Systemic inflammation is a determinant of outcomes of CD40 agonist–based therapy in pancreatic cancer patients

Agonistic anti-CD40 monoclonal antibody (mAb) therapy in combination with chemotherapy (chemoimmunotherapy) shows promise for the treatment of pancreatic ductal adenocarcinoma (PDA). To gain insight into immunological mechanisms of response and resistance to chemoimmunotherapy, we analyzed blood samples from patients (n = 22) with advanced PDA treated with an anti-CD40 mAb (CP-870,893) in combination with gemcitabine. We found a stereotyped cellular response to chemoimmunotherapy characterized by transient B cell, CD56+CD11c+HLA-DR+CD141+ cell, and monocyte depletion and CD4+ T cell activation. However, these cellular pharmacodynamics did not associate with outcomes. In contrast, we identified an inflammatory network in the peripheral blood consisting of neutrophils, cytokines (IL-6 and IL-8), and acute phase reactants (C-reactive protein and serum amyloid A) that was associated with outcomes. Furthermore, monocytes from patients with elevated plasma IL-6 and IL-8 showed distinct transcriptional profiles, including upregulation of CCR2 and GAS6, genes associated with regulation of leukocyte chemotaxis and response to inflammation. Patients with systemic inflammation, defined by neutrophil/lymphocyte ratio (NLR) greater than 3.1, had a shorter median overall survival (5.8 vs. 12.3 months) as compared with patients with NLR less than 3.1. Taken together, our findings identify systemic inflammation as a potential resistance mechanism to a CD40-based chemoimmunotherapy and suggest biomarkers for future studies.

cell and monocyte depletion and CD4 + T cell activation. However, these cellular 48 pharmacodynamics did not associate with outcomes. In contrast, we identified an 49 inflammatory network in the peripheral blood consisting of neutrophils, cytokines (IL-6 and 50 IL-8) and acute phase reactants (CRP and SAA) that was associated with outcomes. 51 Furthermore, monocytes from patients with elevated plasma IL-6 and IL-8 showed distinct 52 transcriptional profiles, including upregulation of CCR2 and GAS6; genes associated with 53 regulation of leukocyte chemotaxis and response to inflammation. Patients with systemic 54 inflammation, defined by neutrophil-lymphocyte ratio (NLR) >3.1, had a shorter median 55 OS (5.8 vs 12.3mo; p=0.0105) as compared to patients with NLR <3.1. Taken together, 56 our findings identify systemic inflammation as a potential resistance mechanism to a 57 CD40-based chemoimmunotherapy and suggest biomarkers for future studies. 58 (27)(28)(29)(30). Furthermore, PDA is often associated with development of a systemic 106 inflammatory response (31) and several markers of systemic inflammation, including 107 neutrophil-lymphocyte ratio (NLR), C-reactive protein (CRP) and serum amyloid A (SAA) 108 are associated with poor outcomes in PDA (32-34). However, whether there is an 109

INTRODUCTION
interaction between systemic inflammation and treatment outcomes to a CD40 agonist 110 remains unexplored in patients with PDA. 111 In this study, we use high-dimensional phenotyping, transcriptional analysis and 112 plasma cytokine analysis to evaluate immune contexture in the peripheral blood of 113 patients with advanced PDA being treated with CD40-based chemoimmunotherapy. We 114 find that although a stereotyped immune response occurs after treatment, cellular 115 pharmacodynamics, including activation of T cells, are not associated with outcomes. 116 Additionally, we show that systemic inflammation defines patients with distinct clinical and 117 biological outcomes after treatment. Taken together, our findings provide novel insight 118 into mechanisms of response and resistance to CD40-based therapy and identify 119 potential biomarkers for future studies. 120 121 Cellular response to  To assess the cellular response to a CD40 agonist in combination with 124 chemotherapy (hereafter referred to as chemoimmunotherapy), we analyzed 125 cryopreserved ficoll-isolated PBMCs from patients (n = 17) with PDA treated with 126 gemcitabine and an agonistic anti-CD40 monoclonal antibody (mAb) (Supplementary 127 Figure S1A). We used a mass cytometry-based (CyTOF) systems approach, which 128 included a 37-marker metal-tagged antibody panel and unsupervised clustering 129 (Phenograph (35)) and meta-clustering (FlowSOM (36) We then studied changes in immune cell meta-clusters representing >1% of baseline 132 PBMCs and saw dynamic remodeling of peripheral blood immune cell composition 133 following chemoimmunotherapy ( Figure 1C). After administration of gemcitabine on day 134 1 of treatment, depletion of monocytes (CD14 + ) was observed on days 3 and 5 with 135 recovery to baseline levels by day 8. Additionally, monocytes were significantly increased 136 at cycle 2, day 1 and cycle 3, day 1 as compared to baseline ( Figure 1D). A minor CD14 + 137 monocyte population, which expressed relatively higher levels of CD66a and CCR6 as 138 compared to the major monocyte population, decreased in frequency on days 5 and 8 139 and then recovered to baseline levels thereafter (Supplementary Figure S1C). A 140 CD56 + CD11c + HLA-DR + CD141 + population also appeared to be impacted by gemcitabine 141 administration and showed reduced frequencies on days 3 and 5, with recovery to 142 baseline by day 8 (Figure 1E). Furthermore, anti-CD40 mAb therapy (administered on 143 day 3) was associated with a transient decrease in B cells (CD19 + ) on day 5 with return 8 to near baseline by day 8, as has been demonstrated previously ( Figure 1F) (25). There 145 was no change in natural killer (CD16 + CD56 + ) cell frequency ( Figure 1G). Granulocytes 146 (CD14 neg CD15 + CD66a + ), which do not represent a major population in ficoll-isolated 147 PBMCs, did not change significantly over the course of treatment (Supplementary 148 Figure S1D). Additionally, there was a relative increase in the frequency of CD4 + T cells 149 among CD45 + cells but not CD8 + T cells at day 5 of treatment ( Figure 1H-I). Finally, a 150 rare population expressing CD56, HLA-DR, CD11c, CD206, CD141, CD86, CX3CR1 and 151 CCR6 was decreased on day 8 as compared to baseline (Supplementary Figure S1E). 152 153 Treatment with CD40-based chemoimmunotherapy is associated with CD4 + T cell 154 activation, which is uncoupled from outcomes 155 Chemoimmunotherapy generates T cell dependent anti-tumor immunity in mouse 156 models of PDA (20, 21). Thus, we next asked whether chemoimmunotherapy impacts T 157 cell activation. To do this, we performed manual gating of the CyTOF data set to assess 158 dual expression of CD38 and HLA-DR by T cells over the course of one cycle of treatment. 159 Gemcitabine administration was followed by a transient decrease in HLA-DR + CD38 + CD8 + 160 T cells on day 3 of treatment, as compared to baseline (Figure 2A-B). Four patients had 161 an increase of CD8 + T cells expressing CD38 and HLA-DR at day 28 of treatment. The 162 overall survival (OS) for these patients was 3.4, 5.1, 8.4 and 8.8 months. HLA-163 DR + CD38 + CD4 + T cells significantly decreased on days 3 and 5 following gemcitabine 164 administration and then significantly increased on day 8 following anti-CD40 mAb 165 treatment, suggesting CD4 + T cell activation (Figure 2C-D). Furthermore, we found 166 heterogeneity in the CD4 + T cell response among patients ( Figure 2E). However, there Similarly, when patients were dichotomized as having an increase or decrease in HLA-169 DR + CD38 + CD8 + T cells at day 8 from baseline, there was no difference in OS among the 170 two groups (Figure 2G). 171

172
An inflammatory network is active in a subset of patients with advanced pancreatic ductal 173 adenocarcinoma 174 We also assessed pre-treatment immune characteristics to define patient specific 175 determinants of outcomes to chemoimmunotherapy. We examined baseline levels of 176 inflammatory cells, cytokines, and acute phase reactants in the peripheral blood of 177 patients. CD4 + T cells, CD8 + T cells and B cells were defined by manual gating of the 178 CyTOF data set (Supplementary Figure S2). Although inter-patient variability in levels 179 of inflammatory markers was present, we found positive correlations among neutrophils, 180 inflammatory cytokines (IL-6 and IL-8) and acute phase reactants (SAA and CRP), 181 suggesting the presence of an inflammatory network ( Figure 3A). Importantly, NLR, 182 which is an established surrogate of systemic inflammation (37), showed a positive 183 correlation with IL-6, IL-8, SAA and CRP and a negative correlation with albumin, absolute 184 lymphocyte count and absolute CD8 + T cell count. These data demonstrate the presence 185 of systemic inflammation in untreated patients with PDA and identify NLR as a measure 186 of systemic inflammation in our patient cohort. 187 We next calculated pre-treatment NLR using clinical blood counts and 188 dichotomized patients using a previously established cutoff of 3.1 (32). Patients with NLR > 189 3.1 were defined as being systemically inflamed (NLR high ) and patients with NLR < 3.1 were defined as being non-inflamed (NLR low ). Classification of patients based on NLR 191 identified biologically distinct groups based on pre-treatment inflammatory factors. 192 NLR high patients had significantly higher levels of IL-6, IL-8, SAA, CRP and lower levels 193 of albumin as compared to NLR low patients (Figure 3B-C). Other cytokines associated 194 with immune activation, including IL-2, IL-4, IL-5, IL-1b, IFN-g, IL-10, IL-12 and TNF, were 195 not found to be elevated at baseline (Supplementary Figure S3A).  clinical blood counts, we found NLR high patients to have significantly higher numbers of 197 total white blood cells and neutrophils, numerically higher numbers of monocytes and 198 significantly lower numbers of lymphocytes as compared to NLR low patients and healthy 199 volunteers (HVs) ( Figure 3D). Both NLR high and NLR low patients had similar numbers of 200 eosinophils, basophils and platelets (Supplementary Figure S3B). Using manually 201 gated CyTOF data from the pre-treatment time point, we detected lower absolute 202 numbers of CD8 + T cells and NK cells in the peripheral blood of NLR high patients 203 compared to NLR low patients, but this was not significant (Supplementary Figure S3C). 204 In addition, we observed no significant difference in the percentage (of CD45 + cells) of B 205 cells, T cells, NK cells, DCs or the CD4:CD8 T cell ratio among NLR high and NLR low 206 patients (Supplementary Figure S3D-E). We also analyzed baseline cell clusters 207 defined by FlowSOM that represented >1% of CD45 + cells among the two groups and 208 found increased CD14 + monocytes in NLR high patients as compared to NLR low patients. 209 However, this difference was not significant after corrections for multiple testing 210 (Supplementary Figure S3F). Together, these data show the presence of an active 211 inflammatory network in the peripheral blood of a subset of patients with PDA.  The kinetics of peripheral blood inflammatory markers suggest distinct responses to 237

CD40-based chemotherapy among patients with systemic inflammation 238
We next evaluated treatment associated changes in inflammatory markers among 239 NLR high and NLR low patients. We first studied cellular dynamics in the peripheral blood 240 based on clinical blood counts. In both groups, neutrophils decreased on treatment days 241 8 and 15. However, neutrophils were significantly higher in NLR high patients at all time 242 points of cycle 1 ( Figure 5A). Monocytes were also found to decrease on treatment day 243 3, after gemcitabine administration. Additionally, in NLR high patients, monocytes 244 recovered to levels significantly higher than seen in NLR low patients on day 8 and 245 remained significantly elevated at the end of cycle 1 ( Figure 5B). Lymphocytes were 246 significantly higher in NLR low patients at baseline but became similar among the groups 247 during treatment ( Figure 5C). Given these changes in neutrophils and lymphocytes with 248 treatment, we next analyzed the dynamics of NLR after beginning treatment 249 Figure S5A). Over the course of one cycle of treatment, NLR remained 250 significantly higher in the NLR high group as compared to the NLR low group. Additionally, in 251 both groups, there was a transient decrease in NLR at day 15 after treatment, which 252 coincided with a treatment related decrease in neutrophils. At the end of one cycle of 253 treatment, NLR in all NLR high patients remained >3.1, while in NLR low patients, 3 of 8 254 patients had converted to a NLR >3.1. We also assessed whether there were differences 255 in the pharmacodynamic response of FlowSOM defined clusters (>1% of CD45 + cells) 256 from the CyTOF data set among NLR high and NLR low patients. This analysis was limited 257 by small numbers of patients in each group and heterogeneity in cluster frequency. characteristic of a CD40 agonist, appeared to recover more rapidly to baseline by day 15 261 in NLR low patients, whereas NLR high patients continued to have B cell frequencies 262 significantly lower than baseline on day 15 (Supplementary Figure S5H). 263

(Supplementary
Furthermore, we assessed for acute changes in inflammatory cytokines by 264 analyzing patient plasma collected pre-treatment and between 5 min and 24 hours after 265 treatment with gemcitabine or anti-CD40 mAb therapy. While modest changes in IL-6, IL-266 8 and IL-10 plasma levels were observed after gemcitabine administration, anti-CD40 267 mAb therapy was associated with significant increases in plasma concentrations of IL-6, 268 IL-8 and IL-10 with a peak at 2-6 hours after treatment. Notably, baseline and peak IL-6 269 levels were highest in NLR high patients ( Figure 5D). In contrast, although baseline IL-8 270 levels were higher in NLR high patients, peak IL-8 levels were similar among the two groups 271 ( Figure 5E). There was no difference in IL-10 plasma levels between NLR high and NLR low 272 patients ( Figure 5F). Additionally, the fold change in inflammatory cytokine concentration 273 (peak relative to baseline) was different among NLR high and NLR low patients. For example, 274 while there was no difference in fold change for IL-6 or IL-10, there was a significantly 275 higher fold change in plasma IL-8 in NLR low patients as compared to NLR high patients 276 ( Figure 5G-I). Taken together, these data show that distinct cellular and cytokine 277 pharmacodynamics are present in NLR high and NLR low patients after treatment with 278 chemoimmunotherapy. 279 groups. Of the 22 patients included in our study, 12 patients were NLR high and 10 patients 284 were NLR low ( Figure 6A). Patient characteristics were well balanced, although more 285 NLR high patients had liver metastases (100% vs. 70%) and more NLR low patients had 286 peritoneal metastases (30% vs. 0%) (Supplementary Table S1). In a univariate analysis, 287 OS was significantly shorter in NLR high patients as compared to NLR low patients (5.82 vs 288 12.3 months; p = 0.0105) ( Figure 6B). Additionally, we performed a multivariate analysis 289 including sex, ECOG performance status, tumor burden and age and found NLR > 3.1 290 continued to correlate with worse OS (HR 3.87; CI 1.04 -14.38; p = 0.043) 291 Figure S6A). Intriguingly, when patients were dichotomized using the 292 median of acute phase reactants (SAA and CRP) and inflammatory cytokines (IL-6 and 293 IL-8), only elevated acute phase reactants, not inflammatory cytokines, were significantly 294 associated with poor OS (Figure 6C-F). 295

296
In this study, we used high-dimensional cellular phenotyping and plasma cytokine 298 analysis to evaluate the immune response to a CD40 agonist in combination with 299 gemcitabine chemotherapy in the peripheral blood of patients with advanced PDA. 300 Notably, CD40-based chemoimmunotherapy was associated with transient activation of 301 CD4 + T cells and changes in monocytes and B cells. However, T cell activation in 302 response to therapy was not associated with outcomes. In contrast, the presence of a 303 pre-existing systemic inflammatory response was found to associate with reduced 304 survival. Taken together, our data suggest that although a CD40 agonist can induce T 305 cell activation in patients, additional determinants of response exist. Furthermore, our 306 findings identify systemic inflammation as a potential resistance mechanism to CD40-307 based chemoimmunotherapy. chemotherapy as compared to gemcitabine alone in combination with a CD40 agonist 314 might generate distinct immune responses as has been suggested in preclinical models 315 (20, 43). In this regard, an ongoing Phase Ib/II trial (NCT03214250), studying the 316 combination of a CD40 agonist (APX005M) and gemcitabine plus nab-paclitaxel with or 317 without nivolumab, will be informative (18).
conclusions regarding efficacy measures. However, a subset analysis of the MPACT 320 Phase III trial which examined NLR as a determinant of outcomes to gemcitabine plus 321 nab-paclitaxel versus gemcitabine monotherapy provides some context for our findings 322 (44). In this study, a NLR cutoff of 5 was used. For patients with NLR < 5, treatment with 323 gemcitabine plus nab-paclitaxel compared to gemcitabine monotherapy was associated 324 with median OS of 10.9 and 7.9 months, respectively. In contrast, median OS for patients 325 with NLR > 5 was 5.6 months for gemcitabine plus nab-paclitaxel and 4.3 months for 326 gemcitabine monotherapy. Although we used a lower NLR cutoff in our study, we have 327 also examined survival outcomes based on a NLR of 5 (Supplementary Table S2). We 328 found that median OS was 11.7 months (NLR < 5) and 5.8 months (NLR > 5) for CD40-329 based chemoimmunotherapy which compares favorably and suggests that CD40-based 330 treatment may be most effective in patients with a low NLR. 331 Cytotoxic chemotherapy can have both immunosuppressive and immune 332 stimulating capacity (45). Importantly, the optimal sequencing of chemotherapy in 333 combination with CD40-based immunotherapy remains ill-defined. In our study, we found 334 near complete depletion of monocytes and a CD56 + CD11c + HLA-DR + CD141 + population 335 in the peripheral blood after chemotherapy administration, which was transient, but in the mechanism of action of a CD40 agonist (20, 22). Thus, chemotherapy, when contrast, administration of anti-CD40 therapy prior to chemotherapy may leverage the 343 anti-stromal effects of a CD40 agonist, thereby potentiating the activity of chemotherapy 344 (22,23). To this end, treatment with a CD40 agonist delivered at least 4 days prior to 345 chemotherapy is safe and produces promising anti-tumor activity in mouse models of 346 PDA (23). However, the timing of chemotherapy treatment is critical, as delivering a CD40 347 agonist within 3 days prior to chemotherapy can trigger lethal hepatotoxicity in mice (23, 348 48). An alternative strategy that remains unexplored clinically, is whether CD40-based 349 immunotherapy might provide benefit in the maintenance setting after induction 350 chemotherapy. Maintenance immunotherapy with checkpoint inhibition was recently 351 established in the JAVELIN-100 trial, which showed improved survival in patients with 352 advanced bladder cancer treated with maintenance anti-PD-L1 therapy following 353 induction chemotherapy (49). Our results suggest further study is warranted to determine 354 the optimal sequencing of anti-CD40 therapy and chemotherapy. 355 Pre-clinical evidence shows that a CD40 agonist, especially when combined with 356 checkpoint inhibition, leads to CD4 + T cell mediated anti-tumor immune responses (50, 357 51). We observed a CD4 + T cell response in the peripheral blood of patients following 358 treatment with CD40-based chemoimmunotherapy, providing evidence that this biology 359 can also be observed in humans. Intriguingly, bona fide cytotoxic CD4 + T cells have been 360 described in patients with bladder cancer and when present intratumorally are associated 361 with improved outcomes to checkpoint inhibition (52). Additionally, we saw no consistent 362 evidence of CD8 + T cell activation. In mouse models of PDA, both CD8 + and CD4 + T cells 363 are required for the activity of CD40-based chemoimmunotherapy (21). It remains possible that absence of CD8 + T cell response limits the full therapeutic potential of CD40-365 based treatment and contributes to the lack of association between cellular 366 pharmacodynamics and outcomes. 367 One limitation to our study, is that tissue biopsies were not available for analysis 368 and we cannot confirm if peripheral blood immune dynamics are representative of 369 responses occurring in secondary lymphoid organs or tumor. To this end, pre-clinical 370 models show that a CD40 agonist can trigger systemic T cell responses without impacting 371 the intra-tumoral T cell compartment (21,22). However, activation of circulating 372 monocytes by a CD40 agonist is correlated with myeloid cell activation within tumors (23). 373 Moreover, CD40-activated monocytes are functionally important and can sensitize tumors 374 to chemotherapy (23). Additionally, an association between peripheral blood leukocyte 375 composition and outcomes in patients with PDA has been shown by others. For example, 376 the presence and diversity of peripheral blood T cells reactive against the tumor-377 associated antigen mesothelin is associated with prolonged disease-free survival in 378 patients with PDA treated with immunotherapy (53, 54). Taken together, these data 379 highlight the potential of peripheral blood leukocyte changes to associate with immune 380 cell dynamics in the TME and correlate with clinical outcomes. 381 Inflammatory monocytes and tumor associated macrophages are intimately 382 associated with PDA resistance to productive T cell immunosurveillance (55). Consistent 383 with the reports of others, we found monocytes to be elevated in patients with poor 384 outcomes (40). Moreover, we identified upregulation of CCR2 and GAS6 in CD14 + 385 monocytes from patients with elevated plasma levels of inflammatory cytokines. One 386 potential limitation of our approach was that we evaluated monocytes in patients defined by plasma cytokine levels rather than NLR. Nonetheless, our findings provide insight into 388 associations among specific inflammatory cytokines and monocyte phenotype. To this 389 end, targeting of CCR2 + macrophages using CCR2 inhibitors is an effective method of 390 tumor control in mouse models of PDA and has shown safety and potential clinical activity 391 in combination with FOLFIRINOX in patients (40, 56). However, we have also shown that 392 CCR2 inhibition can impair the capacity of a CD40 agonist to improve the efficacy of 393 chemotherapy in mouse models of PDA (23). Gas6, which is an AXL kinase ligand, may 394 be an alternative target. Notably Gas6 has been implicated in PDA tumor progression (57, 395 58). Our findings suggest that inflammatory monocytes may be a source of GAS6. 396 Furthermore, blockade of AXL has shown promise in preventing PDA tumor growth (57). Immunotherapy has thus far failed to improve outcomes for patients with PDA. 441 However, myeloid targeted immunotherapy is a distinct treatment approach that has 442 shown promise. In this study, we examined the activity of a CD40 agonist, which can drive 443 innate and adaptive immunity. Unexpectedly, we saw no consistent evidence of CD8 + T 444 cell activation, and CD4 + T cell activation did not correlate with outcomes. Furthermore, 445 our data suggests chemotherapy may have a detrimental impact by eliminating 446 monocytes and DCs, which are cells that are fundamental to facilitating T cell dependent 447 immune responses. Thus, non-T cell-based mechanisms may govern the therapeutic 448 activity of systemic CD40 activation in combination with gemcitabine. Our data also 449 suggest that acute phase reactants (SAA and CRP) and monocyte transcriptional 450 programming may be determinants of response to CD40-based treatment. Overall, our 451 study provides insight into the cellular and biological mechanisms of response and 452 resistance to a CD40 agonist combined with chemotherapy in patients with advanced