Post6 Pre Post3 Post6 60 ● ● ● ● ● ● ● ● 40 3 ● ● ● ● 2 ● ● ● ● ● ● ● ● ● ● 30 ● 40 ● 2 ● ● ● ● ● ● ● ● ● ● 1 ● ● ● ● ● ● ● 20 ● ● ● ● ● ● ● ● ● ● ● 1 ● ● ● 20 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 10 ● ● ● ● ● ● ● 0 ● ● ●

Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Pre Post3 Post6 Figure 6 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.01.071613; this version posted May 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A. B. *** *** HR = 2.057, p = 3.2e-06 *** * *** 95% CI = 1.52 - 2.79 *** *** *

MDSC_INT MDSC_INT MDSC_INT enrichment LowMed MDSC_INT enrichment High

Low Medium High

C. D. R = 0.75 p < 2.2e-16 MDSC_INT MDSC_INT enrichment

Epithelial mesenchymal transition

R = 0.59 p < 2.2e-16 Pearson’scorrelation coefficient MDSC_INT MDSC_INT enrichment

Angiogenesis E. Univariate Multivariate Variables HR (95% CI) p HR (95% CI) p

~ MDSC signature 2.057 1.478 3.2e-06 *** (High versus LowMed) (1.519-2.787) (1.081- 2.019) 0.01425 * ~ Pathologic Stage 4.054 3.125 <2e-16 *** (III&IV versus I&II) (2.922-5.625) (2.215- 4.410) 8.93e-11 *** ~ Histological grade 2.778 1.761 1.34e-08 *** (G3&G4 versus G1&G2) (1.953- 3.951) (1.213- 2.554) 0.00289 **

Figure 7 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.01.071613; this version posted May 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A. NK cell signaling pathways (3 vs 0)

B. Antigen presentation pathway (6 vs 3)

C. Cross talk between dendritic cells & NK cells(6 vs 0)

Supplementary Figure 1 : Top pathway in each time point comparison. bioRxiv preprint doi: https://doi.org/10.1101/2020.05.01.071613; this version posted May 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A. B. C.

High versus LowMed: HR = 2.057, p = 3.2e-06 *** High versus LowMed: HR = 1.1, p = 0.554 High versus LowMed: HR = 1.35, p = 0.0583 High versus Low: HR = 1.95, p = 2.66e-04 *** High versus Low: HR = 1.08, p = 0.677 High versus Low: HR = 1.19, p = 0.337

MDSC_INT G.MDSC MDSC Angelova

Low Low Low Medium Medium Medium High High High

MDSC_INT High versus LowMed: HR = 2.057, p = 3.2e-06 ***, 95% CI = 1.52 - 2.79 High versus Low: HR = 1.95, p = 2.66e-04 ***, 95% CI = 1.36-2.8

G.MDSC High versus LowMed: HR = 1.1, p = 0.554, 95% CI = 0.8-1.51 High versus Low: HR = 1.08, p = 0.677, 95% CI = 0.75-1.56

MDSC_Angelova High versus LowMed: HR = 1.35, p = 0.0583, 95% CI = 0.99-1.83 High versus Low: HR = 1.19, p = 0.337, 95% CI = 0.84-1.69

Supplementary Figure 2 : Prognostic value of MDSC signatures. Kaplan Meier curves showing overall survival (OS) of patients within tertiles of MDSC_INT (A), granulocytic MDSC (B), and MDSC Angelova (C) enrichment scores. Cox proportional hazards statistics of the high versus LowMed (combination of intermediate and low tertiles) and High versus Low tertiles are shown.

A B

MDSC_INT (25)

S100A8 SDC4 COL3A1, COL5A2, VCAN, LUM, ITGAV, FSTL1, POSTN, SPP1, MDSC_INT (25) TIMP1, VEGFA, MSX1, CXCL6

Supplementary Figure 3: Venn diagrams of intersection between MDSC signatures; MDSC_INT, MDSC_Angelova and G-MDSC (A) , and MDSC_INT and Cancer hallmark pathways (Angiogenesis and Epithelial mesenchymal transition) (B) . The total number of genes in each area is reported between the parentheses. bioRxiv preprint doi: https://doi.org/10.1101/2020.05.01.071613; this version posted May 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Pairwise person correlation between MDSC and flowdata

MDSC_LR

PairwisePairwise person peason correlation correlation between genes Bindea.MDSCup betweenPairwisePairwise MDSCMDSC_Angelova person person and correlation correlationflowdata between between MDSC MDSC and and flowdata flowdata A. B. Celltype PDLIM7 CD15 MDSC STK32B Total monocyte DSE MDSC_AngelMDSC_AngelTotal monocyte CKAP4 MDSC_LR CD14+HLA−DRneg GAS6 MDSC_RivMDSC_RivMDSC_INTCD14+PDL−1+ SH3PXD2B CD15 MDSC Activated T cell DataData LIMK2 MDSC_Angelova Activated NK cell Flow_dataFlow_data MT1F CD15CD15PMN MDSC- MDSC MT1H MDSC_Trajo Cytotoxic NK cell ESES PNP Total monocyte Regulatory T cellCorrelationCorrelation C1RL CD14+HLACD14+HLAMONO-MDSC−DR−DR− − 1 1 MMP9 Correlation MDSC_LR 1 0.50.5 SDC4 G0.5−MDSC G−MDSC PRELID1 0 CD15 MDSC 0 0 −0.5 G_MDSCG_MDSCG-MDSCMDSC_Angelova MXD1 −1 MDSC_Trajo −0.5−0.5 − S100A8 Activated T cell − TPST1 MDSC_Trojanoski Correlation −1−1 DR MCL1 DR 1 − ADPRH − MDSC MDSC MDSC Data

- - - 0.5 PCSK5 CD14+HLAG_MDSC−DRneg G_MDSC G_MDSC SEMA4B G 0 MDSC_Riv MDSC_Riv Flow_data SLC11A1 MDSC_INT CD15 MDSC CD15 MDSC PMN MDSC_Angel SH2B2 MDSC_Angel −0.5 ES

ActivatedMDSC_Angel T cell SLCO2B1 Activated NK cell MONO −1 CD14+HLA TMEM88 CD14+HLA Correlation CD14+HLA−DRneg 1 DSE PNP C1RL MT1F SDC4 MCL1 GAS6 MT1H MXD1 MMP9 LIMK2 CD14+PDL−1+ TPST1 SH2B2 CKAP4 PCSK5 ADPRH S100A8 PDLIM7 STK32B TMEM88 SEMA4B PRELID1 SLC11A1 0.5 SLCO2B1 SH3PXD2B RegulatoryActivated T cell NK cell 0 CD14+PDL−1+ −0.5 Cytotoxic NK cell −1 Supplementary Figure 4 : Pairwise Pearson correlation matrix betweenRegulatory(A) 25 genes T cellof the top r2 (kmean1) of top 100 up-regulated genes from the data set obtained from human MDSC generated in vitro according to a model developed in our laboratory [Huber et al. JCI 2018]. (B) enrichment scoreCytotoxicof MDSC NK cellsignature and percentage expression cell type by flow cytometry . 1+ − DRneg − G_MDSC MDSC_LR CD15 MDSC Total monocyte Total Activated T cell Activated CD14+PDL Cytotoxic NK cell Cytotoxic Regulatory T cell Activated NK cell Activated MDSC_Angelova MDSC_Trojanoski CD14+HLA bioRxiv preprint doi: https://doi.org/10.1101/2020.05.01.071613; this version posted May 1, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A B Live cells gate C Live cells gate H - L1 - CD14 PD SSC

FSC-H HLA-DR CD14

Debris exclusion gate CD14+ gate CD14+ gate CD14+HLA-DRneg CD14+PD-L1+ CD15+ PMN-MDSCs MONO-MDSCs Inflammatory L1

- monocytes

CD15 + CD14

CD14 PD Monocyte s

CD14 HLA-DR CD14 D Lymphocyte gate CD3+ gate

CD3+PD-1+ H - Activated T cells CD3 SSC

CD3+

CD3 PD-1

E CD4+CD25highFoxP3+ CD4+CD25+ Regulatory T cells H - H - CD4 SSC FoxP3 SSC CD4+ CD4+CD25high

FSC-H CD4 CD25 CD4

F Lymphocyte gate CD16+CD56+ gate NK cells CD3-CD16+CD56+PD-1+ Activated NK cells A - CD16 FSC CD3neg CD16

CD3 CD56 PD-1

CD3-CD16+CD56dim Cytotoxic NK cells CD16

CD56

Supplementary Figure 5: Flow cytometry gating strategies. Myeloid cell populations: CD15+ PMN-MDSCs, CD14+ monocytes, (A); CD14+HLA-DRneg M-MDSCs (B); CD14+PD-L1+ inflammatory monocytes (C). T and NK cell populations: CD3+PD-1+ activated T cells (D); CD4+CD25hi FoxP3+ regulatory T cells (E); CD16+CD56+PD-1+ activated NK cells and CD3-CD16+CD56dim cytotoxic NK cells (F).