Abnormal HSF1 activation is an early molecular event in pancreatic cancer tumorigenesis.
As in our previous study, activated HSF1 level was elevated in human/mice PDAC and was associated with it progression [13]. In this study, compared to normal pancreatic acinar, we found that the starting significant accumulation of HSF1 in the cytoplasmic of human pancreatic ADM structure (Fig. 1A); however, these activated HSF1 were quickly translocated into nucleus of precancerous lesions even as the in early PanINs formation and lasted until invasive PDAC stage, along with an abundant desmoplastic reaction (Fig. 1B), which means HSF1 and its activation may participate in the initiation of human pancreatic cancer. Next, by using GEPIA and Kaplan-Meier Plotter, we found that HSF1 mRNA level has limited correlation to poor prognosis (overall survival (OS) and disease free survival (DFS)) of pancreatic cancer patients (Additional file 1: Fig. S1F-H), and HSF1 target genes, such as HSPA1A/HSP90AA1, were elevated in pancreatic cancer compared to in the normal pancreas (Additional file 1: Fig. S1I and J). Besides, we also found there was no relationship between the HSF1 expression and the clinical characteristics of pancreatic cancer patients such as tumor stage (Additional file 1: Fig. S1K and Additional file 2: Table S1), and HSF1 was not an independent risk factors for OS of pancreatic cancer patients (Additional file 1: Fig. S1L and Additional file 2: Table S1) according to TCGA datasets. To sum up, all the findings suggested that the central roles of HSF1 may mainly enrichment in the initiation of human pancreatic cancer rather than its progression.
Besides, by using the well tumorigenesis model of pancreatic cancer called KC mice (which harbor a pancreas specific mutant KrasG12D oncogene), we found the same phenomenon that there was little HSF1 in normal acinar; however, the accumulation of HSF1 was founded in murine ADM cytoplasmic and in the nucleus of precancerous lesions/invasive PDAC tissues (Fig. 1C). Indeed, we next want to investigate whether the HSF1 expression was correlated with KRAS oncogene mutation. In mRNA level, according to TCGA datasets, pancreatic cancer samples which harbored mutant KRAS have a higher level of HSF1 compared to no mutation samples (Fig. 1D and Additional file 2: Table S1). However, we also found there have no significant difference between normal pancreas and pancreatic tumor tissues in mRNA level (Fig. 1E). Hence, to further illustrate above doubt, in protein level, we found that compared to pancreatic ductal progenitor cell line hTERT-HPNE (which harbor KRASWT and can reflect the properties of intermediary cells produced during ADM[21, 22]), pancreatic cancer cell lines harbor relative high expression of HSF1 (Fig 1F). Mentionable, similarly to Liang W and colleagues findings[23], as for the difference of pancreatic cancer cell lines in Fig 1F, we also detected that HSF1 expression in mutant KRAS cell lines[24] (AsPC-1/KRASG12D, MIAPaCa-2/KRASG12C, PANC-1/KRASG12D, CFPAC-1/KRASG12V) were relative higher than wild-type KRAS cell line[24] (BxPC-3/KRASWT). Interestingly, we also demonstrated that in mice pancreas, C mice acinar cells (KrasWT) have scarcely any HSF1 expression but KC mice acinar cells (KrasG12D) have little HSF1 expression in their cytoplasm (Fig. 1G). As we all know, KRAS oncogene mutation was a switch of pancreatic cancer initiation[25], hence, above findings suggested that HSF1 may be a critical molecular for pancreatic cancer tumorigenesis again.
In addition, studies have proven that both acute and chronic pancreatitis are the key risk factors for pancreatic cancer[26-28], so we treated 8-week-old KC mice (when PanINs lesions starting formation, as shown in Additional file 1: Fig. S1E) with cerulein to induce pancreatitis, and we found that HSF1 was expressed at a high level and had translocated from the cytoplasm to the nucleus during the pancreatitis (Fig. 1H and I). In conclusion, abnormal HSF1 activation is an early molecular event in pancreatic cancer initiation and may participate its tumorigenesis.
Pharmacological inhibition of HSF1 suppresses pancreatic cancer initiation.
To illuminate above hypothesis, we next treated KC mice with KRIBB11 (a well-known HSF1 inhibitor) to pharmacologically inhibit the activation of HSF1 (Additional file 1: Fig. S1C left), and we found that KRIBB11 reduced the area of pancreatic precancerous lesions (both the number and the area of PanINs at all stages) compared with that in vehicle KC mice (Fig. 2A-C). Along with this, we found that there were no low-grade pancreatic precancerous lesions (AB/PAS+ blue area) in vehicle C mice pancreases. However, vehicle KC mice had a number low-grade PanINs which indicated the initiation of pancreatic cancer; nevertheless, after treatment with KRIBB11, the AB/PAS+ blue area was obviously reduced (Fig. 2A and D), which demonstrates that HSF1 inhibition may suppress the formation of PanINs. Indeed, by using IHC staining for amylase (acinar marker) and CK19 (ductal marker), we found that KRIBB11 elevated the normal amylase+ acinar area and reduced the abnormal CK19+ ductal area in the KC mice pancreases compared to those of the control (Fig. 2A, E and F); consistently, by using western blotting assay, we found that the CK19 expression and the activation of HSF1 was inhibited by KRIBB11 (represented by the reduction of HSP70) in KC mice pancreas (Fig. 2I), but the amylase expression was upregulated in KRIBB11 treated KC mice compared with vehicle KC mice (Fig. 2I). All the findings indicated that HSF1 inhibition may suppress the initiation of pancreatic cancer. Notably, studies have proved that the growth-promoting effect of HSF1 was mainly due to its regulatory activity for proliferation and metabolism[7], and we found that KRIBB11 treatment lead to a reduced number of Ki67+ proliferating cells in PanINs compared to that in the vehicles (Fig. 2G and H). Taken together, HSF1 plays a vital role in the initiation of pancreatic cancer.
HSF1 is a critical participant in the fate change of pancreatic acinar cells.
In the abovementioned initiation assay, we preliminarily illuminated that HSF1 might be related to the initiation of pancreatic cancer through pathological analysis. Next, we used double-label IF staining to verify these findings, and the results showed that some high-grade PanINs existed in vehicle KC mice pancreases (along with a loss of acinar structures); nevertheless, these effects were partly reversed after the inhibition of HSF1 (Fig. 3A). Specifically, compared to vehicle KC mice, KRIBB11 attenuated acinar loss; and interestingly, the precancerous lesions in the KRIBB11 group were mainly composed of ADM (both amylase+ and CK19+) and low-grade PanINs rather than high-grade PanINs (Fig. 3A), which insisted that HSF1 may participated in the fate change of pancreatic acinar cells. Hence, we used AP-ADMs (an in vivo ADM formation assay, see in Additional file 1: Fig. S1C middle) and pancreatic acinar cell 3D culture (an in vitro ADM formation assay, see in Additional file 1: Fig. S1B and Additional file 1: Fig. S2A) to further investigate this phenomenon in vivo and in vitro.
For the in vivo assay, after treating with cerulein for 2 days, both KC mice and C mice developed acute pancreatitis and raised lots of ADM (data not shown) in their pancreases. After 5 days of recovery, the pathologic morphology of the C mice pancreases had basically returned to a normal acinar structure (there were no ADM or PanINs in the pancreas, only the edema of acinar cells); however, the KC mice pancreases still exhibited a large amount of ADM and even PanINs, and this phenomenon was attenuated by treatment with KRIBB11 compared to that in the vehicles (Fig. 3B and C). Similarly, IHC staining showed that HSF1 inhibition increased the normal amylase+ acinar area and reduced the abnormal CK19+ ductal area in the KC mice pancreases (Fig. 1B, D and E). Consist of these findings, by using western blotting assay, we also found that KRIBB11 reversed the acinar loss and ductal gain which caused by short-term cerulein induced acute pancreatitis tissues (Fig. 3F). Synchronously, the expression HSP70 were also downregulated by KRIBB11 which insisted that the anti-ADM formation effect of KRIBB11 may related to its role on inhibiting HSF1 (Fig. 3F).
For the in vitro assay, we plated acinar cell clusters from KC/C mice in 3D conditions (Additional file 1: Fig. S2A) and treated them with KRIBB11 or not to explore the role of HSF1 on the initiation of pancreatic cancer in vitro. Before starting, we first excluded the toxic effect of KRIBB11 on KC mice acinar cells because primary pancreatic acinar cells were vulnerable even in 3D culture condition. The results showed that lower KRIBB11 consecration (≤ 10 μM) have no obvious toxic effect on the growth and ADM formation of acinar cells (Additional file 1: Fig. S2B, we observed the existence of phenotypic normal acinar (green), acinar with a phenotype that are undergoing ADM (blue) and ductal-like sphere (in vitro ADM structures, red) in the background of KRIBB11 2 μM after 3 days’ culture). Secondly, we treated acinar cells from C mice with KRIBB11 and found that HSF1 inhibition accelerated the death/necrosis and structural collapse of acinar cells (Additional file 1: Fig. S2C), which suggested that the core role of HSF1/HSPs on the protection of pancreas in stress condition[29]. Next, we treated acinar cells from KC mice with KRIBB11 and found that HSF1 inhibition suppressed the formation of ductal-like spheres, in other words, KRIBB11 inhibited the pancreatic cancer tumorigenesis in vitro (Additional file 1: Fig. S2D-F). By using qRT-PCR assay, we found that HSF1 inhibition reduced the expression of the ductal marker CK19 and HSP70; in contrast, KRIBB11 evaluated the expression of the acinar marker amylase (Additional file 1: Fig. S2G). Besides, we also corresponding lentiviral expression system to knock-down/over-expression of HSF1 to repeat above experiments and came out the same phenomena (Additional file 1: Fig. S2H-K, though the expression of amylase has no statistical difference, there was still a trend). In conclusion, all in vivo and in vitro findings indicated that HSF1 is related to the formation of ADM, in other words, HSF1 is a critical participant in the fate change of pancreatic acinar cells.
HSF1 is a key molecule for the formation of pancreatic precancerous lesions.
In the initiation assay described above, we preliminarily illuminated that HSF1 might be related to the formation of PanINs. Next, we used CP-PanINs (Additional file 1: Fig. S1C right) to further investigate this phenomenon in vivo. Compared with vehicle KC mice, the pathological morphology of the pancreases changed greatly in KC mice with chronic pancreatitis. Specifically, 12-week-old vehicle KC mice developed a certain amount of ADM and PanINs (mainly low-grade PanINs), and normal acinar structures still dominated the pancreas (Fig. 4A and D-F); however, in chronic pancreatitis KC mice, few normal acinar structures existed, and numerous abnormal ductal-like structures (predominantly belonging to high-grade PanINs and even local PDAC tissues, Fig. 4A-D and G) were present in the mice pancreases, along with abundant desmoplastic reaction. These findings are consistent with previous findings that demonstrated that chronic pancreatitis accelerates the development of pancreatic cancer[30]. Notably, the inhibition of HSF1 partially inhibited these processes; specifically, KRIBB11 reduced the loss of acinar cells/low-grade PanINs (Fig. 4A and D-F) and inhibited the formation of high-grade PanINs (Fig. 4A-D and G). Consist of these findings, by using western blotting assay, we also found that KRIBB11 reversed the acinar loss and ductal gain which caused by long-term cerulein induced chronic pancreatitis tissues (Fig. 4H). Synchronously, the expression HSP70 were also deregulated by KRIBB11 which insisted that the anti-PanINs formation of KRIBB11 may due to its HSF1 inhibition effect (Fig. 4H). To sum up, these phenomena demonstrated that HSF1 is a key molecule for the formation of pancreatic precancerous lesions.
HSF1 is a potential downstream molecule of EGFR in pancreatic cancer tumorigenesis.
Through the above experiments, we found that HSF1 a key molecule in the initiation of pancreatic cancer. Next, to further analyze the possible molecular functions and biological pathways of HSF1 and its related proteins, we adopted DAVID to perform GO/KEGG analysis and the result showed that core relationship between HSF1/HSF1 related proteins and EGFR signaling (Fig. 5A and Additional file 2: Table S1). Besides, studies have shown that EGFR and its downstream pathways are involved in the tumorigenesis of many cancers (such as Ras-MEK-HSF1 axis regulates the development of melanoma[6]). So, we hypothesized that there may have an interaction between EGFR pathway and HSF1 in PDAC initiation. Firstly, we found that both pancreatic precancerous lesions and invasive PDAC harbored high levels of EGFR compared with normal pancreas (Fig. 5B), and this phenomenon was confirmed by TCGA datasets of pancreatic cancer (Fig. 5C). Interestingly, according to this datasets and Kaplan-Meier Plotter, we found high expression of EGFR was correlated with poor prognosis of pancreatic cancer patients (Fig 5D and Additional file 3 Fig. S3A and B); besides, we found that EGFR expression was no obvious difference among pancreatic cancer patients with different clinical characteristics (such as age, clinical stage, pathological grade and lymph node metastasis status, Additional file 3: Table S2); however, we also found EGFR level, age and lymph node metastasis status were independent risk factors for OS of pancreatic cancer patients in the TCGA datasets (Fig. 5E). Indeed, GSEA analysis based on above TCGA datasets also revealed the core role of EGFR in pancreatic cancer related pathways and other well-known tumorigenesis related pathways (Fig. 5F and Additional file 3: Table S2). In conclusion, EGFR was important for pancreatic cancer initiation and progression.
However, whether EGFR pathway and HSF1 interact in pancreatic cancer and their “boss-subordinate relationship” remains unclear. Hence, to illuminate the above ambiguous relationship, firstly, by using GEPIA websites, we found EGFR pathway related molecules such as EGFR and MEK were correlated with HSF1 and its target genes (Fig. 5G-J) such as HSPA4 (HSP70) and HSP90AA1 (HSP90); next, above co-expression relation were confirmed by the IHC staining results that EGFR, HSF1 and HSP90 increased in parallel during the disease progression (Fig. 5K); indeed, by using GSEA analysis (based on TCGA datasets), we found HSF1 binding motifs/signature (target genes gene sets, represented by RGAANNTTC_V$HSF1_01 and HSF1_01) were enriched in group with high EGFR expression (Fig. 5L and Additional file 3: Table S2), which means EGFR pathway may has a positive regulating effect on HSF1. In conclusion, all the findings indicated that there was a correlation between EGFR and HSF1 in the tumorigenesis of pancreatic cancer. Interestingly, another GSEA analysis (based on GSE98399) also revealed that the acute inhibition of EGFR plus MEK caused the suppression of HSF1 binding motifs/signature as we mentioned before in a certain degree (Additional file 1: Fig. S3C and Additional file 4: Table S3, although NOM p-value > 0.05, it still has the expected trend). Combined with the preceding results, we concluded that EGFR may be the upstream molecular of HSF1.
EGFR-HSF1 axis is momentous in pancreatic cancer initiation.
To verified above assumption, we treated KC mice with erlotinib (an inhibitor of EGFR) for 1 month and found that erlotinib suppressed the formation of pancreatic precancerous lesions (Fig. 6A-C), reduced the loss of amylase+ acinar structures, the gain of CK19+ ductal structures and the expression of HSF1 target gene HSP70 (Fig. 6A and C). Besides, EGFR inhibition suppressed the activation of HSF1 and the Ki67+ proliferation of cells in PanINs (Fig. 6A and D). Consist of these findings, western blotting assay also showed erlotinib inhibited the activation of EGFR (represented by the desecration of p-EGFR) and HSF1 (represented by the desecration of HSP70), and reduced the proliferation marker PCNA (Fig. 6E), which revealed the core role of EGFR on modulating pancreatic cancer initiation and HSF1 activation in vivo. However, does EGFR mediated pancreatic cancer initiation was HSF1 dependent remains unclear. Next, we adopted pancreatic acinar cell 3D culture to determine the effect of EGFR-HSF1 axis on the initiation of pancreatic cancer in vitro. We treated acinar cell clusters from KC mice with EGF (a natural activate legend of EGFR) and erlotinib/KRIBB11 and the results showed that EGFR activation significantly increased the number (ADM formation) and diameter (ADM growth) of ductal-like spheres, and this “sphere/ADM-promoting” ability was suppressed by erlotinib and KRIBB11 (Fig. 6F-H). Besides, by extracting acinar protein/mRNA, we found EGFR activation increased the expression of the ductal marker CK19 and HSF1 target gene HSP70, on the contrary, reduced the expression of the acinar marker amylase compared to those in the vehicles; however, these effects were partly blocked by EGFR/HSF1 inhibition (Fig. 6I-J). In conclusion, all the findings above suggested that EGFR mediated HSF1 activation play a key role in pancreatic cancer initiation in vivo and in vitro.
EGFR stimulation activated HSF1 doubly in pancreatic acinar cells.
Above findings suggested that EGFR-HSF1 axis is important for pancreatic cancer initiation both in vivo and in vitro. However, the mechanisms of HSF1 abnormal activation during pancreatic cancer initiation and the role of EGFR in this process remains unclear. Hence, we next aim to clarify above phenomena. Above in vivo studies showed that EGFR inhibition reduced the production of HSF1 (Fig. 5E); similarly, in this term, we found that KC mice acinar cells which harbor high level EGFR also have more HSF1 production compared with EGFR low expression acinar in vivo (Additional file 1: Fig. S3D). For in vitro assay, we also found that EGF/TGFα (24 hours intervention) evaluated the level of HSF1 in both KC (relative severer) and C (relative slight) mice pancreatic acinar cells (Additional file 1: Fig. S3E); besides, to further investigate this phenomenon was EGFR and downstream pathway dependent, we treated acinar cells form C/KC mice with erlotinib/selumetinib (MEK inhibitor) and the results showed that EGFR activation elevated the production of HSF1 which can be suppressed by EGFR-MAPK pathway inhibitors (Additional file 1: Fig. S3F). Mentionable, this phenomenon was severer in KC mice acinar cells than C mice acinar cells (consist with previous findings in fig. 7B) and same as our findings before (Fig.1F), KC mice (KrasG12D mediated hyper-activation of Ras-MAPK pathway) harbor relative high level HSF1 expression in acinar cells compared to C mice (KrasWT mediated normal-activation of Ras-MAPK pathway), to sum up, above results suggested that apart from the activity of EGFR, abnormal Ras-MAPK can also influence HSF1 via unknown mechanism. Interestingly, we also found that acute EGF stimulation (30 min) on KC mice pancreatic acinar cells promoted the phosphorylation (the most common activate form) and nuclear translocation of HSF1 Additional file 1: Fig. S3G), however, these effects were inhibited by selumetinib and mTOR inhibitor torkinib (Additional file 1: Fig. S3G). Similarly, we also found the same phenomenon that EGFR inhibition can suppress the phosphorylation and translocation of HSF1 in to acinar nuclear obviously in vivo (Additional file 1: Fig. S3H). In conclusion, EGFR stimulation activated HSF1 doubly (both in total and phosphorylation level) in pancreatic acinar cells.
HSF1 acts as the sensor of EGFR-Ras-MAPK hyper-activation induced PTSs during pancreatic cancer tumorigenesis.
Above findings suggested that EGFR and downstream Ras-MAPK pathway play a vital role in protein production of HSF1 in an unknown mechanism. Obviously, we first thought about whether the activation of EGFR pathway would increase protein production of HSF1 via promoting its mRNA transcription and the answer is no, detailly, 24 hours’ intervention of EGF/TGFα on KC mice pancreatic acinar cells have no influence on its HSF1 mRNA expression (Additional file 1: Fig. S3I), which suggested that EGFR and its downstream pathways influence the protein level of HSF1 in a non-transcription way. Interestingly, we suggested that the protein level of HSF1 have more clinical significance in pancreatic cancer initiation and progression previously, and as researcher mutually agreed, HSF1 and its target HSPs were key effectors that induce HSRs to maintain cellular proteostasis by perceiving HSSs and other PTSs (Additional file 1: Fig. S3J). Hence, we hypothesized that the abnormal activation EGFR and its downstream pathways may induce PTSs in pancreatic acinar cells and the accumulation/activation of HSF1 in pancreatic cancer tumorigenesis may be a passive response of the cells against to increasing PTSs.
To elucidate this assumption primarily, we chose IRE1α/PERK (two stress-sensing components of unfolded protein reaction (UPR)) as the markers to reflect the level of PTSs indirectly[31, 32]. The GSEA analysis based on a TCGA PAAD datasets was carried out to found the influence of EGFR/KRAS/MEK/ERK mRNA expression on these two molecular mediated UPR and the results showed that high expression of EGFR-Ras-MAPK pathways have a positive enriched tendency on these two UPR pathway (Additional file 1: Fig. S3K-N and Additional file 3: Table S2); besides, another GSEA analysis based on GSE98399 showed that the inhibition of EGFR plus MEK can partly have opposite effect against to above phenomenon (Additional file 1: Fig. S3O and Additional file 4: Table S3, although NOM p-value > 0.05, it still has the expected trend), which means higher activation level of EGFR-Ras-MAPK pathways may induce more serious UPR reaction (symbol of PTSs). Indeed, to further demonstrate whether the induction of IRE1α/PERK mediated UPR response or PTSs in pancreatic cancer can active HSF1, similar before, we adopted a well-known gene set from GSEA database (specifically, in REACTOME subsets and named REGULATION_OF_HSF1_MEDIATED_HEAT_SHOCK_RESPONS) to reflect the HSF1 and analyzed the enrichment of this sets in the background of different IRE1α/PERK mRNA expression according to a TCGA PAAD datasets. The results showed that high expression of PERK can positively enrich the activity of HSF1 (Additional file 1: Fig. S3P and Additional file 3: Table S2), but the phenomenon was not obvious in IRE1α (Additional file 1: Fig. S3P and Additional file 3: Table S2). Hence, above findings suggested that EGFR and its downstream Ras-MAPK pathway may activate HSF1 via induce the PTSs of pancreatic acinar cells indirectly, in other words, HSF1 may acts as the sensor of EGFR-Ras-MAPK hyper-activation induced PTSs during pancreatic cancer tumorigenesis, however, the exact mechanism has yet to be elucidated.