Tissue Resident Macrophages in Pancreatic Ductal Adenocarcinoma Originate from Embryonic Hematopoiesis and Promote Tumor Progression

Pancreatic Ductal Adenocarcinoma Tissues are Infiltrated with Macrophages

High amounts of fibrosis and extensive immune cell infiltration characterize the tumor microenvironment of human PDAC. Among the infiltrating cells, macrophages are frequently a dominant component. This can be readily observed by comparing the expression of pan-macrophage markers CD68, CD163, and CD206 in paired human pancreatic tissue samples containing both PDAC lesions and adjacent normal tissue (Figure 1A–B). This increase in macrophage number parallels the amount of tissue fibrosis. While tumor infiltration by macrophages is well characterized, the sources of these cells are not well understood.

To explore the ontogeny of TAMs in PDAC, we analyzed p48-Cre⁺ Lox-Stop-Lox (LSL)-Kras^G12Dp53^flox/+ (KPC) genetically engineered mouse models (GEMMs), which undergo stepwise progression through stages of pancreatic intraepithelial neoplasia (PanIN) and ultimately develop PDAC (Hingorani et al., 2005). The KPC model faithfully recapitulates many pathological features of human PDAC, including progressive development of fibrosis and extensive accumulation of macrophages (Figure 1C). By flow cytometry, we identified CD45⁺CD11b⁺CD3⁻CD19⁻Ly6G⁻Ly6C^lo/−F4/80⁺MHCII⁺ macrophages in normal pancreas and PDAC tissues of KPC mice and syngeneic orthotopic tumors established from KPC-derived cell lines (KPC-1) (Figures 1D, S1). We confirmed macrophage identity in our gating strategy based on the expression of CD68, MerTK, CD64, colony-stimulating factor-1 receptor (CSF1R), and CX3C chemokine receptor-1 (CX3CR1), but not Siglec-F or Zbtb46 using Zbtb46^gfp/+ reporter mice (Figures 1E, S1C–E). Flow cytometry analyses of tissues from the KPC model demonstrated a 24-fold increase in total number and 5-fold increase in density of macrophages in end-point PDAC compared to normal pancreas (Figure 1F). Similarly, macrophage numbers expand in orthotopic PDAC tissues as the tumors progress (Figure 1G). Taken together, the marked increase of macrophages in PDAC mouse models mimics that seen in human pancreatic cancer.

Subsets of Macrophages in PDAC Are Maintained Independently of Blood Monocytes

To test the contribution of blood monocytes to macrophages in tumors, we first performed parabiosis by surgically joining the CD45.2⁺ KPC mice with congenic wild-type mice that are CD45.1⁺. Parabiotic pairs were set up at 3.5 months of age when KPC mice have developed extensive PanIN lesions, but lack frank PDAC lesions. We then analyzed chimerism after up to 6 weeks, when KPC mice had progressed to late-stage PDAC. We observed 28% chimerism of Ly6C^hi monocytes in the blood of KPC mice within the first 2 weeks, which did not increase further after 6 weeks of parabiosis (Figures 2A–2B). However, in steady-state pancreas, tissue macrophages only achieved 2.5% chimerism after 6 weeks (Figure 2B), suggesting that the majority of macrophages were maintained independently of circulating monocytes during this time period. Interestingly, the chimerism of macrophages in KPC tumors was only 8.8%, compared to 27% for Ly6C^hi monocytes in the same tissue. We observed similar chimerism at both 2 and 6 weeks in spite of significant increases in macrophage numbers during this time period (Figures 1F, 2B). These data suggest that while monocytes contribute to the expansion of macrophages during PDAC progression, they might not be the sole source of macrophages.

Macrophages in tumors exist in subsets with distinct molecular phenotypes. One approach to identity these TAM subsets is based on differential expression of major histocompatibility class (MHC) II (Movahedi et al., 2010). To determine if monocytes replenish these subsets differentially, we analyzed the tumor tissue following parabiosis. We found that the MHCII^hi subset constituted up to 80% of the donor-derived TAMs, suggesting that monocytes preferentially, but not exclusively, replenish MHCII^hi TAMs (Figure 2C).

To complement these parabiosis studies, we established bone marrow chimeras by lethally irradiating wild-type CD45.2⁺ mice and adoptively transferring bone marrow cells from CD45.1⁺ wild-type mice. Six weeks after transfer, >99% of the immune cells in the blood and bone marrow of host mice were donor-derived (Figure 2D). Despite potential elimination of radio-sensitive tissue-resident macrophages, 30% of macrophages in the pancreas were host-derived (Figure 2D). In parallel, we lethally irradiated 3.5-month-old KPC mice at the PanIN stage, adoptively transferred bone marrow cells from CD45.1⁺ wild-type mice, and analyzed chimerism 6 weeks later in fully established PDAC. Similarly, a significant portion (>15%) of TAMs in KPC tumors were host-derived (Figure 2G). These data confirmed the presence of pancreas-resident TAMs that are not rapidly replaced by blood. Consistent with the parabiosis studies, host-derived (CD45.2⁺) macrophages were preferentially, but not exclusively, MHCII^lo (Figures 2E–F, 2H).

To further assess the short-term contribution of blood monocytes to TAMs, we used fluorescently labeled latex beads to selectively trace Ly6C^hi blood monocytes in tumor-bearing mice, and analyzed fluorescent signals in the TAM compartments at 12, 24, 48, and 72 hours after labeling (Figure S2C) (Tacke et al., 2006). While fluorescent beads labeled 25% of blood monocytes within 24 hours, which is the peak time for fluorescence detection in blood (Tacke et al., 2006), bead signal was observed in a negligible amount of macrophages (<0.5%) in normal pancreas (Figure 2I) and 12% of TAMs in orthotopic PDAC tumors (Figure 2I). These data suggest that tumors up-regulate the recruitment of circulating Ly6C^hi monocytes to replenish a portion of TAMs. Consistent with BMT and parabiosis results, bead-labeled monocytes almost exclusively replenished the MHCII^hi TAM subset (Figures 2J–K), and this replenishment persisted through 72 hours after bead injection (Figure S2D).

Impairment of Circulating Monocytes Alone Does Not Impact PDAC Progression

To determine the long-term contribution of blood monocytes to macrophages in the pancreas, we evaluated pancreas and PDAC macrophages in CCR2- and Nur77-deficient mice, which have impairments in circulating Ly6C^hi and Ly6C^lo monocytes, respectively (Boring et al., 1997; Hanna et al., 2011). Despite >95% reduction in the number of circulating Ly6C^hi monocytes (Figures 3A and S3A), the frequency of macrophages in steady-state pancreas was not changed in Ccr2^−/− mice (Figure S3B) and PDAC-infiltrating macrophages were decreased by only 50% compared to Ccr2^+/− mice (Figures 3B and S3C). Consistent with the monocyte labeling experiments, CCR2 deficiency decreased MHCII^hi TAM frequency, while MHCII^lo TAMs were not affected (Figure 3C). Surprisingly, despite the reduced macrophage numbers, CCR2 deficiency did not impact tumor growth in two independent PDAC models (Figures 3D and S3D). Additionally, we observed no changes in tumor burden in KPPC GEMMs treated with CCR2 inhibitors continuously for 45 days (Figure 3K).

To address the possibility that Ly6C^lo monocytes might contribute to macrophages in PDAC tissues, we analyzed nur77^−/− mice. Similar to data from a previous report (Hanna et al., 2011), we observed a 75% decrease in circulating Ly6C^lo monocyte numbers in nur77^−/− mice compared to nur77^+/− littermates (Figures 3E and S3A). However, neither macrophage numbers nor their MHCII-subset distribution was changed in normal pancreas or PDAC tissues (Figures 3F–G). Additionally, Nur77 deficiency did not impact PDAC tumor growth in three PDAC models (Figures 3H and S4E). Together, these data suggest that circulating Ly6C^hi monocytes are important for sustaining MHCII^hi macrophages, but dispensable for tumor progression.

Tissue-Resident Macrophages Promote PDAC Progression

To determine whether tissue-resident macrophages regulate tumor growth, we treated tumor-naïve mice with CSF1 neutralizing antibodies in combination with clodronate-liposomes (CLD), followed by a 10-day chase period to allow mice to recover circulating monocyte numbers. Following the 10-day recovery, we found that circulating monocyte numbers in αCSF1-CLD-treated animals were restored to control levels (Figure 3I). By contrast, pancreas-resident macrophages were depleted as early as 12 hours after injection and remained depleted by 85–95% after 10 days of recovery (Figures 3I and S3F). This suggests this regimen allows us to test the loss of tissue-resident macrophages while leaving circulating inflammatory monocytes intact. To study how loss of resident macrophages affects tumor progression, we established orthotopic PDAC tumors 10 days after αCSF1-clodronate or IgG-PBS treatment. Loss of resident macrophages implantation resulted in a 50% reduction in TAMs in established tumors (Figures 3I and S3F), suggesting that loss of resident macrophages was not fully compensated for by monocyte recruitment. In contrast to observations in CCR2-null mice, depletion of pancreas-resident macrophages led to a significant reduction in tumor burden, as measured by both bioluminescence imaging (BLI) and tumor weight in two distinct PDAC models (Figure 3J, S3G). To exclude the possibility that the impaired tumor progression was due to deficient tumor “seeding” upon implantation, we treated two genetic PDAC models (KPC and KPPC mice) with αCSF1-CLD at the premalignant PanIN stage, and analyzed tumor burden after mice had progressed to PDAC. In both KPC and KPPC models, depletion of resident macrophages resulted in significant reduction in tumor burden (Figures 3K). By contrast, continuous treatment of KPPC-mice with CCR2 inhibitors during the same time period, did not impact tumor burden (Figures 3K). Additionally, analysis of tumor pathology in KPPC mice showed that depletion of resident macrophages dramatically reduced the development of high-grade invasive tumors and PDAC cell proliferation (Figures 3L–M). Taken together, these data suggest that pancreas-resident macrophages are more potent drivers of PDAC progression compared to their monocyte-derived counterparts.

Embryonically Derived Macrophages are Expand During Tumor Progression

To determine whether tissue-resident macrophages are derived from the adult hematopoietic system, we performed lineage tracing using Flt3-Cre⁺ LSL-YFP reporter mice (Flt3-Cre^YFP). Flt3 is upregulated at the multipotent progenitor stages of hematopoietic stem cell (HSC) differentiation. HSC-derived cells that have gone through the Flt3⁺ stage are YFP⁺, whereas macrophages derived from embryonic progenitors outside HSCs are YFP⁻ (Schulz et al., 2012). To validate this model, we analyzed circulating leukocytes and found that >95.5% of cells in the blood, including both Ly6C^hi and Ly6C^lo monocytes, were YFP⁺ (Figure 4A). As controls, we analyzed colon macrophages and brain microglia. Consistent with previous reports (Bain et al., 2014; Ginhoux et al., 2010), in adult mice, 93% of the macrophages in the colon were YFP⁺ and 98.8% of brain microglia were YFP⁻ (Figure 4B). In contrast, we observed heterogeneity of macrophage ontogeny in pancreatic tissues, with 32.4% of tissue macrophages labeled as YFP⁻. To determine if this heterogeneity is retained in aged mice, we analyzed 15-month-old Flt3-Cre^YFP mice and found that 30% of pancreas macrophages were still YFP⁻ (Figures S4A–B). These data suggest that embryonically derived pancreas-resident macrophages persist with age.

To determine whether these YFP⁻ macrophages persist during tumor progression, we established orthotopic KPC tumors in Flt3-Cre^YFP mice. Surprisingly, despite the known contribution from circulating monocytes, 35.4% of the macrophages in these KPC tumors remained YFP⁻, similar to the frequency in normal pancreas (Figure 4C–D). More strikingly, the number of YFP⁻ macrophages was elevated by >29-fold in tumors compared to normal pancreas (Figure 4E). Immunofluorescence analysis also identified clear subsets of both YFP⁺ and YFP⁻ macrophages in normal pancreas and PDAC tissues (Figure 4F). These data suggest that a significant portion of TAMs in PDAC tissues are derived independently of the HSCs, and expand rapidly during tumor progression.

The presence of large numbers of YFP⁻ macrophages suggests that many TAMs could be derived during embryonic hematopoiesis. Alternatively, these cells could have originated from adult HSCs without going through extensive Flt3⁺ stages. To distinguish between these two possibilities, we treated Flt3-Cre^YFP mice with one dose of a CSF1R antibody (αCSF1R) at 13.5 days post coitum (E13.5) to deplete macrophages derived from embryonic sources (Hoeffel et al., 2015). We found that treatment with αCSF1R on E13.5 resulted in 80% reduction in the density of YFP⁻ macrophages in the pancreas of F1 progeny at 6 weeks of age (Figure 4G). To further confirm the contribution of embryonically derived macrophages to TAMs in PDAC, we orthotopically implanted KPC cells in adult mice following αCSF1R treatment on E13.5. Embryonic αCSF1R treatment resulted in a 40–60% reduction in the number of macrophages in established tumors (Figure 4H). Additionally, the loss of embryonically derived macrophages led to delayed tumor progression in two distinct syngeneic PDAC models (Figure 4I). Taken together these data suggest that embryonically derived macrophages facilitate PDAC progression.

To further assess the specific contribution of embryonic hematopoietic progenitors to PDAC TAMs, we administered one dose of tamoxifen in Csf1r-mer-iCre-mer/LSL-tdTomato mice at E8.5, E9.5, E10.5, E11.5, or E13.5 to span yolk sac and fetal liver stages. Using this model, we observed that HSC-derived circulating monocytes were labeled at all time points later than E9.5 (Figures S5A–B). Using E8.5 or E9.5 tamoxifen pulsing, we observed labeling in 4% and 10% of macrophages, respectively, in normal pancreas retained the label as mice reached 6 weeks of age (Figures 5A–B). To confirm this, we used Cx3cr1-CreERT2/LSL-eYFP reporter mice and found after tamoxifen administration on E13.5 at fetal liver stage, the majority of Ly6C^hi monocytes were not labeled (Yona et al., 2013), but significant portion of pancreas tissue macrophages were labeled (Figures S5C). To assess if the embryonically labeled macrophages would expand during tumor progression, we established orthotopic Kras-INK (KI)-derived PDAC tumors in Csf1r-mer-iCre-mer LSL-tdTomato mice. Consistent with results in Flt3-Cre^YFP mice, tdTomato⁺ macrophages labeled with a tamoxifen pulse at E8.5 or E9.5 expanded in number by 6.8- or 13.5-fold, respectively, during PDAC tumor progression (Figures 5C). These data suggest that yolk sac-derived macrophages undergo significant numerical expansion during tumor progression.

In both CSF1R- and CX3CR1-driven lineage-tracing models, we observed higher labeling in the MHCII^lo macrophage subset (Figures S5D–E). Similarly in the Flt3-Cre reporter mice, significantly larger portions of YFP⁻ macrophages constitute the MHCII^lo subset in both normal pancreas and PDAC tissues (Figures S4C–D) suggesting that embryonically derived TAMs are preferentially but not exclusively enriched in the MHCII^lo subset. Interestingly, we found that in both HSC-derived and embryonically derived TAMs, the MHCII^lo subset experiences higher levels of hypoxia, but TAM hypoxia level was independent of origin (Figure S4E–F). These data suggest that macrophage origin might drive intrinsic differences in macrophage phenotype that can be further molded by the tumor microenvironment.

Embryonically Derived Macrophages Expand through in situ Proliferation

To determine if tissue-resident macrophages expand through local proliferation, we analyzed Ki67 expression and short-term 5-Bromo-2′-deoxyuridine (BrdU) incorporation. Analyses of normal pancreas demonstrated that <1% of macrophages incorporated BrdU following a 3-hour pulse and <3% were Ki67⁺ (Figure 6A–D). These data suggest that pancreas-resident macrophages in steady state are mostly quiescent. By contrast, >15% of TAMs in either autochthonous KPC tissues or orthotopic KPC-1 tumors were Ki67⁺, and 3.5–4% were labeled with BrdU within 3 hours (Figures 6A–D). Of note, the 3-hour pulse resulted in no detectable BrdU signal in circulating monocytes (Figure S6A), suggesting that BrdU signals in pancreatic macrophages reflect in situ proliferation. Immunofluorescence staining also identified a significant portion of Ki67⁺F4/80⁺ cells in autochthonous KPC PDAC tissues (Figure 6C), but not in normal pancreas. Interestingly, the majority of these Ki67⁺F4/80⁺ cells localized to fibrotic tumor areas (Figures 6C and S6C). Consistent with this, macrophages cultured on high-density collagen I gels had higher proliferation rates (Figure S6D), suggesting that there may be cross talks between tumor fibrosis and macrophage proliferation. Microarray analysis of TAMs from autochthonous KPC tumors demonstrated distinct changes in cell cycle regulatory genes when compared to macrophages from normal pancreas (Figure 6E). To assess if embryonically derived TAMs proliferate at higher rates than HSC-derived TAMs, we stained for Ki67 in tumor-bearing Flt3-Cre^YFP mice. Embryonically derived TAMs had a significantly higher frequency of Ki67 positivity than their HSC-derived counterparts (Figure 6F). This increased level of Ki67 in embryonic TAMs was independent of their MHCII status (Figure S6B). In addition, transcriptional profiling by Q-PCR also identified significantly reduced level of genes that negatively regulate cell cycle progression, such as Mafb and c-Maf, but higher levels of cell cycle promoting genes, such as Jun and Ets2, in the YFP⁻ TAM subset (Figure 6G). These data suggest that embryonically derived macrophages in PDAC tissues up-regulate proliferative programs, perhaps in response to fibrosis, to keep pace with tumor progression.

We next tested what signals sustain the survival of these TAM subsets in PDAC tissues. We treated orthotopic PDAC-bearing Flt3-Cre^YFP mice with neutralizing antibodies against CSF1 and CSF2, both of which have been implicated in macrophage survival in mouse models of cancer (Hoeffel et al., 2015; Zhu et al., 2014). Although CSF2 signal blockade did not change the number of TAMs, inhibition of CSF1 signaling led to a 48% reduction in the YFP⁺ and a 75% reduction in the YFP⁻ macrophages (Figure 6H). These data suggest that CSF1 is important for the survival of both TAM subsets, but embryonically derived macrophages are more sensitive.

Embryonically Derived TAMs Have a Distinct Pro-fibrotic Phenotype

We next asked whether distinct macrophage origins correlated with phenotypic differences. Towards that end, we first performed flow cytometry analyses to compare the expression of a panel of cell surface markers in TAM subsets using the Flt3-Cre^YFP mice. Both subsets expressed similar levels of macrophage identity markers, including CD64, CD115, and F4/80, whereas YFP⁻ TAMs expressed lower levels of CD11b (Figures 7A and S7A). YFP⁻ TAMs also expressed significantly lower levels of MHCI and MHCII (Figure 7A). Similar differences were also seen between the two macrophage subsets in normal pancreas (Figure S7B), which suggests possibly inherent differences in antigen presentation. By contrast, co-stimulatory molecules (CD80, CD86), T cell-activating molecules (CD40), and immune checkpoint molecules (PDL1, PDL2, PD1) were expressed at comparable levels (Figure 7A). Embryonically derived TAMs also expressed significantly higher levels of CX3CR1 and lower levels of CD11a and CD49d (Figures 7A and S7C). Interestingly, despite the lack of CXCR4 expression in either macrophage subset in the normal pancreas, CXCR4 was significantly upregulated in YFP⁻ TAMs (Figure 7A and S7B). Taken together, these data suggest that TAMs derived from different origins are phenotypically distinct.

To gain further insight into potential functional differences between embryonically derived and HSC-derived macrophages, we performed transcriptional profiling on macrophages sorted from Flt3-Cre^YFP mice (Figures 7B and S7D). Only a modest number of genes were expressed differentially between the YFP⁺ and YFP⁻ macrophages in steady-state pancreas. However, 660 genes were differentially expressed (>1.5 fold, p<0.05) between the two subsets in orthotopic KPC tumors. Using gene set enrichment analysis, we found that the genes enriched in embryonic-derived macrophages showed a strong trend toward poor survival when mapped to human PDAC data sets (Figure S7E). Analysis of gene ontology demonstrated that embryonically derived macrophages had a higher expression of molecules involved in extracellular matrix (ECM) deposition and remodeling (Figure 7C). Indeed, molecules on top of the list of genes that were expressed at higher levels in embryonically derived TAMs were mostly involved in ECM organization (Table S1). This included genes encoding for ECM molecules (collagen isoforms, nidogen, tenascin C, and elastin), ECM-producing enzymes (hyaluronan synthases 2 and 3), and ECM-remodeling molecules (lysyl oxidase), which we validated using Q-PCR (Figures 7D and S7F). Correlating with their differential expression profiles, embryonically derived TAMs could produce significantly more Collagen I and IV in ex vivo assays (Figure 7F). To correlate these ex vivo results to in vivo impact, we analyzed collagen density in PDAC tissue from mice treated on E13.5 with αCSF1R or control IgGs and compared these results to CCR2-deficient mice. We found embryonic macrophage depletion reduced collagen deposition. By contrast, CCR2-deficient mice had slightly elevated collagen levels (Figure S7G). These data suggest that macrophages of different origins have differential impacts on fibrosis.

To rule out the possibility that the identified YFP⁻ cells contained fibroblasts, we compared cancer-associated fibroblasts (CAFs) to YFP⁻ TAMs. We found that CAFs expressed platelet-derived growth factor receptor-β (PDGFRβ), but not CD45, F4/80, or CD11b (Figure S7H). Similarly, the mRNA expression levels of macrophage/myeloid identity genes (Emr1, Itgam, Csf1r, Csf2r, and Cx3cr1) were comparable in both YFP⁺ and YFP⁻ TAMs but 10- to 1000-fold higher in TAMs than CAFs (Figure S7I). These data confirm that the YFP⁻ cells were not fibroblasts but were macrophages with a pro-fibrotic phenotype. To evaluate how the pro-fibrotic gene signature in embryonically derived TAMs might differ from that in CAFs, we compared selected genes by Q-PCR. As expected, CAFs were the dominant producers of several isoforms of collagens, Elastin, and Sparc while YFP⁻ TAMs expressed more Col4a4, Col10a1, Col17a1, Col18a1, and Has3 (Figure S7J). Taken together, these data suggest that embryonically derived TAMs may be involved in “fine-tuning” fibrotic responses in PDAC tumors.

In contrast to pro-fibrotic genes, the levels of mRNA involved in class I and class II antigen presentation (Erap1, Psme1, and Ciita) were higher in HSC-derived TAMs (Figure 7E). To test the phagocytosis capacity in TAM subsets in vivo, we orthotopically implanted mCherry⁺ KPC-1 cells and determined mCherry positivity in TAMs. Although both TAM subsets demonstrated potent capacity to uptake tumor antigen, the amount of antigen uptake was >2-fold higher in HSC-derived TAMs (Figure 7G). We next tested the ability of each TAM subset to present antigen (ovalbumin) to OT1⁺ CD8⁺ T cells and found that HSC-derived (YFP+) TAMs were far more potent at antigen presentation (Figure 7H). In addition, YFP⁺ TAMs expressed significantly higher levels of Il12a, Il4, Ccl17, and Ifnβ (Figure 7E). Taken together, these data suggest that TAMs derived from HSCs may play more potent roles in regulating adaptive immunity. This is consistent with previous reports showing that monocyte-derived TAMs regulate immunosuppression in PDAC models (Beatty et al., 2015; Sanford et al., 2013).

Subsets of TAMs in Human PDAC Tissue Resemble Murine Embryonically Derived TAMs

To evaluate if the characterization of embryonically derived TAMs in murine PDAC models is relevant for human cancer, we took advantage of the observation that CXCR4 was almost exclusively upregulated in embryonically derived TAMs (Figure 7A), and evaluated human PDAC tissues for CXCR4⁺ TAMs. We found that 10–40% of TAMs expressed high levels of CXCR4 (Figure 7I), and that these CXCR4⁺ TAMs expressed lower levels of HLA-DR in eight out of nine patient tissues (Figure 7J). These results were consistent with our observation that embryonically derived TAMs express lower levels of MHCII. To determine if this subset of human TAMs shared similar pro-fibrotic gene expression profile identified in mice, we isolated mRNA from CXCR4⁺ and negative TAMs from human PDAC tissues. Consistent with murine data, we found that CXCR4⁺ TAMs expressed significantly higher levels of collagens and ECM-remodeling genes (Figure 7K). Collectively, these data suggest that CXCR4⁺ TAMs in human PDAC resemble murine embryonically derived TAMs and may have the ability to fine-tune PDAC fibrosis.

Murine PDAC Models

KPC mice (p48-CRE/LSL-Kras^G12D/p53^flox/+) used in these studies have been previously described (Hingorani et al., 2005) and were backcrossed to C57BL/6 background and screened for C57BL/6 identity using congenic markers. KPC-1 cell line was derived from PDAC tissues of 2.2-month-old p48-CRE⁺/LSL-Kras^G12D/p53^flox/flox (KPPC); the KPC-2 cell line was derived from tumors of 6-month-old p48-CRE⁺/LSL-Kras^G12D/p53^flox/+ (KPC) mice. Cells were grown on collagen-coated tissue culture flasks for <12 passages, and were tested for cytokeratin-19, smooth muscle actin, vimentin, and CD45 to verify their carcinoma identity and purity. To establish orthotopic KPC models, either 50,000 or 200,000 KPC-1 or KPC-2 cells in 50 μL of Cultrex (Trevigen) were injected into the pancreas of 6–12-week-old C57BL/6 mice according to published protocol (Kim et al., 2009). For mCherry uptake analyses or bioluminescence imaging (BLI), KPC-1 or KPC-2 cells were infected with mCherry or click beetle red (CBR)-GFP vector respectively. mCherry^hi or GFP^hi cells were selected by FACS prior to orthotopic tumor establishment.

Other Mouse Models

The following mouse strains were purchased from Jackson Laboratories: CCR2^−/−, Nur77^−/−, Rosa26- LSL-eYFP, and Cx3cr1-CreERT2 (all on the C57BL/6 background). Csf1r-Mer-iCre-Mer mice were purchased from Jackson Laboratories, and were crossed to Rosa26-LSL-tdTomato mice, which were a kind gift from Dr. Gregory D. Longmore; both strains are on the FVB background. Flt3-Cre^tg mice were a kind gift from Dr. Thomas Boehm and were crossed to Rosa26-LSL-eYFP or –tdTomato mice, all under the C57BL/6 background. Mice were maintained in the Laboratory for Animal Care barrier facility at Washington University School of Medicine. Washington University School of Medicine Institutional Animal Studies Committee approved all animal studies.

Human PDAC Tissues

Human PDAC tissues were obtained from surgically resected specimens from patients diagnosed in the Department of Pathology at Washington University (St. Louis, MO). Patients underwent pancreaticoduodenectomy followed by adjuvant chemotherapy. Patients did not receive neoadjuvant therapy. For histological analyses (Figure 1), tissues were embedded in paraffin blocks and processed into 6 μm-thick sections for immunofluorescence staining. For gene expression profiling in PDAC patients (Figure 7), tissues were freshly obtained from 9 patients at pancreaticoduodenectomy, and processed immediately for flow cytometry analysis and/or fluorescence-activated cell sorting. Four out of the nine patients have received neoadjuvant therapies, including FOLFIRINOX or ABRAXANE/gemcitabine prior to being scheduled for resection. All patients were off therapy for 21 days prior to surgery. All tissues were collected under informed consent from patients. Washington University Ethics committee approved the study under IRB protocol #201108117.

Parabiosis

Parabiotic pairs were generated according to established protocols (Peng et al., 2013) from age- (3.5-month-old) and weight-matched female CD45.2⁺ (KPC or wild-type C57BL/6) and CD45.1⁺ (C57BL/6) mice. Mice were injected with Buprenex^® subcutaneously after surgery. Sulfatrim was continuously added in drinking water for 10 days post-surgery to minimize infections at surgical wounds. Mice were separated and perfused with phosphate buffered saline (PBS) containing 0.2% heparin. Single cell suspensions from tissues were stained with antibodies for flow cytometry analyses 2 or 6 weeks after the establishment of parabiosis.

Bone Marrow Transplantation

Three and a half-month-old C57BL/6 mice or KPC mice were exposed to γ-irradiation dosed at 1100 rads. Animals were subsequently injected with 2.5 × 10⁶ bone marrow cells from CD45.1⁺ C57BL/6 mice through n). Leukocyte chimerism was analyzed 6 weeks later.

Macrophage Depletion

To deplete tissue resident macrophages, 8–16-week old C57BL/6 mice were treated with 3 doses of CSF1 neutralizing antibody (clone 5A1, BioXCell) (1 mg, 0.5 mg, 0.5 mg on Days −18, −14, and −11, Figure 3) and 3 doses of clodronate-containing liposome (200 uL/each on Days −17, −13, and −10). Control mice were treated with same doses/volume of IgG (clone HRPN, BioXCell) and liposome (or phosphate buffered saline as indicated). On Day 0, Mice were implanted orthotopically with 200,000 CBR⁺ KPC-2 cells or 50,000 CBR⁺ KPC-1 cells, and subjected to BLI on Days 3 and 7.

Similarly, KPC and KPPC mice were treated with 2 doses of αCSF1 (0.5 mg each, Day 1 and Day 5) and 2 doses of clodronate-loaded liposome (100 uL each, Day 3 and Day 7) starting at 2.5-month and 1-month of age, respectively. Tumor burden was analyzed when mice reached 4.5 months for KPC mice or 2.0 months for KPPC mice.

To deplete embryonically derived macrophages, C57BL/6 or Flt3-Cre^YFP mice were intraperitoneally injected with 3.0 mgs of CSF1R depleting antibody (AFS98 clone, BioXCell) on 13.5 dpc. Surviving mice were implanted with 50,000 CBR⁺ KPC-1 at 6 weeks of age. Mice were sacrificed 12 days after tumor establishment for flow cytometry and tumor burden analyses.

Lineage Tracing of Embryonically Derived Macrophages

Timed breeding was set up by crossing Csf1r-Mer-iCre-Mer mice with Rosa26-LSL-tdTomato mice (both on the FVB background). Embryonic timeline was assessed based on vaginal plug observation: 12pm on the day of plug formation was estimated to be 0.5 day post coitum (dpc). Pregnant mice were treated with tamoxifen (Sigma-Aldrich) (75 μg/g) combined with progesterone (Sigma-Aldrich) (37.5 μg/g) at one of the following time points: 8.5, 9.5, 10.5, or 13.5 dpc. To test the contribution of embryonically derived CSF1R⁺ cells to monocytes or macrophages at homeostatic conditions, tumor-free F1 mice were sacrificed for fate mapping analyses as they reached 6 weeks of age. To test the contribution of embryonically derived CSF1R⁺ cells to TAMs in PDAC tumors, 50,000 syngeneic Kras-INK cells were orthotopically implanted into F1 mice as they reached 6 weeks. Kras-INK cells were derived from p48-Cre/Lox-Stop-Lox (LSL)-Kras^G12D/Ink4a-Arf^lox/lox mice on the FVB background as previously described. Mice were sacrificed approximately 12 days after tumor implantation for the analyses of tdTomato signal in different leukocyte populations.

To trace CX3CR1+ cells, Cx3cr1-CreERT2 mice were crossed with Rosa26-LSL-eFYP mice (both on the C57BL/6 background). Pregnant mice were treated with tamoxifen (75 μg/g) and progesterone (37.5 μg/g) on 13.5 dpc. F1 neonates were fostered by lactating female mice and then aged to 6 weeks for fate mapping analyses.

Labeling of Blood Ly6C^hi Monocytes

To selectively label Ly6C^hi monocytes, 250 uL of liposomes containing clodronate were injected intraveneously (i.v.), followed by i.v. injection of 250 uL of FITC-conjugated plain microspheres 16–18 hours later (1.0 um, 2.5% solids [wt/vol]; Polysciences, diluted 1:4 in PBS). Tissues were processed for flow cytometry analyses at indicated time points after bead injection.

Mouse Tissue Isolation and Flow Cytometry

Mice were euthanized by intracardiac perfusion using 20 mL of PBS-heparin under isoflurane anesthesia. Tumor tissues and colon were manually minced and digested in 25 mL of Dulbecco’s Modified Eagle Medium (DMEM) (Thermo Fisher) containing 2 mg/mL of collagenase A (Roche) and 1X DNase I (Sigma) for 30 minutes at 37°C with constant stirring. Normal pancreas was digested in the same buffer for 15 minutes at 37°C. Digestion was quenched in 5 mL of fetal bovine serum (FBS) (Atlanta Biologicals) and filtered through 40 μm Nylon mesh, pelleted through centrifugation (2000 rpm for 5 min at 4°C), and resuspended in staining buffer (PBS containing 1% BSA). Brain and spleen were minced and triturated through 40 μm filters.

Single cell suspensions were blocked with rat anti-mouse CD16/CD32 antibodies (eBioscience) (1/200) for 10 minutes, pelleted by centrifugation, and subsequently labeled with 100 uL of fluorophore-conjugated anti-mouse antibodies at recommended dilutions for 20 minutes on ice, and washed with staining buffer. For proliferation assays, mice were injected with 5-bromo-2′-deoxyuridine (BrdU, 1 mg) i.p. 3 hours prior to sacrifice. BD Bioscience Cytofix/cytoperm kit was used to stain for BrdU. eBioscience transcription factor staining buffer set was used to stain for Ki67. Data were acquired on LSR-II (BD Biosciences), and analyzed using FlowJo software (Tree Star). To quantify proliferating/cycling cells in Flt3-Cre^YFP mice, tumors were digested as described above, stained with fluorophore-conjugated antibodies for 20 minutes, fixed in 4% formaldehyde for 10 minutes on ice, permeabilized in ice-cold 70% ethanol for 3 hours, stained with Ki67 antibody diluted in staining buffer for 20 minutes on ice, pelleted and resuspended in staining buffer, and immediately processed for data acquisition on LSR-II. To assess hypoxia, Flt3-Cre^YFP mice were intraperitoneally injected with pimonidazole hydroxychloride (60 mg/kg) 1 hour prior to sacrifice (Hypoxyprobe). Single cell suspensions were stained with antibodies against pimonidazole adducts for 20 minutes at 4°C without fixation or permeabilization prior to data acquisition.

To quantitate blood monocytes, 200 uL of blood was obtained by intracardiac puncture prior to perfusion, incubated in red blood cell lysis buffer (BioLegend) for 15 minutes on ice, and stained with fluorophore-conjugated antibodies for 20 minutes on ice. Stained cells were counted using LSR-II. For blood analysis that does not require numeration, blood was drawn via tail vein bleeding, followed by RBC lysis, antibody staining, and data acquisition.

Fluorescence-Activated Cell Sorting (FACS)

Normal or tumor-bearing mice (Flt3-Cre^YFP, KPC, or C57BL/6 control) were perfused with 20 mL of heparin-containing PBS. Tissues were digested in 25 mL of DMEM containing 2 mg/mL of collagenase A (Roche) and 1X DNase I (Sigma) for 30 minutes at 37°C with constant stirring. CD45⁺ cells were enriched from single cell suspensions through MACS^® magnetic selection using anti-mouse CD45 microbeads (Miltenyi). Elutes were blocked with rat anti-mouse CD16/CD32 antibodies (eBioscience) (1/200) for 10 minutes, pelleted by centrifugation, and labeled with fluorophore-conjugated anti-mouse antibodies at recommended dilutions for 20 minutes on ice. Cells were filtered through 40 μm Nylon mesh, and immediately sorted using Aria-II (BD Biosciences). For microarray analyses and gene expression analyses, cells were sorted directly into RNA lysis buffer (Omega Biotek); RNA was isolated using the EZNA kit (Omega Biotek) according to instructions by the manufacturer. For ex vivo assays, cells were sorted into DMEM containing 20% FBS before being pelleted and resuspended for subsequent treatments.

Ex vivo Antigen Presentation and Collagen Deposition Assays

For ex-vivo analysis of TAM subsets, macrophages were sorted from Flt3-Cre^YFP mice as described above. To evaluate their potential for antigen presentation, sorted cells were added to 96 well round bottom plates at 40,000 cells per well and allowed to adhere for 2 hours in the presence of 10 ug/ml CSF1 (Peprotech). Cells were then washed with media and incubated with SIINFEKL peptide (Thermo, 2.5 ug/ml). During this time CD8+ OT1s were isolated from the spleen of tumor free mice using CD8⁺ magnetic bead selection (Miltenyi) and these T cells were labeled with Carboxyfluorescein succinimidyl ester (CFSE) according to instructions from the manufacturer (Invitrogen). Macrophages were then washed twice with media and 40,000 CD8+ OT1’s were added. Co-culture was done in Roswell Park Memorial Institute medium (RPMI)-1640 (Gibco) supplemented 5% FBS, 1x beta mercaptoethanol (BME) and 5 ug/ml CSF1. Cells were analyzed for proliferating T cells after 36 hours by flow cytometry analysis of CD8a, CD44, CD62L and CSFE-dilution.

For ex vivo analysis of collagen production, TAMs were isolated and cultured with 10ug/ml CSF1 on glass chamber slides (Falcon). 20,000 TAMs were used per well. After 24 hours cells were fixed in PFA and analyzed for Collagen I or IV by standard immunofluorescent staining, as described above. Importantly cells were not permeabilized with detergent. Collagen content was accessed by fluorescent intensity per 20x field divided by cell number (DAPI⁺ nuclei).

Sirius Red Staining and Immunofluorescence Staining

Tissues were isolated from perfused mice, fixed in 10% formalin overnight, incubated in graded ethanol, embedded in paraffin, and cut into 6 μm-thick sections. To analyze tumor fibrosis, tissue sections were deparaffinized in xylene, rehydrated in graded ethanol, and processed for Picro-Sirius Red staining according to instructions by the manufacturer (Sigma Aldrich). To visualize macrophages in mouse KPC and human PDAC tissues, paraffin sections were steam heated for antigen retrieval in citrate-based buffer (Biogenex), treated with 1% hydrogen peroxide (Invitrogen) for 20 minutes to quench endogenous peroxidase, incubated with blocking buffer (PBS containing 5% goat serum and 2.5% BSA) for 1 hour at room temperature (RT), blocked for biotin/avidin according to instructions by the manufacturer (Vector Labs), and stained with primary antibodies at recommended dilutions at 4°C overnight. Opal 4-color IHC kit was then used according to instructions by the manufacturer (PerkinElmer). When two antibodies were used for co-staining, FITC and Cy5 were used to minimize emission spectral overlap. To visualize YFP-positive macrophages in Flt3-Cre^YFP mice, orthotopic PDAC tissues were fixed in 4% formaldehyde at 4°C overnight, incubated in PBS containing 30% sucrose and 0.1% sodium azide overnight at 4°C, and embedded in OCT compound on dry ice. Frozen sections (6 μm-thick) were air-dried, treated with blocking buffer at RT for 1 hour, and stained with CD68 (1/200) (Abcam) overnight, followed by 1 hour incubation with goat anti-rabbit-Alexa594 (1/400) (Molecular Probes) at RT. All sections were washed in PBS containing 0.05% Tween-20 between incubation steps, and mounted in DAPI-containing media (Vector Labs) for imaging on Nikon 80i microscope.

PDAC Patient Outcome Analysis

We used top 110 genes whose expression was >10-fold higher in the mouse Flt3-Cre^YFP-negative TAM subset than the Flt3-Cre^YFP-positive subset. Mouse gene entrez IDs were mapped to human ortholog based on ENSEMBLE 87 annotation using Bioconductor package “biomaRt”. Genes were queried in the cbioportal TCGA pancreatic RNA-seq data through the R package “cdgsr” on 178 primary PDAC tumors. For association analysis with survival outcomes, we considered the gene signature using the averaged gene expression across all mapped genes (after centering each by their median). The gene signature was dichotomized by associated median to divide patients into the two groups of over- and under-expression. The Kaplan-Meier method was used to estimate empirical survival probabilities and log rank test to compare survival difference between patient groups. Hazard ratio from Cox proportional hazard model was reported with 95% confidence interval.

Human PDAC TAM Profiling

For gene expression profiling in PDAC patients, tissues were freshly obtained from 9 patients at pancreaticoduodenectomy, digested into single cell suspension using the protocol described above, stained with indicated antibodies, and processed for cell isolation using Aria-II (BD Biosciences). Cells were directly sorted into RNA lysis buffer (Omega Biotek). RNA was isolated using the EZNA kit (Omega Biotek) according to instructions by the manufacturer. cDNA was synthesized using the qScript Supermix (Quanta). cDNA targets were pre-amplified for 12 cycles, followed by QPCR analyses (Taqman®, Thermo Fisher).

Quantitation and Statistical Analysis

Statistical analysis was performed using Unpaired Student’s t-test, Mann-Whitney U test, Wilcoxon matched-pairs signed rank test, or ANOVA analysis as appropriate for the data set. Data in bar graphs are displayed as means ± SEM. Statistical significance is displayed as * p<0.05.