Enhanced Complement Expression in the Tumor Microenvironment Following Neoadjuvant Therapy: Implications for Immunomodulation and Survival in Pancreatic Ductal Adenocarcinoma

Background Neoadjuvant therapy (NAT) is increasingly being used for pancreatic ductal adenocarcinoma (PDAC) treatment. However, its specific effects on carcinoma cells and the tumor microenvironment (TME) are not fully understood. This study aims to investigate how NAT differentially impacts PDAC’s carcinoma cells and TME. Methods Spatial transcriptomics was used to compare gene expression profiles in carcinoma cells and the TME between 23 NAT-treated and 13 NAT-naïve PDAC patients, correlating with their clinicopathologic features. Analysis of an online single-nucleus RNA sequencing (snRNA-seq) dataset was performed for validation of the specific cell types responsible for NAT-induced gene expression alterations. Results NAT not only induces apoptosis and inhibits proliferation in carcinoma cells but also significantly remodels the TME. Notably, NAT induces a coordinated upregulation of multiple key complement genes (C3, C1S, C1R, C4B and C7) in the TME, making the complement pathway one of the most significantly affected pathways by NAT. Patients with higher TME complement expression following NAT exhibit improved overall survival. These patients also exhibit increased immunomodulatory and neurotrophic cancer-associated fibroblasts (CAFs); more CD4+ T cells, monocytes, and mast cells; and reduced immune exhaustion gene expression. snRNA-seq analysis demonstrates C3 complement was specifically upregulated in CAFs but not in other stroma cell types. Conclusions NAT can enhance complement production and signaling within the TME, which is associated with reduced immunosuppression in PDAC. These findings suggest that local complement dynamics could serve as a novel biomarker for prognosis, evaluating treatment response and resistance, and guiding therapeutic strategies in NAT-treated PDAC patients.


INTRODUCTION
Pancreatic ductal adenocarcinoma (PDAC) is a highly aggressive cancer with a five-year survival rate of 11% 1 .It is projected to become the second leading cause of cancer-related deaths in the United States by 2030 2 .Only 15-20% of patients are eligible for surgery, as most patients present with locally advanced unresectable disease or distant metastases 3 .Furthermore, since most patients who undergo surgical resection ultimately succumb to their disease, systemic chemotherapy is a necessary therapeutic component for most patients 4 .
PDAC presents a significant therapeutic challenge due to its resistance to various treatment modalities including immune checkpoint inhibitor therapy [5][6][7] .It exhibits extensive genetic diversity, leading to the development of subpopulations with diverse molecular alterations and resistance mechanisms 8 .The tumor microenvironment (TME) in PDAC is highly immunosuppressive, characterized by the presence of immunosuppressive cells, cytokines, and chemokines that impede effective anti-tumor immune responses despite the presence of large numbers of effector T cells [9][10][11] .Additionally, the TME plays a pivotal role in influencing tumor growth and metastasis through its impact on mitochondrial biogenesis, blood supply, and nutrient availability 12 .These factors collectively contribute to the high therapeutic resistance observed in PDAC, highlighting the need for innovative combinational strategies to improve treatment outcomes.
Neoadjuvant therapy (NAT) has emerged as the preferred approach for managing resectable and borderline resectable PDAC over adjuvant therapy for several reasons.NAT offers better tolerability for patients by administering chemotherapy and/or radiation prior to surgery, taking advantage of patients' relatively better overall health status and minimizing treatment-related complications 13,14 .NAT also effectively increases tumor resectability and R0 resection rate by downstaging locally advanced disease and facilitating complete tumor removal 7,15 .Moreover, NAT may target early metastases, potentially improving treatment outcomes 7,15 .NAT also provides an opportunity to observe an individual patient's response to therapy and to select suitable candidates for surgery 7,15 .Some studies suggested that NAT may be associated with improved overall survival rates compared to adjuvant therapy 13,14 .Furthermore, recent studies suggested that NAT may have the potential to prime the stroma and enhance the response to immunotherapy in various cancers 16,17 .By modulating TME, NAT may render PDAC more susceptible to immunotherapeutic interventions 16,17 .
Single-cell RNA sequencing (scRNA-seq) is a valuable tool for characterizing TME as it provides insights into cellular heterogeneity and molecular subtypes 18,19 .Several scRNA-seq studies have explored the PDAC TME, offering valuable insights into this complex ecosystem [20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36] .However, scRNAseq techniques lack spatial context, and are susceptible to cell dissociation-associated artifacts.Spatial transcriptomic analysis complements and extends the scope of single-cell studies 37,38 by providing valuable insights into the spatial distribution of gene expression and a holistic characterization of the PDAC TME.To obtain a comprehensive view of the PDAC TME and its response to NAT, we performed a comparative spatial transcriptomic study using formalin-fixed paraffin-embedded (FFPE) specimens from PDAC patients who received NAT before surgery versus those who underwent upfront surgery without NAT.

Intra-and intertumoral heterogeneity of NAT response in PDAC carcinoma and TME
The clinicopathologic characteristics of the 36 patients in our study cohort are summarized in Figure 1A and Supplementary Table 1.This cohort includes 13 patients who had upfront surgery (the naïve group) and 23 age-, gender-, and stage-matched patients who had surgery after receiving neoadjuvant chemotherapy (the NAT group).The NAT consisted of either FOLFIRINOX or gemcitabine treatment, with or without radiation.Only resected primary PDAC specimens were used in this study except for 4 patients whose pre-NAT biopsies were included in the naïve group and post-NAT resection specimens were included in the NAT group.A separate pairwise longitudinal comparison of the pre-NAT biopsy versus the post-NAT resection from the same patient was also done for these 4 patients (Figure 1E and Supplementary Tables 2).To assess the relative extents of intra-vs.intertumoral heterogeneity, 2 to 5 ROIs (region of interest) were selected from each case, amounting to a total of 119 ROIs from the 36 patients in the entire cohort (Supplementary Figure 1).Using immunofluorescence staining, each ROI was segmented into two AOIs (Area of Illumination): the carcinoma AOI (pan-CK+, SMA-) containing only malignant cells, and the TME AOI (pan-CK-, SMA+) which comprised of various types of stroma cells and immune cells (Figure 1B).RNA libraries were prepared from these 238 AOIs, of which 223 (93.7%) passed the quality control and were used for analysis, including 42 carcinoma AOIs and 43 TME AOIs from the naïve group, and 72 carcinoma AOIs and 66 TME AOIs from the NAT group.
First, we compared the gene expression profiles between carcinoma AOIs versus TME AOIs.
Carcinoma AOIs showed significantly higher expression of 302 and 321 genes than TME AOIs in the naïve group and the NAT group, respectively, with an overlap of 261 genes which included KRT19, KRT7, KRT18, MUC1, CDH1, EPCAM, JUP, ITGA3, ITGB4, ERRB2, LAMB3, CAPN2, LGALS3, LCN2 etc.In contrast, TME AOIs showed significantly higher expression of 286 and 294 genes than carcinoma AOIs in the naïve group and the NAT group, respectively, with an overlap of 250 genes including PDGFRB, COL6A3, FCGR2A/B, IL1R1, etc. (Supplementary Figure 2, Supplement Data 1).Hierarchical clustering analysis successfully grouped these 223 AOIs into the carcinoma cluster and the TME cluster, with a limited number of datapoints appearing in between these two distinct clusters.There is no discernible differentiation between the naïve and NAT AOIs within both the carcinoma cluster and TME cluster (Figure 1C).

Next, we used a dimension reduction technique to visualize the clustering pattern of these 223
AOIs in two major UMAP dimensions based on their transcriptomics profiles.Like the results from the hierarchical clustering analysis, these 223 AOIs were largely separated into carcinoma cluster and the TME cluster, with only a limited number of AOIs from each category intermingling at the peripheries of these two clusters (Figure 1D).However, notable differentiation based on treatment status (i.e., naïve vs. NAT) was once again not evident, suggesting both carcinoma cells and the TME responded to NAT heterogeneously at the transcriptomics level.
To investigate the roles of intra-versus intertumoral heterogeneity in NAT response, we also made a UMAP plot of the four pairs of samples in which the pre-NAT biopsy was compared with the post-NAT resection specimen from same patient.In addition to the clear separation of carcinoma AOIs from TME AOIs, a near-complete separation was also noted between the naïve group (i.e., pre-NAT biopsy) and the NAT group (i.e., the post-NAT resection) in both carcinoma AOI cluster and TME AOI cluster (Figure 1E).In accordance with the anticipated relative magnitudes of intra-versus intertumoral heterogeneity, the AOIs from the same patient tended to aggregate more closely together than those from different patients (Figure 1E).
To compare the cellular composition within carcinoma AOIs and TME AOIs, we calculated the gene signature scores for different types of stromal cells and immune cells using our transcriptomics data.
As compared to carcinoma AOIs, TME AOIs contained significantly more fibroblasts and immune cells across the board (Supplementary Figure 3).
Next, we investigated the impact of NAT on carcinoma cells in more detail by utilizing the recently defined gene signatures for the seven malignant cell programs representing the lineages of carcinoma cells and the seven malignant cell states 25 .Signature scores for each of these malignant programs and cell states were calculated for all carcinoma AOIs based on the geometric mean of expression levels for their respective signature genes set.Their association with NAT was assessed after controlling for age, gender, cancer grade and stage.We found that there was a significant increase of acinar-like program (p = 0.011, Figure 2C), a significant decrease of cyclining_G2/M state (p = 0.001), and a marginally significant decrease of TNF-NFkB signaling (p = 0.057) in the NAT group.These findings were also supported by our morphologic analysis, which showed increased apoptosis along with decreased proliferation and mitosis in the carcinoma cells of the NAT group compared to the naïve group (Figure 2D).
Additionally, there were functional terms related to immunity, including KW-0399 Innate immunity (FDR = 9.61e-3); KW-0391 Immunity (FDR = 5.04e-2), and multiple pathways related to cancer signaling such as hsa04115 p53 signaling pathway (FDR = 1.73e-4) (Figure .3B).To illustrate the gene components for these enrichment terms, chord maps were made for the most significantly enriched GO, KEGG and UP_KW functional terms and their related genes.The 5 complement genes (C1S, C1R, C3, C4B and C7) contribute to not only complement-related pathways, but also other key functional terms related to innate and adaptive immunity (Figure 3C).These findings imply that NAT-induced upregulation of complement expression within TME may serve as a local regulatory hub in TME and contribute to the overall immunomodulation in PDAC.
To investigate how NAT influences the tumor immune microenvironment (TIME) in PDAC, signature scores for all major immune cell types were calculated for each TME AOI and tested for association with NAT treatment while controlling age, gender, carcinoma grade and stage.There was no statistically significant change of signature scores for most types of immune cells in the TIME of the NAT group compared to the naïve group, except for mast cells which were increased with a marginal statistical significance (p=0.09)(Figure 4A).Next, we studied the effect of NAT on the cancer-associated fibroblasts (CAFs) within PDAC TME using the expanded CAF classification system recently defined by Hwang et al. in 2022, 25 in which CAFs are categorized into 4 specific programs based their gene expression profiles, including myofibroblastic, adhesive, immunomodulatory, and neurotrophic CAFs.
We computed signature scores for all four programs of CAFs within each TME AOI and analyzed the association with NAT treatment while controlling for age, gender, carcinoma grade and stage.We found that immunomodulatory (p=0.023) and neurotropic CAFs (p=0.002) were significantly increased in the TME of the NAT group compared to the naïve group (Figure 4B).
To validate these findings of our spatial transcriptomics study and also identify which specific cell types in the TME are primarily responsible for the upregulation of complement genes following NAT, we analyzed a single-nucleus RNA (snRNA-seq) sequencing dataset from Hwang et al. 25 , which was archived in the single cell portal.This dataset encompassed the gene expression profiles of 22,164 genes across 88031 cells from 15 NAT-naïve patients and 50516 cells from 11 NAT-treated patients.Our analysis revealed a significant increase of C3 expression in CAFs of the NAT-treated patients compared to NAT-naïve patients (p = 0.044).Based on generalized linear mixed model analysis, the proportion of C3 high expression CAFs was found to be significantly higher in NAT-treated patients than NAT-naïve patients, (10.66% vs. 5.03%, p = 0.012).(Figure 4C).In contrast, no notable changes in C3 expression level were observed in other types of cells including tumor cells, immune cells and endothelial cells.A more detailed analysis on subtypes of CAFs showed that all subtypes of CAFs exhibited increased complement C3 gene expression, with the most pronounced increases were observed in immunomodulatory and myofibroblastic CAFs.These results not only validated our spatial transcriptomics findings on NAT-induced C3 upregulation in a different cohort of patients, but also provided direct evidence that the CAFs may be the primary contributing cells of the complement differential expression in TME.

NAT-induced upregulation of TME complement expression is associated with improved survival and immunomodulation
To assess the clinical and prognostic significance of NAT-induced upregulation of TME complement, we stratified the NAT group into high and low complement subgroups based on clustering analysis of the mean expression levels of the five complement genes upregulated in the TME by NAT (C3, C4, C7, C1S and C1R).Out of 23 NAT-treated patients in our cohort, 9 were clustered into the high complement subgroup and 14 into the low complement subgroup (Figure 5A).Kaplan-Meier analysis indicated that the high complement subgroup demonstrated a significantly better overall survival compared to the low complement subgroup (p=0.023)(Figure . 5B).Additionally, compared to the naïve group, the high complement subgroup showed a marginally better survival trend (p<0.1), while the low complement subgroup's survival rates were not significantly different (Figure . 5B).Multivariate Cox proportional hazard ratio analysis further identified a high TME complement level as an independent favorable prognostic factor (hazard ratio = 0.21; 95% Confidence Interval (CI): 0.06 to 0.81, p=0.023) for NAT-treated patient (Figure . 5C).In contrast, a positive surgical resection margin (R1) was recognized as an independent adverse prognostic factor (HR=5.45, 95% CI=1.28-23.27)(Figure .5C).
To further clarify the link between the elevated complement expression in TME and survival benefits in post-NAT patients, we also conducted a longitudinal paired analysis of complement C3 level in the TME of the four patients for whom we had both pre-NAT biopsy and post-NAT resection samples.
We examined the changes in TME C3 levels and their correlation with clinical outcome.Among these 4 patients, two of them displayed a significantly increased TME C3 level following NAT and showed improved overall survival compared to the rest two patients who did not show increase of TME C3 level post-NAT (Supplementary Figure 6).These results imply that patients with a significant upregulation of TME complement level in response to NAT could be indicative of improved overall survival.
To elucidate the mechanism behind the association between the NAT-induced upregulation of TME complement and the observed improvement in patient survival, we compared the signature scores for all immune cells and CAF subtypes between the high and low complement subgroups.Our results showed that the high complement subgroup exhibited significantly higher signature scores for mast cell (p = 0.037) and CD4 + T cells (p = 0.039) compared to the low complement subgroup (Figure . 6A).
Moreover, the high complement subgroup showed significantly higher signature scores for immunomodulatory and neurotrophic subtypes of CAFs (p = 3.46e-4, Figure 6B).In addition, the high TME complement subgroup also showed a notable reduction of expression for immune exhaustion gene signature (p = 0.038) while maintaining stable expression of immune cell cytotoxicity gene signature (Figure . 6B).Further analysis using snRNA-seq data from Hwang et al. 25 revealed reduced expression of immune checkpoint gene CD274 (p=0.019) and immune exhaustion gene TIGIT (p=0.034) in NATtreated patients compared to NAT-naïve patients.These results suggest that NAT-induced upregulation of TME complement expression may lead to reduced immune exhaustion and immunosuppression in PDAC, thereby contributing to the improved survival in these patients (Figure 6C).

DISCUSSION
Our results revealed that NAT differentially remodels the carcinoma cells and the TME of PDAC through distinct mechanisms.By comprehensively comparing the spatial transcriptomics profiles of more than 1800 genes related to cancer biology, TME and immune responses between a sizable cohort of patients treated with NAT versus those who were NAT-naïve, we demonstrated that NAT induced 44 DEGs in carcinoma cells and 52 DEGs in the TME.Gene set enrichment analysis revealed 9 significantly up-or downregulated gene sets and functional terms in NAT-treated carcinoma cells, compared to 61 in NAT-treated TME.These results indicate that NAT may exert a more extensive and profound remodeling effect on the TME than on the carcinoma cells per se.In the NAT-treated carcinoma cells, most DEGs are associated with apoptosis regulation, cell cycle control, and DNA repair.Among them, the most significantly upregulated DEG was DDB2 (p=0.00077), which encodes DNA damage-binding protein 2, a nuclear protein crucial for DNA repair and genomic stability.Additionally, DDB2 has been shown to suppress epithelial-to-mesenchymal transition (EMT) and enhance sensitivity to chemotherapy in PDAC 39 .Conversely, in the TME of NAT-treated PDAC, most DEGs are associated with complement signaling and regulation of innate and adaptive immune function.Among them, the most significant downregulated DEG was HDAC2, which encodes histone deacetylase 2, a transcription factor capable of inducing inflammatory gene expression in CAFs and supporting carcinoma growth in PDAC.Recent studies have demonstrated the role of HDAC2 in facilitating pancreatic cancer metastasis 40 .
Notably, we observed a coordinated upregulation of five key complement genes (C3, C4B, C7, C1R, C1S) in the TME of NAT-treated PDAC as compared to those not treated by NAT.Due to the collective upregulation of these five complement genes, the complement pathway emerged as the most significantly enriched pathway (FDR=1.04x10 - ) among the NAT upregulated gene sets.It is worth emphasizing that this NAT-induced upregulation of complement expression was spatially confined to TME.Traditionally, complement proteins were primarily considered to be produced in the liver and circulating in the bloodstream to aid the immune function by opsonization 41 .However, our results suggest that the complements may also be produced locally within TME and play an important role in regulating immune response.Analysis of snRNA-seq data validated our spatial transcriptomics findings, and pinpointed CAFs as the primary mediators of this NAT-induced upregulation of complement in TME, while complement gene expression in carcinoma cells and immune cells were not notably changed.As fibroblasts are the most abundant cell type in TME, our results suggest that CAFs are likely the principal factor contributing to the local production of complement proteins within PDAC TME.
Our findings align with several recent single-cell or single-nucleus RNA-sequencing (sc/snRNA-Seq) studies.In 2021, Chen et al. identified a distinctive subtype of CAF, which they named "complement-secreting CAFs" (cs-CAFs), characterized by their enrichment with complement system genes.They found that cs-CAFs were predominantly located near malignant glands, and mainly existed in early-stage PDAC and diminished in late-stage PDAC 42 .More recently, Croft et al. demonstrated that some CAFs located farther from malignant glands can also produce complement proteins, and their increase was associated with improved patient survival 43 .In 2022, Hwang et al. 25 expanded the CAF classification system to include four subtypes: myofibroblastic, adhesive, immunomodulatory, and neurotrophic.They noted an overlap between the latter three subtypes and the previously defined inflammatory type CAFs, with immunomodulatory CAFs showing enrichment in complement genes 25 .
Similarly, Werba et al. also observed complement gene enrichment in inflammatory CAFs 34 .However, none of these sc/snRNA-Seq studies specifically investigated the impact of NAT on complement gene expression within TME.Our research, for the first time, demonstrated a NAT-induced upregulation of TME complement in CAFs.Further studies, especially those using single-cell spatial transcriptomics, are warranted to characterize how NAT influences the spatial distribution and adaptability of different subtypes of CAFs.A recent study by Shiau et al. using single cell spatial transcriptomics revealed that NAT induced significant changes in interaction strength between malignant cells and inflammatory CAFs mediated by enriched IL-6ST signaling 31 .Correspondingly, our study also showed a significant upregulation of IL-6ST expression in post-NAT TME, contributing to the regulation of immune function (Figure 3A and 3C).Taken together, these results suggest that NAT can substantially remodel the TME and influence the interplay between carcinoma cells and inflammatory CAFs.
As another interesting finding of our study, NAT-induced upregulation of complement expression in TME was found to be associated with modulation of the TIME and decreased expression of immune exhaustion markers, which was demonstrated in our spatial transcriptomics analysis (Figure 6B) and also validated by the snRNA-seq data.Importantly, the NAT-treated patients with higher TME complement demonstrated improved overall survival (Figure 5B).The detailed cellular and molecular mechanisms underlying these observations are yet to be further clarified.It is known that some cleaved complement proteins (such as C3a, C5a, etc.) are potent chemoattractant and can recruit immune cells into TIME [44][45][46] .
Recent evidence also indicated that a tonic level of complements in the microenvironment is required for maintaining immune cell homeostasis and function via the non-canonical intracellular complement (i.e., complosome) signaling 47,48 .Our data suggests that TME complement may serve as a regulatory hub in the local tumor microenvironment for both innate and adaptive immune functions via both canonical and non-canonical signaling mechanisms.Further studies are warranted to elucidate the specific details of this novel mechanism.
Our study is the first spatial transcriptomics study specifically aimed at comparing the differential remodeling effect of NAT on carcinoma cells and the TME using FFPE tissue of primary PDAC.The cohort size is relatively limited, yet comparable to multiple scRNA-seq studies recently published in the field 23,42 25 34 43 . 49The main limitation of our approach is that the spatial transcriptomics method used here does not provide single-cell resolution.On the other hand, since there is no cell isolation and processing needed as in single-cell or single-nucleus RNA-seq, the method used in this study reduces technical variability and batch effect and is not impacted by the artifactual transcriptomics changes due to the procedure-induced cell stress.This is especially important for PDAC in which all cellular components are buried in excessive dense extracellular matrix which makes cell isolation particularly challenging.In addition, by pooling mRNAs from each AOI, this method can achieve higher mRNA concentration per sample, thus facilitating deeper sequencing and reduced dropouts, thereby enhancing the signal-to-noise ratio.Consequently, it enables more robust detection of weakly expressed genes.Recently, the scRNAseq study by Werba et al. observed generally decreased cell-cell interaction and high volume of putative interactions between CAFs and tumor-associated macrophages (TAMs) in the TME of NAT-treated PDAC 34 .No difference was noted in CAF subpopulation distributions between NAT-treated vs. naive groups 34 .Hwang et al. discovered that NAT could increase the adhesive CAFs and decrease the myofibroblast CAFs.They also found that the number of immunomodulatory CAFs was higher in patients treated with regular NAT than those with NAT plus losartan 25 .In the most recent study using single-cell spatial transcriptomics, Shiau et al showed a marked increased ligand-receptor interaction between CAFs and malignant cells in NAT-treated PDAC 31 .Therefore, our data have not only confirmed previous sc/snRNA-seq studies but also provided additional valuable insights thanks to the preserved spatial context.
The interesting findings from our study call for further investigations on the functional details of the molecular mechanisms, by which local complement signaling modulates immune cells functions within TME.The complement system plays a pivotal role in regulating both innate and adaptive immune functions in inflammatory and neoplastic conditions through both canonical and non-canonical pathways.
For instance, Kumar S. et al. demonstrated that adventitial fibroblasts activated by pulmonary hypertension could release complement proteins through paracrine secretion of extracellular vesicles, subsequently reprogramming macrophage metabolism. 50Under neoplastic condition, studies have showed that chemotherapy can enhance extracellular vesicle release of complement proteins in breast cancer models and paradoxically increase the risk of lung metastasis 51 52 .On the other hand, other studies demonstrated that complement was a central mediator of radiotherapy-induced anti-tumor immunity 53 .
These studies highlight the dual functionality of complement signaling as a double-edge sword.Further research is needed to gain deeper understanding of this complex context-dependent effect of complement signaling on tumor and immune cells within TME.Such knowledge could pave the way for new therapeutic strategies that target the dynamics of local complement.
In conclusion, by using spatially resolved transcriptomics, we demonstrated that NAT remodels the carcinoma cells and the TME of PDACs through different mechanisms.NAT can significantly upregulate complement gene expression within TME, which is associated with reduced immune exhaustion and improved survival.These results suggest that NAT may potentially attenuate immune cell exhaustion and enhance immune surveillance in some PDAC patients by upregulating local complement production and signaling.Those patients who did not respond to NAT with a significant upregulation of TME complement may have more immune exhaustion and thus benefit from combinational immune checkpoint blockade therapy.Therefore, assessment of the local complement dynamics in the TME may provide more accurate prognostic guidance for NAT-treated patients, which may help to stratify them for more tailored downstream interventions.Further analyses, especially single-cell spatial transcriptomics with intact spatial context, are warranted to further elucidate the detailed cellular and molecular mechanisms of complement-mediated immunomodulation and its potential therapeutic implications.

Patient cohort and data collection
We retrospectively reviewed all PDAC patients who underwent surgical resection of PDAC at UW-Madison between 2015 to 2020.The inclusion criteria were PDAC patients aged 18 to 80 years from both genders.Patients with incomplete clinicopathologic information or those with incomplete archived slides or blocks were excluded.From this review, 36 patients were identified as meeting the study criteria, including 23 patients who had NAT (either chemotherapy alone or in combination with radiation) before surgery, and 13 patients matched by age, gender, and stage and underwent upfront surgery without any forms of NAT.All included patients had a minimum follow-up period of 38 months.The pertinent clinicopathological information was retrospectively collected by reviewing electronic medical records under the IRB-approved protocol.These data were de-identified prior to analysis.Among the 23 patients in the NAT group, 16 patients received multiple cycles of FOLFORINOX, and 7 patients received multiple cycles of gemcitabine and Abraxane.Eight patients received concurrent radiation therapy.For all patients, only resection specimens were used, except for 4 patients, for whom the pre-NAT biopsies were also included for a paired comparison with their corresponding post-NAT resection specimen.

Sample preparation for NanoString GeoMx DSP
Serial sections (4 μm in thickness) were cut from the selected formalin-fixed, paraffin-embedded (FFPE) tissue block for the consecutive H&E staining and immunofluorescence staining for DSP (Digital Spatial Profiler) experiment.For DSP experiment, sections were baked at 60⁰C for 1.5 hours and deparaffinized in CitriSolv.After rehydration, sections were incubated in 100 °C 1xTris-EDTA/pH9 buffer for 15 minutes for antigen retrieval.After digestion and post-fixation, all sections were hybridized to UV-photocleavable barcode-conjugated RNA in situ hybridization probe set (Nanostring CTA, 1812 gene targets) overnight at 37 °C.Then, all slides were washed to remove off-target probes.Next, slides were used for immunofluorescence staining with morphology markers set, which contains 1:10 SYTO13 (Thermo Fisher Scientific, cat.no.57575), 1:20 anti-panCK-Alexa Fluor 532 (clone AE-1/AE-3; Novus Biologicals, cat.no.NBP2-33200AF532), and 1:100 anti-αSMA-Alexa Fluor 647 (clone 1A4; Novus Biologicals, cat.no.IC1420R) in blocking buffer (NanoString).For each slide, 2-5 ROIs (Regions of Interest) with the most representative morphology were selected.Each ROI was segmented into a carcinoma AOI (area of illumination, panCK+, SMA-) and a TME AOI (panCK-, SMA+) using the scanned immunofluorescence image.Photocleavage of spatially indexed barcode and sample collection were done on GeoMx DSP equipment (NanoString).Samples obtained from each AOI were collected in a 96-well plate.RNA Libraries were prepared according to the manufacturer's instruction.Then, Next Generation Sequencing (NGS) read-out was done using an Illumina NextSeq 550AR sequencer.The data was exported as digital count conversion (DCC) files which were used for analysis using R and Bioconductor packages.

Figure. 1 Intra-and intertumoral heterogeneity of NAT response carcinoma cells and the TME in
PDAC revealed by spatial transcriptomics.A. Clinical and pathologic characteristics of the 36 patients in our study cohort.For patients 14, 15, 16 and 17, pre-NAT biopsies were included in the naïve group, and the post-NAT resection specimens were included in the NAT group.For all other patients, only resected primary PDAC specimens were used.B. Immunofluorescence segmentation of each ROI into carcinoma AOI and TME AOI (pan-CK in green, α-SMA in yellow, DAPI in blue).Total mRNA libraries were prepared separately from each of these two AOIs for every ROI.C. Heatmap of gene expression profiles in all 223 carcinoma and TME AOIs.Distinct separation between carcinoma AOIs and TME AOIs is evident in the unsupervised clustering analysis, with only a limited overlap.There is no discernible separation between the naïve group and the NAT group in either the carcinoma AOI cluster or the TME AOI cluster.D. UMAP plot demonstrates the near-complete separation of carcinoma AOIs (below the dashed line) and TME AOIs (above the dashed line).95.1% (137/144) of the TME AOIs cluster above the dashed line, and 94.5% (103/109) of the carcinoma AOIs cluster below the dashed line.
Approximately 5% of the AOIs overlap at the peripheries of these two clusters.There is no noticeable separation between the NAT group and the naïve group within either cluster.E. UMAP plot of the four pairs of pre-NAT biopsy (as naïve) and post-NAT resection (as NAT-treated) in patients 14, 15, 16, 17 (∆=naïve carcinoma; ▲=NAT-treated carcinoma.○=naïve TME; •=NAT-treated TME.Datapoints from the same patient are shown in the same color).For both carcinoma AOIs and TME AOIs, datapoints from NAT group are separated from those from the naïve group.The dashed lines connect each datapoint to its nearest neighbor.Most of the datapoints are closer to datapoint from same patient than to those of the different patients, suggesting datapoints from same patient tend to aggregate together.

Figure. 2
Figure. 2 Impact of NAT on PDAC carcinoma cells.A. Volcano-plot showing up-and down-regulated genes in carcinoma cells induced by NAT.B. Chord map illustrating the relationship between individual DEGs and the enriched pathways.C. Impact of NAT on malignant cell lineage programs (left) and malignant cell states (right).NAT treatment significantly upregulated the acinar program and downregulated the cycling_G2/M state.TNF-NFκB signaling is also decreased with marginal statistical significance (Linear modeling analysis, * p<0.05; *** p<0.001, # p<0.1).D. Histologic evaluation confirmed that the NAT-treated carcinoma cells show significantly increased apoptosis (solid arrowhead), decreased mitosis (open arrowhead), greater nuclear pleomorphism, and cytoplasmic vacuolization, in line with the transcriptomics findings (scale bar: 50 μm).

Figure. 3
Figure. 3 Impact of NAT on PDAC TME. A. Volcano plot showing the significantly up-and downregulated DEGs in TME induced by NAT.B. Dot Plots summarizing the most upregulated pathways in TME following NAT.Three complement related pathways (highlighted in red) were

Figure 4 .
Figure 4. Impact of NAT on CAFs and immune cells in PDAC TME. A. Analysis of changes in immune cell composition in the tumor immune microenvironment (TIME) post-NAT.No statistically significant changes were observed except a marginal increase in the number of mast cells.B. NAT significantly increases the immunomodulatory CAF and the neurotrophic CAF, with no significant changes in other CAFs.C. snRNA-Seq analysis of data from Hwang et al. demonstrates that the number of complement C3 expressing CAFs was significantly increased following NAT.The percentages of C3 expressing CAFs and p-values are labeled on the upper right corner.

Figure. 5
Figure. 5 The prognostic relevance of NAT-induced upregulation of TME complement. A. A Heatmap displays the expression levels of the five key complement genes upregulated by NAT in the TME of the NAT group.Unsupervised clustering segregates the patients in NAT-group into the high and low TME complement subgroups.B. Kaplan-Meier survival analysis reveals that the high TME complement subgroup (red line) has a significantly improved overall survival compared to the low TME complement subgroup (blue line, p<0.05) and a marginally improved overall survival compared to the naïve group (green line, p<0.1) (* p<0.05; # p<0.1).There is no statistically significant difference in overall survival between the low TME complement subgroup and the naïve group.C. A forest plot of multivariate Cox proportional hazards analysis reveals the TME complement level is an independent favorable prognostic factor for NAT-treated patients (HR=0.21;95% CI: 0.06-0.51,p=0.023).Positive surgical resection margin (R1) is an independent poor prognostic factor for NAT-treated patients (HR=5.46;95% CI: 1.28-23.27,p=0.022).

Figure. 6
Figure.6 Association of NAT-induced upregulation of TME complement with immunomodulation and reduced immune exhaustion.A. A Box-Whisker plot shows that the high TME complement subgroup has significantly increased mast cells, CD4+ T cells and monocytes compared to the low TME

Figure. 5 Figure. 6
Figure. 5 The prognostic relevance of NAT-induced upregulation of TME complement.A.A Heatmap displays the expression levels of the five key complement genes upregulated by NAT in the TME of the NAT group.Unsupervised clustering segregates the patients in NAT-group into the high and low TME complement subgroups.B. Kaplan-Meier survival analysis reveals that the TME complement subgroup (red line) has a significantly improved overall survival compared to the low TME complement subgroup (blue line) and a marginally improved overall survival compared to the naïve group (green line) (* p<0.05; # p<0.1).There is no statistically significant difference in overall survival between the low TME complement subgroup and the naïve group.C. A forest plot of multivariate Cox proportional hazards analysis reveals the TME complement level is an independent favorable prognostic factor for NAT-treated patients (HR=0.21;95% CI: 0.06-0.51;p=0.023).Positive surgical resection margin (R1) is an independent poor prognostic factor for NAT-treated patients (HR=5.46;95% CI: 1.28-23.27,p=0.022).

1 Intra-and intertumoral heterogeneity of NAT response in PDAC carcinoma cells and TME revealed by spatial transcriptomics. A.
Clinical and pathologic characteristics of the 36 patients in our study cohort.For patients 14, 15, 16 and 17, pre-NAT biopsies were included in the naïve group, and the post-NAT resection specimens were included in the NAT group.For all other patients, only resected primary PDAC specimens were used.B. Immunofluorescence segmentation of each ROI into carcinoma AOI and TME AOI (pan-CK in green, α-SMA in yellow, DAPI in blue).Total mRNA libraries were prepared separately from each of these two AOIs for every ROI.C. Heatmap of gene expression profiles in all 223 carcinoma and TME AOIs.Distinct separation between carcinoma AOIs and TME AOIs is evident in the unsupervised clustering analysis, with only a limited overlap.There is no discernible separation between the naïve group and the NAT group in either the carcinoma AOI cluster or the TME AOI cluster.D. UMAP plot demonstrates the near-complete separation of carcinoma AOIs and TME AOIs.95.1% (137/144) of the TME AOIs cluster above the dashed line, and 94.5% (103/109) of the carcinoma AOIs cluster below the dashed line.Approximately 5% of the AOIs overlap at the peripheries of these two clusters.There is no noticeable separation between the NAT group and the naïve group within either cluster.E. UMAP plot of the four pairs of pre-NAT biopsy and post-NAT resection in patients 14, 15, 16, 17 (∆=naïve carcinoma; ▲=NAT-treated carcinoma.○=naïve TME; •=NAT-treated TME.Datapoints from same patient are shown in same color).For both carcinoma AOIs and TME AOIs, datapoints from NAT group are separated from those from the naïve group.The dotted lines connect each datapoint to its nearest neighbor.Most of the datapoints are closer to datapoint from same patient than to those of the different patients, suggesting datapoints from same patient tend to aggregate together.