KLF15 expression is downregulated in pancreatic cancer tissues and predicts a relatively good prognosis

To gain a systematical expression profile of KLF15, we analyzed its expression in pancreatic cancer tissues. First, we tested the level of KLF15 in 6 pairs of clinical tissues, the results showed, the expression level of KLF15 in PDAC tissues was significantly lower than that in peri-tumor tissues (Fig.S1A). We performed immunohistochemical (IHC) staining of KLF15 in 60 surgical samples from PDAC patients those underwent radical resection in TJMUCIH. Tissue samples were divided into two groups according to KLF15 expression as follows: a low-staining group (IHC score < 4) and a high-staining group (IHC score ≥ 4) (Fig. 1A). PDAC patients with low KLF15 expression level had significantly poorer OS (P < 0.01) and RFS (P < 0.01) than those with high KLF15 protein expression, The median overall survival (OS) and recurrence-free survival (RFS) for the high-KLF15 group were significantly longer than those for the low-KLF15group (OS: 25.73 months vs. 12.367 months, P < 0.001, RFS: 12.0 months vs. 7.0 months, P < 0.001) (Fig. 1B). Univariate analysis and multivariate analysis suggested KLF15 was an independent prognostic factor in PDAC (Tables 1 and 2).

Fig. 1

KLF15 expression is down-regulated in pancreatic cancer tissues and predicts a relatively poor prognosis. (A) IHC staining of KLF15 in surgical samples from PDAC patients. Examples of low-expression and high-expression of KLF15 are shown. (B) Kaplan-Meier curves of OS and RFS in pancreatic cancer patients according to the different levels of KLF15 based on the log-rank statistic test (P < 0.001). (C) IHC staining of KLF15 in specimens via paracentesis of advanced pancreatic cancer patients. Examples of low-expression and high-expression of KLF15 are shown. (D) Kaplan-Meier curves of OS according to the different levels of KLF15 based on the log-rank statistic test (P < 0.001). (E) The relationship between the chemotherapy response and the expression of KLF15 (P < 0.001). (F-H) The expression of KLF15 in PDAC and the relationship between OS, RFS and KLF15 expression in GEPIA database (P < 0.05)

Table 1 Univariate Cox proportional hazards analysis of clinicopathological factors for overall survival and relapse free survivalTable 2 Multivariate Cox proportional hazards analysis of clinicopathological factors for overall survival and relapse free survival

To determine whether KLF15 expression was associated with clinicopathological parameters in PDAC, patients were divided into two groups according to IHC scores for KLF15 expression, the level of KLF15 was negatively correlated with Tumor size (c2 = 7.716, P = 0.007, r =-0.346), Histological grade (c2 = 4.007, P = 0.044, r =-0.261), TNM stage (c2 = 7.163, P = 0.007, r =-0.346) and lymph node metastasis (c2 = 6.524, P = 0.010, r = -0.330) in PDAC patients (Table 3). In order to investigate the role of KLF15 in advanced pancreatic cancer patients, we also performed IHC staining of KLF15 in specimens via paracentesis, the high KLF15 expression group displayed a trend of no progressive disease (NPD) (Fig. 1C-E). These results agree with the analysis of GEPIA dataset (Fig. 1F-H). These observations suggested that KLF15 may act as a tumor suppressor gene.

Table 3 Corralationship of tissue KLF15 expression with clinicopathological parametersLow expression of KLF15 is associated with the stemness phenotype in PDAC

As cancer stemness is a feature of chemoresistance and disease progression, we focus on the role of KLF15 in pancreatic cancer stemness. To investigate the association between KLF15 and stemness phenotype in PDAC, an immunohistochemical multiplex assay was performed in tissues from a retrospective cohort of 80 patients with PDAC. The number of CD133 + cells in the high-KLF15 group significantly decreased compared with those in the low-KL15 group (Fig. 2A). Furthermore, the prospective cohort was applied, in fresh PDAC tissues of 20 patients, the expression of KLF15 was negatively correlated with ESA + CD24 + CD44 + cells and CD133 + cells (Fig. 2B-C). In vitro experiments were conducted to further explored the association between the expression level of KLF15 and the stemness characteristic. Related to the basic expression of KLF15 in PDAC cells (Fig. S2A), we established overexpressed KLF15 in PANC-1 cells with low expression, down-expressed KLF15 in BxPC-3 cells with high expression, and overexpressed KLF15 and down-expressed KLF15 in SW1990 (Fig. S2B-C).

Fig. 2

Low expression of KLF15 is associated with the stemness phenotype in PDAC. (A) Multiplex fluorescent IHC staining of KLF15 expression and the accumulation of CD133 cells. Bars, 200 µm. Nonpaired Student’s t-test was used as statistical analysis; **** P < 0.0001. (B-C) The proportion of ESA + CD24 + CD44 + cells (B) and CD133 + cells (C) by flow cytometry; Spearman’s correlation analysis between KLF15 IHC score and the proportion of ESA + CD24 + CD44 + cells (B, right) and CD133 + cells (C, right); P < 0.001. (D) The proportion of ESA + CD24 + CD44 + cells in PANC-1-Vector, PANC-1-KLF15-OE, BxPC-3-Scramble and BxPC-3-KLF15-KD detecting by flow cytometry (left) and the proportion of ESA + CD24 + CD44 + cells analysis (right). (E) The proportion of CD133 + cells in PANC-1-Vector, PANC-1-KLF15-OE, BxPC-3-Scramble and BxPC-3-KLF15-KD detecting by flow cytometry (left) and the proportion of CD133 + cells analysis (right). (F-G) The proportion of ESA + CD24 + CD44 + cells and CD133 + cells in PDX-1-Vector, PDX-1-KLF15-OE, PDX-2-Scramble and PDX-2-KLF15-KD detecting by flow cytometry. (H-I) ELDA for PANC-1-Vector, PANC-1-KLF15-OE, BxPC-3-Scramble and BxPC-3-KLF15-KD using the ELDA software (http://bioinf.wehi.edu.au/software/elda). (J) Respective images of spheroid formation of different groups (left) and the spheroid formation abilities analysis (right). (K-L) The protein expression of KLF15 in CD133- and CD133 + cells by western blot. (M) In vivo limiting dilution assay of PANC-1-Vector and PANC-1-KLF15-OE. All in vitro experiments were repeated three times independently. Paired Student’s t-test was used for statistical analysis. ** P < 0.01*** P < 0.001

The percentage of ESA + CD44 + CD24 + cells significantly decreased in overexpressed KLF15 compared with that in the PDAC-vector control group (Fig. 2D). Similarly, KLF15 also suppressed the expression of CD133 in PDAC (Fig. 2E). The finding was confirmed in two PDX cells (Fig. 2F-G). Extreme limiting dilution analysis (ELDA) was used in further testing the effect of KLF15 on the stemness phenotype in PDAC cells. Correspondingly, KLF15 restrained the self-renew ability in PANC-1 and BxPC-3 cells (Fig. 2H-I). We also confirmed the conclusion in SW1990 cells (Fig. S3A-C). In vitro sphere formation assay demonstrated that KLF15 negatively regulated the cellular sphere formation capacity of PDAC (Fig. 2J). As CD133 + pancreatic cancer cells serve as pancreatic cancer stem cells (PCSCs), we sorted CD133 + cells and CD133- cells from PANC-1 and BxPC-3 cells through fluorescence-activated cell sorting (FACS). Compared with CD133- cells, the CD133 + cells had lower protein levels of KLF15 (Fig. 2K-L). The role of KLF15 in regulating the stemness properties of PDAC was further investigated via subcutaneous inoculation of cells into NOD/SCID mice. Mice injected with over-expressed KLF15 of PANC-1 cells exhibited appreciably decreased tumor incidence compared to control group (Fig. S4A-B) The limiting dilution xenografted assays indicated that KLF15 can decreased the tumorigenic capacity in vivo (Fig. 2M). Therefore, our results indicated that tumoural KLF15 suppressed the stemness in PDAC.

The expression level of KLF15 is negatively correlated with the stemness factor Nanog

We examined the expression of PDAC stemness-associated proteins, KLF15 overexpression reduced the expression levels of CD24, CD44, SOX9 and Nanog in the PDAC cells, on the hand, the suppression of KLF15 significantly increased the expression of stemness markers in PDAC cells (Fig. 3A). Surprisingly, the mRNA levels of CD24, SOX9 and Nanog had no any change after KLF15 overexpression or knockdown (Fig. 3B). In different PDAC cells, we observed Nanog, which was closely correlated to cancer stem cell characteristics, seemed to displayed the most significant change. Thus, we selected Nanog as the candidate of KLF15 in regulating pancreatic cancer stemness.

Fig. 3

The expression level of KLF15 is negatively correlated with the stemness factor Nanog. (A)Western blot on KLF15, SOX9, SOX2, Nanog, CD44 and CD24 were analyzed in PDAC cell lines. GAPDH was used as loading control. (B) Q-PCR on KLF15 and stemness markers of SOX9, Nanog and CD24 were performed in PDAC cell lines, GAPDH as the control; ns: no significant. (C) IHC staining on KLF15 and Nanog in surgical samples from PDAC patients (left), (D)Spearman’s correlation analysis between KLF15 expression and Nanog expression (right); P < 0.001. (E) The expression of Nanog in mice tissues of the group of KLF15-overexpession and the group of Vector

To verify this hypothesis, we performed IHC staining of KLF15 and Nanog in the serial sections of PDAC samples. The result showed that KLF15 expression was negatively correlated with Nanog expression (KLF15, Nanog: spearman r = − 0.451, P < 0.001;(Fig. 3C-D). We confirmed the relationship between KLF15 and Nanog in tissues of NOD/SCID mice by performing Western blot. The results showed that in the group of KLF15-overexpession, the protein level of Nanog was significantly decreased compared with the control group (Fig. 3E).

KLF15 regulates the protein level of nanog by promoting ubiquitination degradation

Additionally, the mRNA level of Nanog had no change after KLF15 overexpression or knockdown. This result is consistent with the finding that KLF15 regulating Nanog via post-transcriptional modification. Thus, we investigated whether KLF15 suppresses stemness by promoting the degradation rate of Nanog. When cells were treated with cycloheximide (CHX), which is used to block the synthesis of new proteins, KLF15 overexpression significantly promoted the degradation of Nanog in PANC-1 cell and in SW1990 cells (Fig. 4A, Fig. S5A). In contrast, knockdown of KLF15 in BxPC-3 cells prolonged the half-life of Nanog protein (Fig. 4B). Moreover, KLF15 overexpression-induced downregulation and KLF15 knockdown-induced upregulation of Nanog was reversed by the treatment with the proteasome inhibitor MG132 in PANC-1 cells and in BxPC-3 cells (Fig. 4C-D). Consistent with the above results, the level of ubiquitinated Nanog was increased by KLF15 overexpression in PANC-1 cells and decreased by KLF15 knockdown in BxPC-3 cells (Fig. 4E, Fig. S5B).

Fig. 4

KLF15 regulates protein level of Nanog by promoting ubiquitination degradation. (A-B) The degradation of the Nanog protein was measured by western blot when cells were treated by CHX. The level of Nanog was quantified after normalization to that of Tubulin using Image J. (C-D) Western blot of Nanog in PANC-1-KLF15-OE, PANC-1-Vector, BxPC-3- KLF15-KD and BxPC-3-Scramble, followed by treatment with MG132. (E) Ubiquitylated Nanog was detected by western blot using an anti-ubiquitin antibody in PANC-1 cells which was treated by MG132 for 6 h. Data were performed three times. (F) Co-IP assays showing the interaction between KLF15 and Nanog. (G) Co-IP assays showing the interaction between KLF15 and USP21 and between Nanog and USP21. (H-I) PANC-1 cells (H) and BxPC-3 cells (I)were subjected to IP with anti-Nanog Ab, followed by IB with indicated Abs. KLF15-OE decreased the interaction of Nanog and USP21, and KLF15-KD enhanced the interaction of Nanog and USP21. (J) Schematic model of the domain structure of wild-type Nanog (Nanog -WT) and Nanog mutants. (K) PANC-1 cells cotransfected with Flag-tagged Nanog mutants as indicated and KLF15 were collected for immunoprecipitation. (L-M) Vector or KLF15-overexpressed PDAC cells transduced with USP21-OE or vector plasmid were treated with cycloheximide (CHX) for indicated times. The degradation of the Nanog protein was measured by western blotting analysis (L). The level of Nanog was quantified after normalization to that of Tubulin using Image J (M)

To understand the mechanism of the regulation the regulation of Nanog protein stability of KLF15, we performed the Co-Immunoprecipitation (Co-IP) assay for KLF15 and Nanog. The results indicated that in PDAC cells, KLF15 co-precipitate with Nanog (Fig. 4F, Fig. S5C).

USP21, as a deubiquitinase, deubiquitylates the K48-type linkage of the ubiquitin chain of Nanog, stabilizing Nanog. USP21-mediated Nanog stabilization is required to maintain the pluripotential state of the ESCs [22]. Co-IP assays validated that KLF15 did not co- precipitate with USP21 whereas Nanog co-precipitate with USP21 in PDAC cells (Fig. 4G, Fig. S5D). Furthermore, the protein level and mRNA level of USP21 was not changed after KLF15 overexpression or knockdown in PDAC cells (Fig. S5E-F). When Nanog co-precipitate with USP21, ubiquitination of Nanog will be inhibited, which leading the increase of protein stability of Nanog. Consistent with the notions, when KLF15 overexpression, the co-precipitation between the Nanog and USP21 was significantly weakened (Fig. 4H). Correspondingly, when KLF15 knockdown, the co-precipitation between the Nanog and USP21 was significantly enhanced (Fig. 4I)Next, to identify the binding region that mediate the interaction between KLF15 and Nanog, we constructed a series of deletion mutations lacking residues ( Nanog -∆N, Nanog-∆H and Nanog-∆C ) (Fig. 4J) . Co-immunoprecipitation assays showed that N-domain and H-domain of Nanog were required for its interaction with KLF15 (Fig. 4K)Furthermore, rescue experiments were performed in PDAC cells when USP21 is co-overexpressed with KLF15, western blot assay showed in double overexpressed KLF15- USP21 cells, the complement of USP21 restored the expression of Nanog (Fig. S6A-B). In addition, CHX assays showed that the complement of USP21 in cells could attenuate the decrease of stability caused by KLF15 overexpression, which also indicated that KLF15 could interfere with the binding between USP21 and Nanog (Fig. 4L-M, Fig. S6C-D). Thus, KLF15 possibly directly interacts with Nanog, inhibiting interaction between Nanog with USP21, and promoting the ubiquitylation and proteasome-mediated degradation of Nanog in PDAC cells.

KLF15 inhibits stemness profile in PDAC by downregulating protein level of Nanog

To determine whether KLF15 suppresses PDAC stemness by downregulating Nanog expression, we performed response and blocking experiments in vitro and in vivo by knocking down or overexpressing Nanog.

Firstly, we overexpressed Nanog in PANC-1 cells both in PANC-1-Vector and PANC-1-KLF15-OE group, the role that KLF15-induced reduction in the proportion of ESA + CD44 + CD24 + cells was abolished by Nanog overexpression (Fig. 5A). We then knocked down Nanog in SW1990 cells both in the scramble and KLF15-KD group. No difference in the proportion of ESA + CD44 + CD24 + cells was found between the two groups (Fig. 5B). The protein expression of Nanog and KLF15 in Nanog OE cells in the presence of KLF15, as well as in double depleted KLF15- Nanog cells was shown in Fig. S7A-B. These results were also confirmed by conducting flow cytometry detecting the percentage of CD133 + cells and using extreme limiting dilution analysis (ELDA) (Fig. 5C-F).

Fig. 5

KLF15 inhibits stemness profile in PDAC by downregulating protein level of Nanog. (A-B) The proportion of ESA + CD24 + CD44 + cells in PANC-1-Vector and PANC-1-KLF15-OE with or without overexpression of Nanog (A, left), and in SW1990-Scramble and SW1990-KLF15-KD with or without down-expression of Nanog (B, left), detecting by flow cytometry and the proportion of ESA + CD24 + CD44 + cells analysis (A-B, right). (C-D) The proportion of CD133 + cells in PANC-1-Vector and PANC-1-KLF15-OE with or without overexpression of Nanog (C, left), and in SW1990-Scramble and SW1990-KLF15-KD with or without down-expression of Nanog (D, left), detecting by flow cytometry and the proportion of ESA + CD24 + CD44 + cells analysis (C-D, right). (E-F) ELDA for PANC-1-Vector, PANC-1-KLF15-OE, PANC-1-Vector-Nanog-OE, PANC-1-KLF15-OE-Nanog-OE, SW1990-Scramble SW1990-KLF15-KD SW1990-Scramble-shNanog and SW1990-KLF15-KD-shNanog using the ELDA software (http://bioinf.wehi.edu.au/software/elda). All above experiments were repeated three times independently. Paired Student’s t-test was used for statistical analysis. ** P < 0.01*** P < 0.001 ns: no significant. (G-H) In vivo limiting dilution assay was performed to determine the effects of Nanog in SW1990-Scramble and SW1990-KLF15-KD cell. Tumor incidence and CSCs probabilities were shown

To determine whether KLF15 suppresses stemness by downregulating Nanog in vivo, the limiting dilution assays were performed in pancreatic xenograft tumor models, as shown in (Fig. 5G). When KLF15 was down-expressed, CSC frequency in PDAC cells significantly increased (from 1/716277 to 1/156139). When down-expressed the Nanog in both Scramble and KLF15 knockdown groups, KLF15-knockdown-induced self-renew ability was abrogated after the silencing of the Nanog expression (Scramble-Nanog-KD vs. KLF15-KD -Nanog-KD: 1/2477943vs 1/3376726) (Fig. 5H). Collectively, our data supported that KLF15 decreased the stemness of PDAC by suppressing the Nanog expression.

Tazemetostat sensitizes PDAC cells to gemcitabine by upregulating the KLF15 expression

As chemoresistance is a key feature of PDAC patients, stemness is one of the most paramount reasons. We investigated whether KLF15 plays an active role in sensitivity to gemcitabine (GEM), which is the standard chemotherapy drug for PDAC patients after radical resection. Flow cytometry and dose-response curves were performed. The results showed that the overexpression of KLF15 sensitized PDAC cells to gemcitabine, as reflected by decreases in IC50 values, which ranged from 115.3nM to 57.38nM, and increased apoptotic rate when KLF15 was overexpressed. However, down-expressed KLF15 had the opposite effect, as reflected by increases in IC50values (Scramble IC50 = 42.26nM vs. KLF15-KD IC50 = 80.84nM) and decreased apoptotic rate (Fig. 6A-B). Thus, KLF15 is a promising therapeutic target for PDAC.

Fig. 6

Tazemetostat sensitizes PDAC cells to gemcitabine by upregulating the KLF15 expression (A) Cell viability analyses were performed in PANC-1-Vector, PANC-1-KLF15-OE, BxPC-3-Scramble and BxPC-3-KLF15-KD after Gemcitabine treatment. (B) Apoptotic cell analyses using flow cytometry were performed in PANC-1-Vector, PANC-1-KLF15-OE, BxPC-3-Scramble and BxPC-3-KLF15-KD after Gemcitabine treatment(left), and the proportion of apoptotic cells statistical analysis (right). * P < 0.05. (C-D) Western blot on the expression of H3K27me3, EZH2, KLF15 and Nanog after Tazemetostat treatment. (E-F) Q-PCR on the expression of KLF15 after treatment by different concentrations of Tazemetostat, GAPDH as the control. (G-H) ELDA for PANC-1 and SW1990 cells after Tazemetostat treatment, using the ELDA software (http://bioinf.wehi.edu.au/software/elda). All above experiments were repeated three times independently. Paired Student’s t-test was used for statistical analysis. * P < 0.05 ** P < 0.01*** P < 0.001. (I) The subcutaneous tumors formed by four groups: control, Tazemetostat, gemcitabine and the combination of Tazemetostat and gemcitabine, H&E staining and IHC staining of Ki67 in above four groups. (mean ± SD, n = 3), Scale bar (black), 200 μm. (J) The statistical analysis of Ki67 staining in the above four groups. (K-L) The volume (K) and the weight (L) of subcutaneous tumors in above four groups, (mean ± SD, n = 3), * P < 0.05 ** P < 0.01*** P < 0.001

Epigenetic regulation is a novel approach to cancer treatment. This methylated H3K27me3 chromatin is commonly associated with the silencing of genes related to tumor suppression and cell differentiation [23, 24]. The expression level of enhancer of zeste homolog 2 (EZH2) increases in H3K27me3 [25]. More and more evidences show that inhibiting EZH2 expression has potential therapeutic effects on a variety of cancer [26, 27]. Tazemetostat as the first-in-class targeted epigenetic regulator that specifically inhibits EZH2, has been used as the new FDA-approved oral treatment and received accelerated approval for patients with hematologic and solid malignancies [28]. In recent years, several reports have demonstrated tazemetostat presented good effects in esophageal cancer [29]and in biliary tract cancer [30]by inhibiting the expression of EZH2. However, the effect of tazemetostat combined with other drugs in pancreatic cancer has not been reported. When pancreatic cancer cells were treated by Tazemetastat, which is an EZH2 inhibitor, the protein level of KLF15 was increased with the concentration (Fig. 6C-D), as reflected by the decreased expression levels of EZH2 and H3K27me3. Similarly, the mRNA level of KLF15 increased with concentration (Fig. 6E-F). To analyze whether Tazemetastat can inhibit the stemness of PDAC, extreme limiting dilution analysis (ELDA) was performed in PANC-1 cells and SW1990 cells. The results showed that Tazemetastat can decrease stem cell frequencies (Fig. 6G-H).

To validate whether Tazemetastat can work synergistically with gemcitabine in suppressing PDAC growth. We performed an in vivo study for evaluating the therapeutic effects of a combination of Tazemetastat and gemcitabine. Based on the evaluation of the tumor slides, HE staining and ki-67 IHC staining were performed. The result of ki-67 IHC staining indicated that the combination of gemcitabine and Tazemetastat significantly reduced cancer cell proliferation compared with gemcitabine or Tazemetastat monotherapy (Fig. 6I-J). Consistently, treatment using a combination of Tazemetastat and gemcitabine significantly decreased tumor volume and tumor weight (Fig. 6K-L). Therefore, Tazemetastat can suppress PDAC stemness and may be used to boost the efficacy of standard chemotherapy drugs for PDAC patients.