Obesity-Induced Inflammation and Desmoplasia Promote Pancreatic Cancer Progression and Resistance to Chemotherapy

Pancreatic cancer is the fourth-leading cause of cancer-associated death worldwide, with an overall 5-year survival rate of 7% (1). The risk of pancreatic cancer is about 50% greater for individuals with obesity, particularly those with increased abdominal adiposity (2). Excess body weight also worsens the already dismal outcome of patients with pancreatic ductal adenocarcinoma (PDAC) by increasing the relative risk of cancer mortality by more than 2-fold (3–7). As a consequence of the obesity pandemic—with nearly 70% of the U.S. adult population being either overweight or obese (8)—the majority of patients with PDAC present with excess weight at diagnosis (6). Thus, understanding why obesity is associated with a worse prognosis might lead to novel treatments and enhance the outcome of current therapies.

PDAC is a highly desmoplastic cancer characterized by activated pancreatic stellate cells (PSC) and an excessive accumulation of extracellular matrix (ECM; refs. 9, 10). ECM components and PSCs directly promote the survival and migration of cancer cells (11–13). In addition, the interaction between cancer cells, stromal cells (e.g., PSCs), and the ECM produces physical forces (solid stress—pressure from solid tissue components) which compress tumor blood vessels, thus causing poor and heterogeneous tumor perfusion (14, 15). These mechanically induced changes in vascular perfusion create a formidable barrier to delivery and efficacy of chemotherapeutics, leading to poorer treatment outcomes (11, 16–19). Importantly, obesity itself is a prodesmoplastic condition. The hypoxia that results from abnormal blood vessels and decreased blood flow due to the rapidly expanding white adipose tissue (WAT) in obesity causes adipocyte dysfunction and immune cell recruitment (20–22). The latter leads to cytokine production, inflammation, and ultimately fibrosis (20–22). In particular, adipocytes in WAT abundantly express the angiotensin II type-1 receptor (AT1; refs. 23, 24), a major profibrotic pathway that becomes activated in a proinflammatory environment (23, 24). Obesity also leads to fat accumulation in the normal pancreas (steatosis), which generates a similar inflammatory process within the pancreas itself, with increased expression of cytokines, ECM remodeling, and fibrosis (7, 25, 26). Importantly, cancer lesions in obese mice and patients have an increased adipocyte content (27, 28). Of clinical relevance, the interaction of cancer cells with adipocytes—both in the form of accumulation of fat in the pancreas and at the invasive tumor front into the local adipose tissue—is associated with worse outcomes in patients with PDAC (7, 29). However, the role of adipocytes during obesity-induced PDAC progression remains unclear.

We have previously reported that obesity promotes inflammation (increased IL1β levels) and immune cell infiltration in PDACs, which associates with increased tumor growth and metastasis (30). Here, we hypothesized that the obesity-associated adipocyte accumulation in PDACs generates a proinflammatory and profibrotic microenvironment, which promotes tumor progression, and hinders the delivery and efficacy of chemotherapy. In this study, we tested this hypothesis in patients with pancreatic cancer as well as in clinically relevant orthotopic and genetically engineered mouse models of PDAC. To reduce the obesity-instigated desmoplastic reaction in PDAC, we used AT1 knockout (Agtr1a^−/−) mice and the FDA-approved AT1 blocker (ARB) losartan, which as we previously demonstrated reduces PSC activation, desmoplasia, and solid stress in PDAC (18). In addition, we examined whether the obesity-exacerbated desmoplasia in PDAC results from increased inflammation and immune cell infiltration, and we uncovered the cellular and molecular mechanisms involved.

We fed a high-fat diet (HFD) to four different strains of mice to generate diet-induced obesity (DIO; Fig. 1A). In addition, we used a genetic model of leptin deficiency (ob/ob; Fig. 1A). Consistent with previous studies including those from our laboratory (27, 30, 31), obesity promoted tumor initiation and progression across various tumor models. Using spontaneous PDAC models—KPC (Ptf1-Cre/Kras^LSL-G12D/+/Trp53^LSL-R172H/+) and iKRAS (Ptf1-Cre/ROSA26-LSL-rtTa-IRES-eGFP/TetO-Kras^{TetO-LSL-G12D}/Trp53^L/+) mice (32–35)—we found that obese animals tended to develop tumors earlier than lean mice (Supplementary Fig. S1). Furthermore, DIO and genetically induced obesity accelerated the growth of implanted tumors in two orthotopic syngeneic PDAC models—PAN02 and AK4.4. Similar to our previous observations (30), in comparison with lean mice, obese mice presented with increased tumor weight (Fig. 1B), increased metastatic dissemination to the mesenteric peritoneum in the PAN02 model (Fig. 1Ci and ii), and increased local invasion into the retroperitoneum in both models (Fig. 1Ciii). Taken together, the mouse models used confirmed the tumor-promoting effect of obesity in PDAC.

The dysfunctional hypertrophic adipocytes that accumulate in visceral adipose and pancreatic tissues in obesity lead to the development of a local desmoplastic reaction characterized by fibrosis and inflammation (20–22). In tumors, desmoplasia promotes tumor growth and impairs response to chemotherapy via reduced vessel perfusion (18). We hypothesized that obesity augments desmoplasia in the pancreatic tumor microenvironment, thus stimulating tumor progression. As expected (24), in obese mice we observed hypertrophic adipocytes and associated fibrosis in the visceral adipose tissue (Fig. 2Ai). Importantly, we found that the tumor microenvironment also contained more and larger adipocytes (Fig. 2Ai and iii). In part, this was due to tumors invading the neighboring visceral WAT (Supplementary Fig. S2A and S2B), as reported in patients with pancreatic cancer (7, 29). Furthermore, Masson’s trichrome staining revealed an abundant fibrosis in tumor areas enriched in adipocytes or alongside adjacent visceral adipose tissues (Fig. 2Ai and iv; Supplementary Figs. S2C and S3A). These data suggest that, in obesity, PDACs adopt a fibrotic adipose microenvironment as they invade the adjacent fibrotic adipose tissues. We next determined whether the abundance of fibrotic adipocyte-rich areas in tumors from obese mice led to an overall increase in tumor fibrosis. Using second harmonic generation (SHG) imaging, we found that obesity significantly increased the SHG signal, which represents fibrillar collagen, in orthotopically grown PAN02 and AK4 tumors (Fig. 2Bi and iii). Obesity also significantly increased collagen-I expression (immunofluorescence) in the lesions of orthotopic and spontaneous PDACs (Fig. 2Bii and iv). Of note, within each body-weight setting, collagen-I levels did not correlate with tumor area, indicating that tumor size alone (increased in obesity) is not responsible for the observed increase in fibrosis in obese mice (Supplementary Fig. S3B). The tumor levels of hyaluronan (HA)—an ECM molecule also associated with desmoplasia—also tended to be higher in obese mice (Supplementary Fig. S3C and S3D). We then determined whether the presence of activated PSCs was also increased in tumors in obese mice. In obese animals, immunofluorescence staining revealed a significant increase in the density of alpha-smooth muscle actin (αSMA)–positive PSCs in PAN02 (Fig. 2Ci and ii). The numbers of αSMA-positive PSCs in AK4.4 and KPC models also tended to be higher in obese mice, and immunoblotting confirmed significant increases of αSMA in PAN02 and AK4.4 tumors (Fig. 2Ci and ii; Supplementary Fig. S3E; see alsoFig. 4C). We confirmed that αSMA-expressing PSCs associate with collagen-I expression in our PDAC models (Supplementary Fig. S3F and S3G). Importantly, the percentage of PSCs associated with collagen-I expression increased by 2- to 3-fold in the obese setting (Fig. 2Ciii). Consistent with the increase in the fibrotic reaction in adipocyte-rich regions in tumors, the expression of αSMA also increased in these regions (Fig. 6Bi and ii). Taken together, we found that tumors in the obese setting are enriched in enlarged adipocytes, activated PSCs, and collagen-I.

We have reported that desmoplasia reduces the delivery and efficacy of chemotherapy in PDAC by decreasing vessel perfusion (18). Therefore, we tested if obesity-augmented desmoplasia reduced the efficacy of chemotherapy. Indeed, we found that in obese animals, the tumor vascular perfusion was reduced (Fig. 3Ai–iii). In both PAN02 and AK4.4 tumors, the percentage of perfused vessels per total tumor area significantly decreased from 2.5% to 0.75% (from ∼25% to 10% of total vessel area). Consistent with this finding, we observed an increase in hypoxia markers in tumors from obese mice (Fig. 3B; Supplementary Fig. S4). To determine if reduced perfusion impairs the delivery of chemotherapeutic agents, we measured the uptake of 5-Fluorouracil (5-FU), an approved cytotoxic agent used in the treatment of PDAC (36). Obesity significantly decreased the uptake of 5-FU in PAN02 tumors (Fig. 3C) and reduced the efficacy of 5-FU chemotherapy in both PAN02 and AK4.4 tumors (Fig. 3D). The 5-FU treatment significantly reduced the tumor weight in lean mice (Fig. 3D). In contrast, in obese mice, 5-FU did not significantly affect the growth of PAN02 and AK4.4 tumors (Fig. 3D). Collectively, these data show that, in addition to directly promoting PDAC growth and metastasis, obesity reduces vascular perfusion, drug delivery, and the efficacy of 5-FU chemotherapy.

We have shown that tumor desmoplasia results in part from the activation of PSCs through AT1 signaling (18). We confirmed the expression of AT1 in PSCs (AT1/αSMA double staining), which ranged from 70.3% (SEM = 5.7) in PSCs in PAN02 tumors to 35.3% (SEM = 12.8) in AK4.4 tumors (Supplementary Fig. S5A). We next analyzed the effects of obesity on AT1 downstream pathways. In both PAN02 and AK4.4 tumors, obesity increased the activation of p38, ERK, AKT, and their targets pS6 and 4EBP1 (Fig. 4A; refs. 18, 37–39). In addition, several target genes of the AT1 pathway, including Col1a2, Tgfβ1, and matrix metalloproteinases genes, were upregulated (Fig. 4Bi and ii; Supplementary Tables S1–S2).

To test if the inhibition of AT1 signaling could reverse desmoplasia in obese mice, we used the ARB losartan and mice deficient in AT1 (Agtr1a^−/−). In AK4.4 tumors in obese but not lean mice, the 16-day losartan treatment significantly reduced the gene and protein expression of the PSC-activation marker αSMA (Fig. 4Ci and ii; Supplementary Fig. S5B). Similarly, in obese mice, losartan significantly decreased the SHG signal for fibrillar collagen and reduced collagen-I immunofluorescence, though not significantly (Fig. 4Di–iv). Moreover, in AK4.4, losartan normalized the obesity-induced expression of desmoplasia-related markers (αSMA, decorin, SMAD2, p-p38) and ECM remodeling (MMP9; Fig. 4Ei and ii; and Supplementary Fig. S5B). These effects were absent or only modest in lean animals, consistent with the small reduction in αSMA expression in this weight setting. Similarly, in obese mice with PAN02 tumors, losartan significantly reduced the collagen-I content (Fig. 4Dv), and losartan or AT1 deletion reduced the expression of αSMA (Supplementary Fig. S3Ei and S3Eii; Supplementary Fig. S5C). Furthermore, intimately associated with AT1 signaling, the epithelial-to-mesenchymal (EMT) markers MMP9, SNAIL, and vimentin were increased in AK4.4 tumors from obese mice in both models and were subsequently normalized by losartan (Fig. 4E; Supplementary Fig. S5B). In PAN02 tumors, we confirmed that AT1 deletion normalized the obesity-increased expression of vimentin (Supplementary Fig. S3Ei and S3Eiii).

We then determined whether AT1 blockade could improve the response to chemotherapy, particularly in obese mice with desmoplastic tumors. In a previous study, we found that losartan combined with 5-FU significantly inhibited the growth of AK4.4 tumors in mice fed a regular diet, but losartan alone had no effect on tumor growth (18). Consistent with this, in the current study in lean mice, the genetic deletion of AT1 or the losartan treatment had no effect on tumor size in PAN02 and AK4.4 models. However, in obese mice with a genetic deletion of AT1 or treated with losartan, the tumor sizes were significantly smaller (Fig. 4Fi and ii). Similar to the data of Fig. 3D, 5-FU alone significantly reduced the size of AK4.4 and PAN02 tumors in lean but not in obese mice. The combination of genetic AT1 blockade plus 5-FU significantly reduced tumor size compared with either treatment alone in obese but not in lean animals (Fig. 4Fi). Similarly, in both AK4.4 and PAN02 tumors, losartan significantly enhanced the response to 5-FU in obese but not in lean animals, although it was still somewhat effective in the lean setting in the AK4.4 model (Fig. 4Fi and ii). However, because losartan alone was effective in decreasing tumor growth in obese mice (Fig. 4Fi and ii; Supplementary Fig. S5D), the tumor size in losartan- versus losartan plus 5-FU–treated groups in obese mice did not significantly differ (Fig. 4Fi and ii). Importantly, we confirmed via abdominal ultrasound that, when treatment was initiated, tumors had similar size in all groups (lean and obese, control vs. losartan; Supplementary Fig. S5D). In addition to AT1, we recently observed that PSCs also express the angiotensin II type-2 receptor (AT2), which has antifibrotic effects as opposed to the profibrotic effects of AT1 signaling (18). However, we found here that the PAN02 tumor response to chemotherapy in AT2^−/− mice was similar to that in wild-type (WT) mice, regardless of diet group (Fig. 4Fi).

Similar to our previous study (18), we analyzed whether the increase in 5-FU antitumor efficacy induced by losartan in obese animals could be at least in part due to improvements in vascular perfusion and drug delivery. In both AK4.4 and PAN02 models, we found a trend for increased tumor perfusion and increased delivery of chemotherapeutics by AT1 blockade (Supplementary Fig. S5E and S5F). The mild effects on tumor perfusion could be the consequence of a reduced mean arterial blood pressure (MABP) induced by the relatively high dose of losartan (90 mg/kg/day). In fact, losartan reduced the MABP by approximately 27% in both lean and obese mice (Supplementary Fig. S5G). Of note, AT1 blockade did not induce greater loss in body weight in tumor-bearing mice compared with WT, suggesting that based on this parameter, it does not appear to cause overall toxicity (Supplementary Fig. S5H).

In conclusion, we found that obesity is associated with increased AT1 signaling in tumors; that in obese animals, losartan alone or the genetic blockade of AT1 has direct antitumor effects; and that AT1 inhibition can also enhance the efficacy of 5-FU chemotherapy.

The fibrotic phenotype in adipose tissues and normal pancreas in obesity is largely the consequence of a persistent proinflammatory state, which is characterized by the production of cytokines by hypoxic and dysfunctional hypertrophic adipocytes and the recruitment of immune cells (21, 22, 34). In addition, we have previously described increased inflammation and immune cell recruitment in PAN02 tumors implanted in obese mice (30). Hence, we next determined here whether adipocyte-associated inflammation was responsible for the increased tumor desmoplasia and accelerated tumor growth observed in obese animals. Flow cytometric analysis revealed that obesity promotes the infiltration of CD11b⁺Gr1⁺F4/80⁻ myeloid cells in orthotopic AK4.4 and PAN02 PDAC models in obese mice (Fig. 5A; Supplementary Fig. S6A). We subsequently confirmed that the majority of infiltrated myeloid cells were Ly6G⁺ tumor-associated neutrophils (TAN; Supplementary Fig. S6B). In both PAN02 and AK4.4 tumors, TAN depletion (TAN-D)—with an anti-Ly6G antibody that decreased TAN infiltration by 90% (Supplementary Fig. S7A)—reverted the increased tumor growth in obese mice to levels in lean mice (Fig. 5Bi and ii). This effect occurred only when TAN-D was initiated on day 1 but not on day 7 of the experiment, indicating the relative importance of TANs in tumor progression at an early stage (Fig. 5Bi). In addition to a direct effect on tumor growth, we determined whether TANs could affect desmoplasia. We observed the preferential accumulation of TANs in areas with activated PSCs (Fig. 5C), suggesting a potential paracrine cross-talk between TANs and PSCs. Indeed, TAN-D in PAN02 tumors in obese animals decreased PSCs and the number of activated PSCs associated with collagen-I to levels in lean animals (Fig. 5Di and ii, and E). Consistent with this, we observed a significant reduction in AT1 expression and a trend for decreased collagen-I and MMP9 expression in PAN02 tumors (Fig. 5E). In addition, TAN-D significantly increased the vascular surface area and the number of perfused vessels in PAN02 tumors, but not significantly in AK4.4 tumors (Fig. 5F; Supplementary Fig. S7Bi and S7Bii). Our data suggest that obesity increases TAN recruitment in PDAC, which mediates the activation of PSCs and tumor growth.

The obesity-induced proinflammatory/profibrotic response and immune cell recruitment that occur in adipose tissue are mediated by cytokine/chemokine production from dysfunctional adipocytes, such as IL1β and IL6 (21, 22, 40). Hence, we determined whether these inflammatory cytokines mediated obesity-induced fibrotic processes and TAN infiltration in the tumor microenvironment. We confirmed our previous observation (30) that in PAN02, obesity significantly increased the expression of IL1β by 5-fold and induced a trend toward increased expression of IL6, TNFα, IL12, or CXCL1 (Fig. 6A). Obesity also increased the levels of IL1β in AK4.4 tumors (Supplementary Fig. S8). Not surprisingly, IL1β was abundantly expressed by adipocytes in adipocyte-rich areas where activated PSCs predominate (Fig. 6Bi, top row figure, and ii). Antibody neutralization of IL1β (MM425B; Endogen/Pierce Biotechnology; 2 mg/Kg i.p. q4d)—in obese mice implanted with PAN02 tumors—significantly decreased TAN infiltration. In addition, the decreased TAN recruitment was associated with an increase in CD8⁺ cytotoxic T-cell population (decreased in obese mice compared with lean) and a decrease in regulatory T cells (Treg; Fig. 6Ci and ii). Similar to TAN-D effects, IL1β inhibition in obese animals significantly reduced the growth of PAN02 tumors (Fig. 6D) as well as the expression levels of αSMA (Supplementary Fig. S3E). In addition, the EMT marker vimentin was also reduced by IL1β inhibition in obese mice (Supplementary Fig. S3E). In addition to adipocytes, IL1β was also expressed in about 70% of TANs themselves (Fig. 6E), and TAN-D reduced tumor IL1β levels (Fig. 6F). This suggests the presence of an autocrine mechanism that enables further TAN recruitment and potentiates inflammation and fibrosis. Furthermore, TAN-D significantly reduced the levels of CXCL1, indicating that this cytokine may also play a role in obesity-induced inflammation (Supplementary Fig. S9). In conclusion, our findings suggest IL1β is involved in obesity-induced TAN recruitment, immunosuppressive microenvironment, and PDAC progression.

We observed that αSMA-positive PSCs also abundantly (∼70% of PSCs) expressed IL1β (Fig. 6Bii and G) and that of the total cells expressing IL1β, PSCs correspond to 40% of these cells in adipocyte-poor but 70% in adipocyte-rich (∼70% increase) regions (Fig. 6Bi and ii). Because IL1β recruits TANs, which localize in close proximity to PSCs as shown above, we next examined whether targeting PSCs could also interfere with IL1β production and TAN recruitment, particularly in obese mice. Indeed, in obese mice, prevention of PSC activation by losartan inhibition of AT1 signaling decreased IL1β levels in AK4.4 and PAN02 tumors (Supplementary Fig. S10A). In addition, the genetic deletion of AT1 decreased TAN recruitment in PAN02 tumors in obese but not lean mice (Fig. 6Ci and ii). Consequently, recapitulating IL1β inhibition, AT1 blockade increased CD8⁺ T cells and reduced Tregs in tumors from obese but not lean mice (Fig. 6Fi–iii). Similar trends were observed for losartan in the AK4.4 tumor model (Supplementary Fig. S10Bi and S10Bii). The effects of AT1 blockade on the expression of IL1β and immune cell recruitment are consistent with the prevention of obesity-induced tumor growth (Fig. 4F) and indicate that AT1 signaling mediates obesity-augmented tumor progression via modulation of the immune microenvironment.

To validate the findings from mouse models of PDAC, we obtained samples of PDAC from treatment-naïve patients who presented with a body mass index (BMI) either below 25 (normal weight) or above 30 (obese). As in the mouse models, we found that tumors from obese patients presented with hypertrophic adipocytes (Fig. 7A) and more pronounced ECM deposition—as shown by increased collagen-I and HA expression (Fig. 7B).

To validate our preclinical findings, we analyzed data collected from patients admitted to Massachusetts General Hospital (MGH) with a diagnosis of PDAC who underwent surgical resection. A total of 309 patients were included in this study. The average BMI of patients was 26.25 (BMI distribution in Supplementary Fig. S11). In normal-weight patients (BMI ≤ 25), adjuvant chemotherapy was a significant predictor of overall survival on multivariate analysis (HR, 0.5; 95% confidence interval, 0.3–0.8; P < 0.005), independent of other tumor characteristics including tumor size or lymph node positivity (Fig. 7Ci; Supplementary Tables S3 and S4). In overweight and obese patients (BMI > 25), we did not observe a significant survival advantage of adjuvant chemotherapy on either univariate or multivariate analysis (Fig. 7Cii; Supplementary Tables S5–S6). In conclusion, these results support our preclinical findings and suggest less efficacy of chemotherapy in overweight/obese patients. In conclusion, our preclinical and clinical studies indicate that excess weight alters the tumor microenvironment to augment the cross-talk between CAAs, TANs, and PSCs, which ultimately leads to increased tumor progression and reduced response to chemotherapy (Fig. 7D).

We and others have shown that obesity promotes pancreatic tumor initiation, growth, and metastasis in preclinical models (27, 30, 41–44). Here, we show in established PDAC models that obesity stimulates the activity of IL1β and the cross-talk between adipocytes, TANs, and PSCs to worsen the fibroinflammatory microenvironment. The increased fibroinflammatory activity then promotes tumor progression and reduces vascular perfusion, drug delivery, and the efficacy of chemotherapy. We also uncovered that AT1 signaling can mediate obesity-induced PSC activation and increase collagen-I accumulation. The inhibition of AT1 signaling is particularly effective in preventing these effects in obese animals, reverting the obesity-accelerated tumor progression and resensitizing tumors to the chemotherapeutic agent 5-FU in obese mice.

In more than half of patients with PDAC, adipocytes infiltrate >20% of pancreatic tissue (28). Similar to previous reports (27, 28, 45), we demonstrated that hypertrophic tumor adipocytes—referred to as cancer-associated adipocytes (CAA; refs. 45, 46)—accumulate in murine and human PDAC in the obese setting. It is well known that obesity promotes the secretion of inflammatory cytokines from hypertrophic adipocytes in adipose tissues and the pancreas, which ultimately leads to local tissue fibrosis (21, 22, 25, 47, 48). However, how hypertrophic CAAs facilitate the interaction between inflammation and desmoplasia to ultimately promote PDAC progression in the obese setting was unknown. We observed here an accumulation of activated PSCs and fibrosis surrounding CAA-rich areas. Given the increased adipocyte burden, the overall PSC activation and fibrotic content in tumors were consequently increased. We also found in obese mice that CAAs express IL1β, which increases the production of IL1β in PDAC, especially in CAA-rich areas. In line with IL1β being an activator of PSCs (49), we found that IL1β inhibition reduces PSC activation in PDACs in obese animals, indicating that the elevated production of IL1β in CAA-rich regions leads to activation of PSCs. This is in agreement with previous reports showing that in the pancreas of mice fed a high-fat diet, fibrosis is also accompanied by inflammation in early-stage neoplastic lesions (43, 44), and with our previous observation that metformin reduces IL1β and desmoplasia in tumors implanted in obese mice (50). It was shown before that neutrophil infiltrates can be observed in the vicinity of tumor cells and in the desmoplastic tumor stroma, which correlates with undifferentiated tumor growth and poor prognosis (51). We found that TAN-specific depletion reduces PSC activation and the number of αSMA-positive cells associated with collagen-I. In addition, we discovered that IL1β could recruit/activate TANs. This is consistent with previous reports of increased levels of myeloperoxidase—a marker of intrapancreatic neutrophil sequestration/activation—associating with IL1β in the steatotic pancreas of obese mice (25, 52). Furthermore, our results show that TANs also secrete IL1β, which may then activate PSCs and further increase TAN recruitment. Our results explain the findings that, in patients with pancreatic cancer, the intrapancreatic fat correlates with the increased incidence of PDAC and lymphatic metastasis, and decreased survival (28, 53). In conclusion, in obese mice, the cross-talk between adipocytes, TANs, and PSCs—mediated in part by IL1β—stimulates the activation of PSCs.

We also found here that TAN-D in obese mice reduced tumor growth, which is consistent with the correlation between TAN infiltration and more aggressive types of pancreatic tumors (54). Importantly, the obesity-induced increase in TANs occurred concomitantly with decreased CD8⁺ T cells and increased Tregs, which is typical of an immunosuppressive microenvironment that promotes tumor progression (55, 56). In fact, we previously found that obesity is associated with an increased expression of the immunosuppressive molecules IL4 and IL5 in PAN02 tumors (30), which was confirmed here in the AK4.4 model (Supplementary Fig. S8B). Furthermore, we found here that IL1β inhibition recapitulates the effect of TAN-D on the immune phenotype and tumor growth, consistent with the TAN-recruiting effect of IL1β. These findings extend our previous observations that obesity is associated with inflammation, immunosuppression, and macrophage recruitment in PAN02 tumors, and that reducing IL1β production via VEGFR1 inhibition leads to a reduction in tumor progression in obese mice (30). Considering that the immune environment in obesity appears to be suppressed, and the unprecedented success of therapies that block immune checkpoint pathways (55), it would be interesting to evaluate whether these therapies will be particularly effective in obese hosts.

We found that the reduction in PSC activation after AT1 blockade, in addition to decreasing obesity-associated desmoplasia, also reduces the levels of IL1β, decreases the infiltration of TANs and Tregs, increases the number of CD8⁺ T cells, and reduces obesity-associated tumor growth. These findings suggest that the interaction of inflammatory cells and PSCs is bidirectional—targeting TANs reduces PSC activation, and, in turn, targeting PSCs reduces the recruitment of TANs, with both approaches reducing tumor growth. This is consistent with previous in vitro work demonstrating that neutrophils interact reciprocally with myofibroblasts and PSCs (57, 58). In conclusion, our work reveals a cross-talk between CAAs, TANs, and PSCs that promotes tumor progression in obese hosts, with IL1β (secreted by all these cells) playing a major role in this cooperation (Fig. 7D).

Similar to our previous results in mice fed a regular diet (18), losartan or AT1 deletion did not affect the growth of PDAC in lean mice. In contrast, in obese animals AT1 inhibition reduced PDAC growth, which was likely due to the inhibition of the protumorigenic activity of TANs and IL1β, which modulate the recruitment/activity of CD8-positive T cells and Tregs, as discussed above. The reduction of TAN recruitment after AT1 inhibition may result from a direct effect, because neutrophils have been shown to express AT1 (59, 60). Indeed, we found that approximately 40% of TANs expressed AT1 in the PAN02 model (Supplementary Fig. S12) and similarly in AK4.4 and KPC models (data not shown). Alternatively, the effects of AT1 blockade on immune cell recruitment could also be indirect. We showed that IL1β recruits TANs, and AT1 inhibition reduced the production of IL1β. This could be due to AT1 inhibition of PSCs or tumor adipocytes, as both cells express AT1 and produce IL1β. In fact, AT1 is highly expressed in normal adipocytes (23), its activity is increased in adipose tissues in obesity (23, 24), and we show here that it is expressed in CAA as well (Supplementary Fig. S13). Moreover, AT1 deletion reduced the levels of the TAN chemoattractant CXCL1 and the TAN activation molecule INFγ in tumors (Supplementary Fig. S14A), which can also explain the reduced TAN infiltration.

On the other hand, AT1 inhibition in obese mice reduced the expression of hypoxia and EMT markers in tumors (Supplementary Figs. S3E and S14B), likely due to its effects on the desmoplastic microenvironment (Supplementary Fig. S3E). These effects of AT inhibition may have also contributed to the reduction of tumor progression in obesity. In addition, obesity also increased AKT/ERK, p38, JNK, and pS6/4EBP1 signaling, which is involved not only in AT1 signaling/desmoplasia in PSCs, but also in proliferation and survival of PDAC cells (61–63). Because losartan reduced the activation of some of these signaling pathways (p38/JNK/4EBP1), its effects on tumor growth may also be explained by a direct effect on cell proliferation/survival.

In contrast to the antitumor effects of losartan, 5-FU reduced PDAC growth in lean but not obese mice, which could be due to the significantly lower intratumoral vascular perfusion and 5-FU uptake in obese than lean mice. In obese mice, AT1 deletion significantly increased the antitumor activity of 5-FU, whereas the effect of the combination of losartan and 5-FU was not superior to 5-FU alone. In fact, and in contrast to our previous study (18), the increased 5-FU activity induced by losartan was not related to a significant increase in vascular perfusion or drug uptake. Blood flow in tumors is known to be highly dependent on the MABP (64); thus, it is possible that the significant decrease in MABP induced by the relatively high dose of losartan (90 mg/kg) may have affected the tumor blood flow, thus minimizing the positive effects of losartan on vascular perfusion. Apart from the effects of losartan on TANs or the adaptive immune response, the losartan-induced decrease in SNAIL expression and phospho-p38 may also increase the cytotoxic activity of 5-FU. p38 is known to be involved in chemoresistance (65), and in mouse models of pancreatic cancer, SNAIL deletion increased the antitumor activity of gemcitabine (66).

Of note, consistent with previous work in PAN02 tumors (67), by using both genetic and diet-induced obese mouse models we determined that obesity promotes tumor growth independent of diet. This was also confirmed by the finding that FVB mice that did not gain weight despite an HFD (∼25% of all mice on an HFD) were observed to have tumor growth similar to that in lean animals (not shown).

We found that our key observations in mice were also present in patients. In particular, excess weight was associated with increased number and size of adipocytes and desmoplasia in tumors and reduced response to chemotherapy. Although the specific effect of obesity on patients with PDAC in a neoadjuvant setting is not known, we found that obesity adversely affected patients with PDAC who received adjuvant chemotherapy. Moreover, the high incidence (∼85% in our cohort; Supplementary Fig. S15) of hypertension in obese patients makes it an ideal target population for AT1 inhibitors (ACERi/ARBs), considering the blood pressure–lowering effect of these drugs.

Obesity is considered to be responsible for 14% to 20% of all cancer-related deaths in the United States (4). As a result, the obesity epidemic—which also affects the majority of patients with PDAC—highlights the importance of understanding the pathophysiology underlying the obesity–cancer connection. We had previously found that obesity associates with increased tumor inflammation, immune cell recruitment, and tumor growth. We discovered here that obesity-induced inflammation and TAN infiltration lead to a desmoplastic tumor microenvironment, which directly promotes tumor growth and impairs the response to chemotherapy. Both of these factors may explain the poor outcomes in obese patients. The finding that obesity-induced desmoplasia responds to clinically available antifibrotic therapies (e.g., ARBs) is extremely encouraging and strongly supports the implementation of clinical trials to determine the efficacy of these therapies in obese patients with PDAC in combination with the current standard of care. Because epidemiologic and molecular evidence suggests a link between obesity and other desmoplastic cancer types, the strategies established in this study may also apply to a broader patient population. In conclusion, obesity and the obesity-related markers described here (increased desmoplasia, TAN infiltration, and IL1β expression) may help explain the poorer prognosis in patients with PDAC.

C57BL/6 WT, leptin-deficient (ob/ob), AT1-knockout (KO; agtr1a^−/−) or AT2-KO (agtr2^−/−; all C57BL/6 background), and FVB mice were originally obtained from The Jackson Laboratory and bred and maintained in our defined flora animal colony. KPC (PTF1-Cre/LSL-KRAS^G12D/p53-R172H, mixed C57BL/6 and FVB background) and iKRAS (p48-Cre;R26-rtTa-IRES-EGFP;TetO-Kras^G12D mice, FVB background; refs. 32–35, 68) mice were obtained from our collaborators. To generate an obese model, mice (6 weeks old) were given either a 10% or a 60% fat diet (D12450J and D12492; Research Diets) for 10 weeks (or until tumor collection in KPC spontaneous models), as previously described (68–70). iKRAS mice were given 10% and 60% doxycycline (650 mg/kg)–containing diet for induction of PDAC development (D14071902 and D14071903; Research Diets). ob/ob mice remained on standard chow. For implanted tumor experiments, AK4.4 cells (KRAS^G12D and p53+/-) were kindly provided by Dr. Nabeel Bardeesy and were isolated from mice generating spontaneous pancreatic tumors (PTF1-Cre/LSL-KRAS^G12D/p53^lox/+; ref. 71). PAN02 cells (SMAD4-m174; ref. 72) were obtained from the ATCC. Orthotopic pancreatic tumors were generated by implanting a small piece (1 mm³) of viable tumor tissue (from a source tumor in a separate donor animal) into the pancreas of 6-to-8-week-old male lean or obese FVB (AK4.4 model) or C57BL/6 background (PAN02 model) mice. PAN02 tumor chunks and AK4.4 cells were authenticated by IDEXX laboratories (PAN02: IDEXX RADIL Case # 22366, 2013; AK4.4: IDEXX RADIL Case # 27818, 2014; please see more details in Supplementary Methods). With exception of PAN02, these orthotopically grown pancreatic cancers are characterized by a dense collagenous stroma, a hallmark of pancreatic cancer desmoplasia (73). All experimental uses of animals abide by the Public Health Service Policy on Humane Care of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee at Massachusetts General Hospital.

For all experiments unless specified below, mice bearing orthotropic PAN02 or AK4.4 pancreatic tumors were randomized into treatment groups (or controls), and tumors were collected at day 21 after implantation. For the tumor growth study with chemotherapy, mice were divided into treatment groups 7 days after implantation when tumors in lean and obese mice had similar sizes, as determined by ultrasound (Supplementary Fig. S5J), and treated with either 5-FU (30 mg/kg i.v. every 4 days) or an equal volume of saline by intravenous injection on days 7, 11, and 15 after implantation, with tumors collected at day 19. For the tumor growth study with losartan, mice were treated with losartan (90 mg/kg i.p. every day) or an equal volume of PBS intraperitoneally starting on day 5 after implantation for the duration of the study. In the combined experiment of losartan and 5-FU, the same protocol was used for each drug as described above. TAN-D by a Ly6G-specific inhibitor (BioExcell; 4 mg/Kg i.p. every 2 days) was administered starting at day 1 or day 7. IL1β inhibition (MM425B; Endogen/Pierce Biotechnology; 2 mg/kg i.p. every 2 days) was administered to PAN02-bearing animals starting at day 7. At the completion of the study, adipose tissue and tumor samples were collected, weighed, and processed for further analysis (for information on rtPCRarray, immunohistochemistry/immunofluorescence, Western blotting, ELISA, and flow cytometry, please see Supplementary Methods). When spontaneous models were used, tumors were collected when palpable. For drug delivery in tumors, drug preparation, and ultrasound and blood pressure measurements, please refer to Supplementary Methods.

Human samples of pancreatic cancer were obtained from the MGH tissue repository. Tumors selected received no prior chemotherapy or radiotherapy before the surgical specimen was collected at the time of tumor resection. BMI was obtained for the respective sample. A total of 16 samples were randomly selected from this subset of samples (8 with BMI < 25 and 8 with BMI > 30). Paraffin sections were stained for collagen-I and HA as described below. Images were acquired using confocal microscopy and quantified using Matlab. Data were analyzed anonymously. For survival studies, 309 consecutive patients with pancreatic cancer who received surgical resection who were followed up in MGH between 2006 and 2010 with information including clinicopathologic details, treatment received, and BMI at the time of diagnosis were included. Written informed consent from the patient or the next of kin was obtained for the use of pancreatic cancer samples or clinical data. All studies were conducted in accordance with the Declaration of Helsinki and were approved by an Institutional Review Board. All patients have a pathology-confirmed diagnosis of PDAC.

Statistical analyses were performed using GraphPad Prism Version 6.0f (GraphPad Software) and SPSS software (for clinical studies), version 22 (SPSS). Error bars indicate the SEM of data from replicate experiments. The significance of difference between samples within figures was confirmed using unpaired t tests, one-way ANOVA, χ² test, or two-way ANOVA, depending on the experimental setting. In clinical studies, patients were divided into BMI ≤ 25 (n = 145), or BMI > 25 (n = 164), with mean BMI being 26.25. The values of BMI represent the BMI at the time of diagnosis. Overall survival was defined as the time from date of diagnosis until date of death or last contact. Survival curves were calculated using Kaplan–Meier methods and were compared using the log-rank test. Covariates with a P value < 0.150 in univariate analysis were included into multivariate analyses. Cox proportional hazards regression analyses were used to conduct comparisons for chemotherapy efficacy. The association of adjuvant chemotherapy and overall survival was adjusted for tumor size (continuous variable), lymph vessel and perineural invasion (absent vs. present), lymph nodes (negative vs. positive), surgical margin (negative vs. positive), pathologic grading (1–2 vs. 3–4), adjuvant radiation (yes vs. no), and adjuvant chemotherapy (yes vs. no). Time to event was diagnosis of pancreatic cancer until death or last follow-up with individuals still alive at last follow-up being coded as censored for the event (death) as of that date. For comparison between prevalence of hypertension, a χ² test was performed. In all analyses, a P value of less than 0.05 was considered significant.

V.P. Chauhan has ownership interest (including patents) in XTuit Pharmaceuticals. Y. Boucher is a consultant for Xtuit Inc. R.K. Jain received consultant fees from Ophthotech, SPARC, SynDevRx, and XTuit. R.K. Jain owns equity in Enlight, Ophthotech, SynDevRx, and XTuit and serves on the Board of Directors of XTuit and the Boards of Trustees of Tekla Healthcare Investors, Tekla Life Sciences Investors, Tekla Healthcare Opportunities Fund, and Tekla World Healthcare Fund. No reagents or funding from these companies was used in these studies. R.K. Jain, J. Incio, and D. Fukumura have ownership interest in a provisional patent application filed December 18, 2015, titled “Polyacetal Polymers, Conjugates, Particles and Uses Thereof.” No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Incio, I.X. Chen, V. Desphande, Y. Boucher, D. Fukumura, R.K. Jain

Development of methodology: J. Incio, I.X. Chen, Y. Huang, P. Huang, V. Desphande, D. Fukumura

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Incio, H. Liu, P. Suboj, I.X. Chen, M.R. Ng, H.T. Nia, S. Kao, S. Babykutty, Y. Huang, K. Jung, N.N. Rahbari, J. Kahn, V. Desphande, J. Michaelson, T.P. Michelakos, D. Fukumura, R.K. Jain

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Incio, H. Liu, P. Suboj, S.M. Chin, I.X. Chen, M. Pinter, H.T. Nia, S. Kao, S. Babykutty, Y. Huang, N.N. Rahbari, X. Han, V.P. Chauhan, J.D. Martin, V. Desphande, T.P. Michelakos, Y. Boucher, D. Fukumura, R.K. Jain

Writing, review, and/or revision of the manuscript: J. Incio, H. Liu, P. Suboj, I.X. Chen, M. Pinter, M.R. Ng, J. Grahovac, S. Kao, S. Babykutty, N.N. Rahbari, V.P. Chauhan, J.D. Martin, J. Kahn, V. Desphande, T.P. Michelakos, C.R. Ferrone, R. Soares, Y. Boucher, D. Fukumura, R.K. Jain

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Incio, S.M. Chin, M. Pinter, J. Grahovac, J. Kahn, P. Huang, D. Fukumura, R.K. Jain

Study supervision: J. Incio, D. Fukumura, R.K. Jain

Other (performed the animal surgery): J. Kahn

The authors thank Sylvie Roberge for assistance with tumor implantation and blood pressure measurements, Sergey Kozin for assistance with blood pressure measurements, Penny Woo for assistance with animal husbandry, Carolyn Smith for assistance with immunohistochemical studies, Lance Munn for his contribution to illustrations, and Ana Batista and Peter Blume-Jensen for suggestions and critiques to the manuscript.

This study was supported in part by the NIH (CA80124, CA85140, CA96915, CA115767, and CA126642 to R.K. Jain and D. Fukumura), the Lustgarten Foundation (research grant to R.K. Jain), the Foundation for Science and Technology (Portugal, POPH/FSE funding program, fellowship and research grant to J. Incio), and the Austrian Science Fund (FWF), project number J3747-B28 (to M. Pinter).