Prognostic Implication of pAMPK Immunohistochemical Staining by Subcellular Location and Its Association with SMAD Protein Expression in Clear Cell Renal Cell Carcinoma

Although cytoplasmic AMP-activated protein kinase (AMPK) has been known as a tumor-suppressor protein, nuclear AMPK is suggested to support clear cell renal cell carcinoma (ccRCC). In addition, pAMPK interacts with TGF-β/SMAD, which is one of the frequently altered pathways in ccRCC. In this study, we investigated the prognostic significance of pAMPK with respect to subcellular location and investigated its interaction with TGF-β/SMAD in ccRCC. Immunohistochemical staining for pAMPK, pSMAD2 and SMAD4 was conducted on tissue microarray of 987 ccRCC specimens. Moreover, the levels of pSMAD2 were measured in Caki-1 cells treated with 5-aminoimidazole-4-carboxamide ribonucleotide. The relationship between AMPK/pAMPK and TGFB1 expression was determined using the TCGA database. As a result, pAMPK positivity, either in the cytoplasm or nuclei, was independently associated with improved ccRCC prognosis, after adjusting for TNM stage and WHO grade. Furthermore, pAMPK-positive ccRCC displayed increased pSMAD2 and SMAD4 expression, while activation of pAMPK increased pSMAD2 in Caki-1 cells. However, AMPK/pAMPK expression was inversely correlated with TGFB1 expression in the TCGA database. Therefore, pAMPK immunostaining, both in the cytoplasm and nuclei, is a useful prognostic biomarker for ccRCC. pAMPK targets TGF-β-independent phosphorylation of SMAD2 and activates pSMAD2/SMAD4, representing a novel anti-tumoral mechanism of pAMPK in ccRCC.


Introduction
Clear cell renal cell carcinoma (ccRCC), the most common type of renal cell carcinoma, is characterized by genetic alterations that regulate cellular metabolism [1,2]. For example, accumulation of an oxygen-sensing protein, hypoxia-inducible factor (HIF)-α, by the mutational loss of the VHL tumor-suppressor gene supports ccRCC progression via angiogenesis and epithelial-mesenchymal transition (EMT) [2]. Another metabolic hallmark of ccRCC is a change in the glucose-sensing machinery caused by constitutive activation of the PI3K/AKT/MTOR pathway [1,2]. Activated mammalian target

Patients and pAMPK IHC Staining
The demographic and clinicopathological characteristics of the discovery and validation of ccRCC patient cohorts are summarized in Table 1. In the discovery and validation cohorts, the male-to-female sex ratios were 2.8 and 3.0 and the median ages were 58 (range, 20-81) and 56 (range, 24-84) years, respectively.
IHC staining revealed cytoplasmic ( Figure 1A,B) and nuclear ( Figure 1B,C) pAMPK positivity in 250 (55.2%) and 228 (50.3%) samples, respectively, of the discovery cohort, and in 242 (45.3%) and 231 (43.3%) patients, respectively, of the validation cohort ( Table 1). The results of this dichotomous assessment of cytoplasmic and nuclear pAMPK expression, described in detail in Materials and Methods 4.3., showed high agreement between the tissue microarray (TMA) and matched whole-section slides from 10 randomly selected patients. Both cytoplasmic and nuclear pAMPK positivity was significantly associated with small tumor size (p < 0.001), low TNM stage (p < 0.001), and low WHO grade (p < 0.001) in the discovery cohort, which was verified in the validation cohort (Table 1).  Figure 1. Immunohistochemical staining for pAMPK. (a) More than 50% of tumor cells expressed pAMPK both in the cytoplasm and nuclei; (b) More than 50% of tumor cells are stained for pAMPK in the cytoplasm but less than 10% of tumor cells show nuclear staining for pAMPK; (c) More than 50% of the tumor cells exhibits pAMPK only in the nuclei.

Positive IHC Staining for pAMPK Was Significantly Associated with Improved ccRCC Prognosis
The median follow-up periods from the discovery and validation cohorts were 121 (range, 1-178) and 102 (range, 2-288) months, respectively. The median overall survival (OS) was not reached in either cohort. Kaplan-Meier and log-rank tests showed that pAMPK-positive ccRCC, either in the cytoplasm or in the nuclei, was associated with longer progression-free survival (PFS) (p < 0.001), overall survival (OS) (p < 0.001), and cancer-specific survival (CSS) (p < 0.001) than pAMPK-negative ccRCC, both in the discovery and validation cohorts ( Figure 2). Univariate Cox regression analysis showed that patients with pAMPK-positive ccRCC (with either a cytoplasmic or nuclear pattern) were less likely to experience disease progression (p < 0.001), death (p < 0.001), and disease-specific death (p < 0.001); these findings were true in both the discovery (Table 2) and validation (Table 3) cohorts. A multivariate analysis incorporating TNM stage and WHO grade showed that pAMPK positivity is an independent prognostic factor for favorable PFS (cytoplasmic, hazard ratio [HR] = 0.260, p < 0.001; nuclear, HR = 0.308, p < 0.001), OS (cytoplasmic, HR = 0.656, p = 0.032), and CSS (cytoplasmic, HR = 0.374, p = 0.001; nuclear, HR = 0.232, p = 0.003) in the discovery cohort ( Table 2). The prognostic significance of cytoplasmic and nuclear pAMPK expression adjusted for TNM stage and WHO grade was confirmed in the validation cohort (Table 3).

Positive IHC Staining for pAMPK Was Significantly Associated with Improved ccRCC Prognosis
The median follow-up periods from the discovery and validation cohorts were 121 (range, 1-178) and 102 (range, 2-288) months, respectively. The median overall survival (OS) was not reached in either cohort. Kaplan-Meier and log-rank tests showed that pAMPK-positive ccRCC, either in the cytoplasm or in the nuclei, was associated with longer progression-free survival (PFS) (p < 0.001), overall survival (OS) (p < 0.001), and cancer-specific survival (CSS) (p < 0.001) than pAMPK-negative ccRCC, both in the discovery and validation cohorts ( Figure 2). Univariate Cox regression analysis showed that patients with pAMPK-positive ccRCC (with either a cytoplasmic or nuclear pattern) were less likely to experience disease progression (p < 0.001), death (p < 0.001), and disease-specific death (p < 0.001); these findings were true in both the discovery ( Table 2) and validation (Table 3) cohorts. A multivariate analysis incorporating TNM stage and WHO grade showed that pAMPK positivity is an independent prognostic factor for favorable PFS (cytoplasmic, hazard ratio [HR] = 0.260, p < 0.001; nuclear, HR = 0.308, p < 0.001), OS (cytoplasmic, HR = 0.656, p = 0.032), and CSS (cytoplasmic, HR = 0.374, p = 0.001; nuclear, HR = 0.232, p = 0.003) in the discovery cohort ( Table 2). The prognostic significance of cytoplasmic and nuclear pAMPK expression adjusted for TNM stage and WHO grade was confirmed in the validation cohort (Table 3).

Discussion
Here, we have revealed the prognostic implication of IHC staining for pAMPK in large cohorts of ccRCC patients. pAMPK positivity was related to low-risk pathological traits of ccRCC, including small tumor size, early TNM stage, and low WHO grade, and was an independent favorable prognostic factor in ccRCC. These results were consistent across samples with cytoplasmic and/or nuclear pAMPK IHC staining, suggesting that IHC staining for pAMPK may be a useful prognostic biomarker for ccRCC. Furthermore, pAMPK-positive ccRCC showed higher levels of nuclear pSMAD2 and SMAD4 expression than those in pAMPK-negative tumors. Consistent with this result, Caki-1 cells treated with AICAR showed increased expression of pSMAD2. SMAD2/SMAD4 mRNA also showed a modestly positive correlation with AMPK mRNA/phosphoprotein, whereas TGFB1 showed an inverse correlation, suggesting that pAMPK activates SMAD expression through a non-TGF-β pathway.
pAMPK inhibits growth, invasion, EMT, and metastasis by modulating various signaling axes, including downregulation of mTORC1, HIF, and NF-κB and upregulation of Forkhead box class O (FoxO)3a and p53 [7,12,[23][24][25]. In addition, pAMPK is responsible for restricting ATP-consumption by directly suppressing fatty acid synthesis enzymes, such as acetyl-CoA carboxylase and fatty acid synthase. In other words, pAMPK shifts metabolism away from the Warburg effect-like status, an environment that is supportive of ccRCC progression [1,6,7,24]. Our clinical results as well as previous preclinical data propose that AMPK activation may be useful as a ccRCC treatment strategy [25,26]. pAMPK expression may affect the treatment response to targeted agents in ccRCC. For example, rapamycin, an inhibitor to mammalian target of rapamycin, displayed additive counteraction against the invasion and growth of renal cell carcinoma when this was administered together with AICAR [25]. In addition, because pAMPK inhibits HIF-α activity and HIF-α stably upregulates vascular endothelial growth factor in ccRCC [2,25], the association between pAMPK expression and vascular endothelial growth factor receptor tyrosine kinase inhibitors might be worth investigating. However, the subcellular location-specific tumor-suppressive functions of AMPK have been controversial. For example, nuclear pAMPK increased transcription of pro-cancerous molecules, such as cyclin D1, c-myc, and Oct4, in cancer cells under metabolic stresses (e.g., in glucose-deprived conditions) [13,14]. The clinical significance of this nuclear pAMPK immunoreactivity has never been investigated in vivo because prior studies have reported that IHC staining for pAMPK is confined to the cytoplasm of normal and neoplastic human tissue [10][11][12]27,28]. In the ccRCC TCGA reports, the prognostic significance of pAMPK was demonstrated using protein array, without considering pAMPK subcellular localization [1,6]. We observed that both cytoplasmic and nuclear pAMPK is independently associated with favorable prognoses in ccRCC. Previous studies have revealed that nuclear pAMPK can function in either an inhibitory or a stimulatory manner in renal cell carcinoma, depending on context: the oncogenic function of nuclear AMPK is confined to glucose-deprived renal cell carcinoma cells, whereas regular tumor-suppressive functions predominate under usual conditions [14]. Therefore, it is reasonable to speculate that the hostile environments that encourage nuclear pAMPK to stimulate ccRCC progression were present in only a restricted number of patients in this study. In addition, it is noteworthy that glucose starvation can also lead to the activation of various signals other than AMPK, including Akt and ERK [23].
The TGF-β/SMAD signaling pathway, a major regulator of carcinogenesis and the progression of various tumor types [29], is frequently altered in ccRCC [1]. In fact, pAMPK has been shown to inhibit SMAD-mediated TGF-β signal transduction through various mechanisms, by which pAMPK exhibits anti-fibrogenic and anti-EMT functions [15][16][17][18]20]. For example, pAMPK prevented TGF-β-mediated phosphoactivation of SMAD2 and/or SMAD3 in vascular smooth muscle and breast cancer cells [17,20]. In mesenchymal cells, pAMPK also reduced the stability and nuclear localization of SMAD4 [18] and the interaction of SMAD3 with its coactivator, p300, without affecting the phosphorylation of SMAD2 or SMAD3 [16]. In contrast, we demonstrated that pAMPK-positive ccRCC showed higher levels of nuclear pSMAD2 and SMAD4 than those in pAMPK-negative tumors. We hypothesize that this positive relationship between pAMPK and pSMAD2/SMAD4 may be involved in an important, as-yet-unidentified mechanism that is critical to the inhibition of ccRCC. In addition, this may explain a previous report that ccRCC displaying nuclear SMAD2/SMAD3/SMAD4 protein expression had, for unknown reasons, favorable outcomes [22]. It is known that LKB1, kinase upstream of pAMPK, dictates phosphoactivation of AMPK, except on limited occasions when AMPK might be activated by calcium/calmodulin-dependent protein kinase kinase 2, TGF-β-activated kinase 1, or AMP through phosphorylation or allosteric modification [7,8]. AICAR is an AMP-mimetic that directly binds to AMPK and facilitates its phosphorylation by LKB1 [7,19]. Therefore, it is safe to say that LKB1 played a critical role in activating pAMPK in this study. LKB1 may directly increase TGF-β signaling in an AMPK-independent way, as suggested in other studies [30,31]. Nevertheless, we identified a negative correlation between TGFB1 and both AMPK mRNA and pAMPK proteins in the TCGA ccRCC database, in agreement with a previous report [20]. Low TGFB1 mRNA levels are likely attributable to an attenuated TGF-β/SMAD cascade [31]. The positive correlation between SMAD2/SMAD4 mRNA levels and AMPK/pAMPK in the TCGA database was weak and inconsistent, compared to the tight correlation between pSMAD2/SMAD4 and pAMPK expression in our TMA and Western blot analyses. Therefore, we hypothesize that pAMPK post-transcriptionally or post-translationally activates pSMAD2/SMAD4 by phosphorylating SMAD2 via a TGF-β-independent mechanism. Furthermore, the results suggest that pAMPK probably conducts a tumor-suppressive function through pSMAD2/SMAD4 activation, at least in part. Upregulation of nuclear FoxO3a may connect the coactivation of pAMPK and SMAD. FoxO3 is also a good candidate for an interdependent molecule mediating the antitumoral functions of pAMPK and pSMAD2/SMAD4 in ccRCC. In ovarian cancer, FoxO3a interacted with a SMAD2/SMAD3/SMAD4 complex in the nucleus, which could maintain activated SMAD proteins at high levels, and this interaction in turn promoted the cell-cycle arrest synergistically [32,33]. In addition, previous studies showed that the activation of AMPK increased both the nuclear localization and stability of FoxO3a in cancer and enhanced the transcription of autophagy-associated genes, leading to cell death [23,34]. Taken together, this putative pAMPK/FoxO3/SMAD interaction may lead to favorable outcomes in pAMPK-positive ccRCC with high expression of SMAD proteins. Presumably, this is a novel biological function of pAMPK/pSMAD2/SMAD4 in ccRCC that warrants further studies. Secondly, given that AMPK is a kinase of diverse targets and the TGF-β/SMAD pathway engages in a reciprocal cross talk with various molecules [7,21], it is readily assumable that pAMPK might activate and cooperate with pSMAD2/SMAD4 to downregulate ccRCC, either directly or indirectly by way of modulating other pathways.
There are a few limitations in this study. Although it is well known that pSMAD2 and pSMAD3 are very similar proteins that can form a complex with SMAD4 to activate transcriptional responses [29], the interaction between pSMAD3 and pAMPK was not investigated in this study. Along with the positive correlation between pAMPK and nuclear pSMAD2/SMAD4 in IHC, the Western blot result puts forward the idea that pAMPK overexpression phosphorylates SMAD2 and subsequently activates pSMAD2/SMAD4 in ccRCC. However, more detailed in vitro studies adopting a constitutively active mutant of AMPK or depletion of pAMPK using compound C or siRNA to AMPK would profoundly improve the understanding of the causal relationship between pAMPK and pSMAD2/SMAD4 activation. In addition, despite the unequivocal overexpression of pSMAD2 and SMAD4 in pAMPK-positive tumors, it still remains uncertain whether pAMPK directly induces SMAD-mediated transcriptional activities in ccRCC. Previous reports demonstrated that SMAD activities were regulated by pAMPK at the transcriptional level in various conditions [15,16,20], which accompanied increased expression of the pAMPK and AMPKα2 subunit in the nuclei of human mesangial cells as well [15]. Although these results imply that pAMPK may directly associate with SMAD proteins and modulate their transcriptional function in the nuclei [15], this phenomenon has not been verified in ccRCC. Further investigation will be required to elucidate the detailed functional effects of pAMPK-induced activation of SMAD proteins, including SMAD-responsive transcription and the regulation of its downstream targets.

Patients' Cohorts
In total, 987 ccRCC samples were surgically resected at the Seoul National University Hospital between 1995 and 2008, and separated into discovery (n = 453; 2003-2008) and validation (n = 534; 1995-2004) cohorts. Of these samples, 294 cases of the discovery cohort and 493 cases of the validation cohort were radical nephrectomy specimens, while the others were partial nephrectomy specimens. Each tumor sample was reviewed for histologic type, TNM stage, and WHO grade. The TNM stage was reclassified according to the American Joint Committee on Cancer-Cancer Staging Manual, 8th edition [35]. Clinical data were obtained from medical records. Patients who received neoadjuvant treatment, displayed bilateral disease at the time of diagnosis, or who had Von Hippel-Lindau syndrome were excluded. This study was approved by the Institutional Review Board of Seoul National University Hospital (H-1810-150-983).

Establishment of Cut-Off Criteria for pAMPK IHC Staining Positivity
pAMPK expression was observed in the cytoplasm and the nuclei of tumor cells (Figure 1). Firstly, the extent of tumor cells in the discovery cohort displaying at least moderate immunoreactivity was assessed semiquantitatively as follows: <10%, 10-50%, and >50%. Based on both sample distribution and prognostic significance, 10% staining was defined as the best cut-off value for both cytoplasmic and nuclear pAMPK positivity ( Figure S1). Next, the prognostic significance of pAMPK staining was confirmed in the validation cohort using the same cut-off criteria. Regarding discordant cases from duplicate TMA cores, the lower pAMPK expression level was used. To account for intratumoral heterogeneity, IHC staining for pAMPK was also conducted on whole sections from 10 randomly selected cases in the discovery cohort. The percentage of tumor cells (%) positive for nuclear pSMAD2 and SMAD4 was calculated from both TMA cores, using QuPath version 0.1.2 [36] on digitally scanned slides (Aperio AT2, Leica Biosystem, Wetzlar, Germany). The mean values of pSMAD2 and SMAD4 are shown.

Western Blot Analysis
A human ccRCC cell line, Caki-1, was obtained from the Korean Cell Line Bank (KCLB, Seoul, Republic of Korea) and was cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% fetal bovine serum in a 5% CO 2 humidified incubator. To evaluate the AMPK/SMAD2 signaling pathway, we used the AMPK activator, AICAR (Sigma, St. Louis, MO, USA). Caki-1 cells were treated with 1 mM AICAR for 6 hours, and then, the cells were harvested for Western blot analysis. Cell lysates were electrophoretically resolved on a 10% polyacrylamide gel in a sodium dodecyl sulfate buffer and then transferred onto nitrocellulose membranes. Afterward, the blots were incubated with antibodies against pAMPK T172 (Cat. #2535, Cell Signaling Technology) and pSMAD2 S465/467 (Cat. #3108; Cell Signaling Technology). These experiments were conducted in triplicate and are presented as the mean ± standard deviation.

Characteristics of the TCGA ccRCC Dataset
ccRCC mRNA sequencing and reverse phase protein array data with clinicopathological information generated by the TCGA Research Network were obtained using the Broad Institute GDAC Firehose [37]. The mRNA sequencing dataset was generated from 552 primary ccRCC samples and quantified by RSEM (RNA-Seq by Expectation) [38]. The median age in this sample was 61 years (range, 26-90) and the male-to-female sex ratio was 1.8:1. Approximately, 60% (316/522) and 46% (236/514) of these patients had a low TNM stage (I or II) and low Furhman grade (1 or 2), respectively. We next constructed a design matrix using the DGEList function in the EdgeR module. To filter out genes with low expression levels, we excluded those genes with counts per million (cpm) values <1 in at least half of the samples [39]. As a result of this filtering, AMPKγ3 (PRKAG3) was omitted. RSEM read counts underwent Trimmed Mean of M-values (TMM) normalization and logCPM transformation using voom [40]. Of the patients whose samples were included in the mRNA dataset, 472 also had reverse phase protein array information, including pAMPK T172 protein expression levels. These data were analyzed to examine the correlation between AMPK/pAMPK T172 and TGFB1/SMAD mRNA expression.

Statistical Analysis
The interrelation between pAMPK IHC staining and clinicopathological characteristics was analyzed using Pearson's χ 2 test with Yates' correction for categorical variables and with the Mann-Whitney test for continuous variables. The PFS period was calculated as the interval between surgery and recurrence, progression, metastasis, or the last follow-up visit. The OS duration was defined as the period between surgery and death from any cause or the last follow-up. The CSS duration was defined as the interval between surgery and cancer-related death or the last follow-up visit. Kaplan-Meier analysis and the log-rank test were used to compare survival. A Cox proportional hazard regression model was used for univariate and multivariate survival analyses. The strength and direction of the linear relationships among AMPK/pAMPK, TGFβ1, and SMAD2/SMAD4 expression was assessed using Pearson's r test. All statistical analyses were performed using SPSS Statistics 25 (IBM Co., Armonk, NY) or in R, with a 2-tailed p < 0.05 considered statistically significant.

Conclusions
Both cytoplasmic and nuclear pAMPK immunostaining are independently associated with favorable outcomes in ccRCC. pAMPK positivity is associated with nuclear overexpression of pSMAD2 and SMAD4, through stimulation of TGF-β-independent phosphorylation of SMAD2, which is a novel antitumoral mechanism for pAMPK in ccRCC. Understanding the interaction between pAMPK and SMAD proteins will facilitate the use of AMPK activation as a strategy for ccRCC treatment.