Pyruvate Kinase M2 and Lactate Dehydrogenase A Are Overexpressed in Pancreatic Cancer and Correlate with Poor Outcome

Pancreatic cancer has a 5-year survival rate of less than 4%. Despite advances in diagnostic technology, pancreatic cancer continues to be diagnosed at a late and incurable stage. Accurate biomarkers for early diagnosis and to predict treatment response are urgently needed. Since alteration of glucose metabolism is one of the hallmarks of cancer cells, we proposed that pyruvate kinase type M2 (M2PK) and lactate dehydrogenase A (LDHA) enzymes could represent novel diagnostic markers and potential therapeutic targets in pancreatic cancer. In 266 tissue sections from normal pancreas, pancreatic cystic neoplasms, pancreatic intraepithelial neoplasia (PanIN) and cancer, we evaluated the expression of PKM2, LDHA, Ki-67 and CD8+ by immunohistochemistry and correlated these markers with clinicopathological characteristics and patient survival. PKM2 and LDHA expression was also assessed by Western blot in 10 human pancreatic cancer cell lines. PKM2 expression increased progressively from cyst through PanIN to cancer, whereas LDHA was overexpressed throughout the carcinogenic process. All but one cell line showed high expression of both proteins. Patients with strong PKM2 and LDHA expression had significantly worse survival than those with weak PKM2 and/or LDHA expression (7.0 months vs. 27.9 months, respectively, p = 0.003, log rank test). The expression of both PKM2 and LDHA correlated directly with Ki-67 expression, and inversely with intratumoral CD8+ cell count. PKM2 was significantly overexpressed in poorly differentiated tumours and both PKM2 and LDHA were overexpressed in larger tumours. Multivariable analysis showed that combined expression of PKM2 and LDHA was an independent poor prognostic marker for survival. In conclusion, our results demonstrate a high expression pattern of two major glycolytic enzymes during pancreatic carcinogenesis, with increased expression in aggressive tumours and a significant adverse effect on survival.


Introduction
The aim of our study was to evaluate the expression of PKM2 and LDHA in pancreatic preneoplastic lesions (cysts and pancreatic intraepithelial neoplasia, PanIN) and cancers and to correlate it with patient outcome. Given that PKM2 and LDHA are involved in several cell proliferation signaling pathways [42], we also investigated whether the expression of these glycolytic enzymes correlated with the number of CD8+TILs and markers of tumour proliferation (Ki-67).

Materials and Methods Patients
This study included pancreatic biopsy or surgical resection specimens from 266 patients; 136 from University College London Hospital NHS Foundation Trust (UCLH) and another 130 from a commercially available tissue microarray (TMAs) (AccuMaxArray, ISU ABXIS CO., LTD, USA, and Insight Biotechnology Limited, UK). The UCLH cohort consisted of 136 patients with confirmed pancreatic ductal adenocarcinoma (PDAC) (n = 61), ampullary adenocarcinoma (n = 11), pancreatic cystic lesions (n = 49), chronic pancreatitis (n = 11) and normal pancreatic tissue (n = 4). These patients had been treated at UCLH between January 2005 and January 2010. A clinical database of the UCLH patients was created through the CoPath histology database, which included the following clinicopathological parameters: gender, age at diagnosis, sample type (biopsy or resection), type of tumour (ampullary or ductal), outcome (alive or dead), cause of death, length of follow-up if alive, presence or absence of metastatic disease at the time of histological diagnosis, whether resection was carried out, residual disease (R) status, post-resection recurrence, time to post-resection recurrence (if any), cancer stage, lymph node involvement, presence of perineural or lymphovascular invasion, degree of tumour differentiation, and whether patients received chemotherapy (Table 1). All data were collected from the date of histological diagnosis to the date of death or the end of data collection on 1 February 2015. Ethics committee approval was obtained from the Central London REC 3 Research Ethics Committee to perform immunohistochemistry on stored biopsy and resection specimens linked with a clinical database with the need for consent waived (REC reference 06/ Q0512/106, amendment date 30 July 2010). All patient samples were anonymised and de-identified prior to analysis.

Tissue microarray
A second validation set of 130 tissue samples was derived from commercially available pancreatic TMAs. A total of 206 tissue cores from 130 patients (83 male, 47 female, median age 59, range 32-80 years) were present on the TMAs, which included 63 cases of PDAC in duplicate, 19 PanIN (10 PanIN-1, 7 PanIN-2 and 2 PanIN-3), 10 chronic pancreatitis and 38 normal pancreatic tissue cores. Histological diagnosis of all cases was verified by haematoxylin and eosin staining. Clinical data provided with the TMAs included tumour size and stage, degree of tumour differentiation, lymph node involvement, metastatic status and tumour sites.

Immunohistochemistry
Standard immunohistochemistry methods were used to detect the expression of PKM2, LDHA, CD8 and Ki-67. Briefly, paraffin embedded sections were pre-heated in an oven for 30 minutes at 60°C, and then deparaffinised in xylene and hydrated through a series of graded ethanol (70%-95%). The sections were subjected to heat-mediated antigen retrieval with citrate buffer (pH 4.0) in an autoclave. Endogenous peroxidase activity was inhibited by 3% hydrogen peroxide for 20 minutes, followed by incubation with 2.5% normal horse blocking serum for 20 minutes. The sections were then incubated with primary antibody for one hour at room temperature for CD8 (CD8A Rabbit polyclonal antibody, Abnova, UK, Cat# PAB11235, 1:200) After three washes with PBS containing 0.5% Tween 20, the sections were incubated with horseradish peroxidase (HRP)-conjugated secondary antibody for 30 minutes. Primary antibody was detected using the 3, 3'-diaminobenzidine (DAB) or 3-amino-9-ethyl-carbazole (AEC) detection system kit. The sections were placed in haematoxylin for 3 minutes, then gently washed and mounted.

Double immunohistochemistry
A double immunostaining kit (PicTure double staining kit, Invitrogen, UK) was used for colocalisation of PKM2 and CD8 or Ki-67. Briefly, two different enzyme detection systems were used for the sequential double staining with initial use of the horseradish peroxidase system followed by the alkaline phosphate system. Sections were incubated with the anti-PKM2 primary antibody and HRP secondary antibody, and visualisation was developed with the DAB detection system. Following this step, sections were washed with PBS, incubated with blocking serum, followed by primary antibody (anti-CD8 or Ki-67) and Goat anti-Rabbit IgG alkaline phosphatase secondary antibody. Colour was developed by fast red with counter staining by haematoxylin.

Evaluation of immunohistochemistry staining
Immunohistochemistry was evaluated using a conference light microscope (AXIO Scope, Carl Zeiss MicroImaging GmbH, Gottingen, Germany). The assessment of the immunostained slides was performed independently by two observers, blinded to patient background information, and any disparity between the observers was resolved by using a conference microscope. A comprehensive scoring formula was used for the semi-quantitative evaluation of PKM2 expression as described previously [43], with intensity of staining scored as: 1, weak expression; 2, moderate expression; or 3, strong expression; the extent of staining scored as 1, <33% of tumour cells positive; 2, 33-67% of tumour cells positive or; 3, >67% of tumour cells positive. The intensity and extent scores were then multiplied to obtain a single scale of scores of 1, 2, 3, 4, 6, and 9. The scores of 1-3 were defined as weak (or negative) staining, whereas scores 4, 6 and 9 were considered as strong (or positive) staining. The numbers of CD8+ TILs and Ki-67 positive tumour cells were also counted independently by two observers. Initially, the whole slide was scanned at low magnification (x40) to identify the region with the highest density of intratumoural CD8 or Ki-67-positive tumour cells and then five random areas within that region were counted at high magnification (x400). The average number of CD8+ TILs was calculated and expressed as count per high power field (HPF). The cell proliferation index (PI) was expressed as a percentage of the number of nuclei stained positive for Ki-67 among 1000 tumour cells using a standardised grid.

Western Blot
PKM2 and LDHA expression was evaluated in pancreatic cancer cell lines by Western Blotting. Briefly, following protein extraction, protein concentration was measured by the Bicinchoninic Acid (BCA) assay and 15μg protein was run on a pre-cast gel (NuPAGE Novex 4-12% Bis-Tris 1.0mm, 10 wells gel, Invitrogen, USA) and transferred onto a 0.45μm pore size Polyvinylidene fluoride (PVDF) membrane (Invitrogen, USA). The membrane was blocked with 5% Bovine Serum Albumin (BSA) solution and incubated overnight at 4°C either with mouse anti-PKM2 (DF-4, ScheBo1Biotech, Giessen, Germany, 1:1000) or rabbit anti-LDHA antibody (Cell Signaling, UK, 1:1000), and then incubated with appropriate horseradish peroxidase conjugated secondary antibodies (Cell Signaling, UK, 1:2000). The antigen antibody reaction was detected by enhanced chemiluminescence substrate (Thermo Scientific, USA). Anti-β actin antibody (Cell Signaling, UK) was used as a protein loading control.

Statistical analysis
IBM SPSS Statistical software (Version 22, SPSS Inc., Chicago, IL, USA) was used for data analysis and graphics. Kaplan-Meier plots and log-rank tests were used to analyse survival and to identify differences between groups. One way ANOVA with Bonferroni post-hoc test was used for overall comparison of multiple groups, with the Mann-Whitney U-test for nonparametric tests and the Chi square test for differences between categorical data. Non-parametric correlation analyses between two continuous variables were performed by Spearman test. All test results were two-tailed, with effects summarised using 95% confidence intervals. Statistical significance was set at p<0.05.

Expression of PKM2 and LDHA in pancreatic cancer
Both PKM2 and LDHA were overexpressed in tumour cells compared with normal pancreatic tissue. A variable expression pattern of PKM2 was observed in tumour tissues, with relatively higher expression in poorly differentiated areas, in advancing margins of tumour nodules and in invasive (muscular and blood vessel) tumours (Fig 1A, 1B, 1E and 1F). Overall, PKM2 expression was predominantly associated with aggressive tumours. Preferential expression of PKM2 was observed in binucleated proliferating cells in tumour nodules (Fig 1D). Expression of PKM2 was noted in all cell compartments, including the cell membrane, cytoplasm and/or nucleus (Fig 1C and 1F). In contrast, LDHA expression was generally high in tumour as well as in preneoplastic tissues and pancreatitis without a specific pattern (Fig 2). LDHA expression was also detected in the cell membrane and/or cytoplasm and occasionally in the nucleus.
Similar expression of PKM2 and LDHA was observed in the UCLH cohort and TMA samples, with a staining score of > 3 in 64% and 73% of tumours, respectively, for PKM2, and in 76% of tumours for LDHA in both cohorts. The expression pattern in the pancreatic cancer cell lines was similar for both PKM2 and LDHA, except in the KP4 cell line in which the LDHA level was stronger than PKM2 level (Fig 3). In both cohorts, pancreatitis samples also highly expressed LDHA compared with PKM2 expression (Fig 4).
As shown in Fig 4, progressively higher PKM2 expression was observed along the transition to pancreatic cancer, with the lowest expression in pancreatic cysts (19%), intermediate in PanIN (37%) and highest in cancers (68%). PKM2 expression was approximately four-fold higher in pancreatitis (29%) compared with normal pancreatic tissue. Although LDHA expression was also significantly increased in cancers compared with normal pancreatic tissue (p < 0.0001, ANOVA), there were no significant differences between chronic pancreatitis, pancreatic cysts, PanIN and cancer samples (67%, 59%, 73% and 76%, respectively).

PKM2 and LDHA expression in pancreatic cancer cell lines
By Western blotting, high expression of PKM2 and LDHA was detected in 8 and 9 out of the 10 pancreatic cancer cell lines (Fig 3). The expression level of PKM2 and LDHA corresponded in all cell lines except in the KP4 in which the PKM2 level was weaker than the LDHA level. By immunohistochemistry, both PKM2 and LDHA had strong cytoplasmic and nuclear expression (Fig 3).

Association with clinicopathological parameters
The correlation between PKM2 and LDHA expression with clinicopathological characteristics is shown in Table 1. There was a significant inverse correlation between PKM2 expression and tumour differentiation in the UCLH cohort, with 83% of PKM2 positive tumours being less differentiated compared with 64% of PKM2 negative tumours (p = 0.047, Chi-square test) (data not shown).
A significantly higher number of CD8+ TIL was found in tumours with weak PKM2 or LDHA expression compared with tumours that had a strong expression (p = 0.0001, p = 0.005 respectively, Table 1). Furthermore, a significant association between tumour cell proliferation and expression of both PKM2 and LDHA was observed; the number of tumour nuclei expressing Ki-67 was more than 2-fold higher in PKM2 and LDHA expressing tumours compared  Percentages of PKM2 and LDHA expression in different tissue types. PKM2 was strongly expressed by pancreatic cancer tissue specimens and was significantly higher than in normal, pancreatitis, pancreatic cyst and PanIN tissues (P<0.001). Expression of LDHA was significantly higher in pancreatic cancer than in normal pancreas (P<0.001), whereas, there was no significant differences in LDHA expression between pancreatic cancer, PanIN, pancreatic cysts and pancreatitis.  (Figs 5 and 6). In the TMA cohort, the expression of PKM2 and LDHA correlated with tumour size. PKM2 expression was observed in 54.5%, 77.8% and 90.9% of tumours that were 2.5, 2.6-3.9 and 4 cm in size, respectively. Positive LDHA expression was found in 59.1%, 77.8% and 100% in tumours that were 2.5, 2.6-3.9 and 4 cm in size, respectively. There were no significant differences between PKM2 or LDHA expression and tumour location, lymph node involvement, T-stage and metastatic status.

Correlation between PKM2 and LDHA expression and patient survival
We next examined whether expression profile of PKM2 and LDHA predicted survival. Patients with tumours scoring > 3 for PKM2 or LDHA expression had significantly worse survival compared with those that weakly expressed PKM2 and/or LDHA. Of the 72 pancreatic cancer samples (UCLH cohort), 46 (64%) strongly expressed PKM2 and these patients had a median survival of only 8.9 months compared with 28.9 months in the 26 (36%) patients with weak (negative) PKM2 tumour expression (p = 0.016, log-rank test, Fig 7). Similarly, 55 (76.4%) patients with positive LDHA tumour expression had a median survival of 10.9 months compared with 34.5 months of the 17 (23.6%) patients with weak (negative) LDHA tumour expression (p = 0.029, log-rank test, Fig 7). Moreover, when the expression profile of PKM2 and LDHA was combined, the survival of patients with negative expression for both or positive for  (Table in S1 Table).
Additionally, 27 of the 72 patients (UCLH cohort) underwent surgical resection for PDAC (n = 16) or ampullary adenocarcinoma (n = 11). Those who underwent surgical resection had a significantly longer survival than those who did not undergo surgery (26.5 vs 7.0 months, P<0.0001; log rank test). As expected, patients with ampullary adenocarcinoma had a better prognosis after surgery than those with PDAC, with 5 of 11 (45.5%) ampullary carcinoma patients alive at last contact, compared with only 1 out of 16 (6.3%) patients with PDAC (P<0.0001; log rank test).

Discussion
There is mounting evidence that cancer cells have elevated glucose uptake with a concomitant increase in lactate production through sequential catalytic enzyme mediated processes [44,45]. PKM2 and LDHA are two crucial glycolytic enzymes that facilitate these processes to confer cancer cells with a growth advantage over normal cells. To date, serum PKM2 has been identified as a diagnostic and prognostic marker with comparable sensitivity and specificity to serum CA19-9 marker in pancreatic cancer [46][47][48]. However, there are limited data on the expression pattern and prognostic impact of PKM2 and LDHA in pancreatic cancer [32,49,50]. To the best of our knowledge, our study is the first to evaluate the prognostic impact of combined PKM2 and LDHA expression in the initiation and progression of pancreatic cancer. A recent study showed overexpression and phosphorylation of both of these enzymes in thyroid cancer compared with benign goitre [51]. Our results concur with their findings showing significant overexpression of PKM2 and LDHA in pancreatic cancers compared with normal pancreatic tissue. The step-wise initiation and development of pancreatic cancer often begins with pancreatitis, ductal metaplasia, cyst formation or PanIN lesions, leading to pancreatic cancer. Interestingly, our results demonstrate overexpression of LDHA at a very early stage along the carcinogenetic pathway from pancreatitis through cyst/PanIN to cancer with the highest expression in the most aggressive tumours. In contrast, PKM2 expression increased progressively along the transition to pancreatic cancer and was lowest in cysts, intermediate in PanIN lesions and highest in cancers. Although the exact mechanism of this differential expression pattern remains to be elucidated, it is possible that the pre-neoplastic lesions acquire the glycolytic phenotype through LDHA overexpression and then LDHA itself or other oncogenes induce PKM2 overexpression at later stages when tumour cell proliferation rates are higher. In fact, it has been recently shown that the epidermal growth factor receptor (EGFR) induces βcatenin transactivation and c-myc expression, upregulating LDHA, which in turn induces upregulation of PKM2 expression by alternating splicing of the gene from M1 to M2 type [52]. These findings might partially explain the consistent overexpression of LDHA throughout the tumorigenic process and the progressive overexpression of PKM2 along the carcinogenetic pathway. The expression and enzymatic activity of LDHA and PKM2 can also be modulated by tyrosine phosphorylation at various residues (Y10 and Y105, respectively) by the oncogenic tyrosine kinase fibroblast growth factor receptor-1 [53]. The differential expression pattern suggests that PKM2 (or a combination of PKM2 and LDHA) would be a better choice for discriminating cancer from pre-neoplastic lesions compared with LDHA alone, except in pancreatitis where both markers are highly expressed.
Treating pancreatic cancer is highly challenging due to late diagnosis, and lack of appropriate prognostic markers and effective therapies. Our results demonstrate that both PKM2 or LDHA are significant prognostic markers in pancreatic cancer and the combination provides improved stratification of outcome. These results are in line with previous publications showing a significant prognostic effect of LDHA or PKM2 in other tumour types, including squamous cell carcinoma, cholangiocarcinoma and gastric cancer [43,54,55]. The exact mechanisms associated with the overexpression of PKM2 and LDHA that lead to poor prognosis remain unclear. Very recently, Rajeshkumar et al (2015) found that the LDHA small molecule inhibitor FX11 can impede tumour growth, reduce tumour cell proliferation and induce apoptosis in a patient-derived mouse xenograft model of pancreatic cancer with mutant TP53, while tumours harbouring wild-type TP53 were completely resistant to FX11 [56].
Pancreatic cancer is one of the more aggressive tumour types, with a very poor overall survival. The results of our study clearly show up-regulation of both PKM2 and LDHA in pancreatic cancer and implicate the high expression of these two glycolytic enzymes in the development and progression of pancreatic cancer through enhanced proliferation, migration, invasion and angiogenesis. Feng et al (2014) reported that the knockdown of PKM2 in pancreatic cancer cells reduces cell proliferation, migration and invasion of pancreatic cancer cells in vitro and induces apoptosis by increasing the expression level of BAX protein and reducing the expression level of Bcl-2 protein [57]. In a related study, Azoitei et al (2013) demonstrated that the knockdown of PKM2 by siRNA attenuates pancreatic cancer cell proliferation and migration in vitro and impedes tumour growth, angiogenesis and Ki67 expression in the pancreatic cancer xenograft model [58].
In this study, we noticed a significant direct correlation between staining scores and the tumour cell proliferation index, and by using the double labeling technique, we were able to identify Ki-67 positive proliferating cells topographically co-localising in PKM2 positive areas. It has been well recognized that these glycolytic enzymes translocate to the nucleus and interact with several other oncogenic transcription factors and transcribe several cell proliferating signaling pathways including Stat3, β-catenin, HIF-1, Oct-4, cyclin D1 [42]. Moreover, PKM2 has been implicated in the phosphorylation of the chromosomal spindle checkpoint protein Bub3--Bub1-Blinkin complex, which ensures fidelity of chromosomal segregation during cell proliferation [59]. Although we could not find any significant correlation with tumour stage and metastasis status, strong expression of PKM2 was observed in metastatic tumours invading muscle and blood vessels, indicating the aggressive phenotype of pancreatic tumours expressing the glycolytic enzymes. We and others have previously shown in cholangiocarcinoma and lung cancer, respectively, that tumour-associated angiogenesis is induced by PKM2 and LDHA, which could be another contributing factor to poor prognosis [43,60].
Host immune evasion is a hallmark of the aggressive tumour phenotype. Although a limited number of studies have provided evidence of mechanism between host immune suppression and the tumour glycolytic phenotype, to date, there is no concrete evidence showing a correlation between the expression of PKM2, LDHA and CD8+ effector cell infiltration. Interestingly, we noticed a significant inverse correlation between the PKM2 and LDHA expression and CD8+ cell infiltration, with an accumulation of CD8+ cells in tumours that did not express PKM2. Recently, Crane et al (2014) reported that LDHA secreted by glioblastoma cells downregulated the Natural Killer group 2, member D receptor on natural killer cells and thus subverted host immune surveillance [61]. Similarly, Liu et al (2015) reported that PKM2 expression was related to increased infiltration of primary and metastatic tumours by myeloid derived suppressor cells, responsible for the suppression of NK cells and induction of host immune suppression by regulatory T cells [62]. We postulate that the compromised immune surveillance induced by enhanced expression of PKM2 and LDHA, and reduced CD8+ effector T-cells might be another contributing factor associated with poor prognosis of the glycolytic phenotype of pancreatic cancer.
Limitations of our study include limited sample size and lack of some clinicopathological information. An optimal scoring system for PKM2 and LDHA expression has not yet been established, with various scoring systems having been reported [43,[63][64][65][66]. In the current study, PKM2 and LDHA expression scores were classified as negative or positive, which was based on the intensity and extent of PKM2 and LDHA expression across tumour areas. Moreover, other treatments such as chemotherapy might have affected the levels of PKM2 and LDHA expression and overall survival of patients.
In conclusion, we have shown a differential expression pattern for PKM2 and LDHA from cysts through PanIN lesions to pancreatic cancer, with upregulation of LDHA throughout the carcinogenetic process and a progressive upregulation of PKM2 expression along the carcinogenetic pathway. Moreover, the combined expression of these glycolytic enzymes is a strong independent marker of poor prognosis, attributable to increased cell proliferation, larger tumour size and host immune evasion. Further studies are underway to evaluate these markers as possible targets for therapy in pre-clinical models of pancreatic cancer.