Pyruvate kinase M2 and the mitochondrial ATPase Inhibitory Factor 1 provide novel biomarkers of dermatomyositis: a metabolic link to oncogenesis

Inflammatory myopathies (IMs) is a group of heterogeneous diseases characterized by muscle weakness and inflammatory infiltrates within the skeletal muscle. Despite presenting unknown etiology, inflammatory and bioenergetic disturbances have been argued in most of cases. Due to their similar clinical presentation, polymyositis (PM), dermatomyositis (DM) and sporadic inclusion-body myositis (sIBM) are the three major groups ascribed to IMs [1]. A fourth and fifth subtypes termed necrotizing auto-immune myositis and overlap myositis are also being recognized within the group of IMs [2]. IMs are considered rare diseases due to their low incidence, about of 2.1 to 7.7 new cases per every million inhabitants/year. sIBM is the most common acquired myopathy in patients above 50 years, with some geographical differences [3‐5]. There are few biomarkers that could help the diagnosis and management of patients affected by IMs, in particular myositis-associated or specific antibodies. The elevated serum activity of creatine kinase, lactate dehydrogenase and aldolase are currently used as activity indicators of all subtypes of IMs [2].

Reverse phase protein microarrays (RPPA) is a high-throughput quantitative technique adequate for multiplexed analysis of protein expression in minute amounts of sample in a large variety of biological specimens [6, 7]. Over the last decade, RPPA technique has provided a precious tool in the discovery of biomarkers of disease which might become indispensable in the progress of diagnostic, prognostic and therapeutic fields. The Achilles heel for the development of a reliable RPPA platform, is the availability of high-affinity and specific antibodies against the proteins investigated [8, 9]. Herein, we have studied the putative relevance of proteins of energy metabolism as diagnostic biomarkers in IMs using RPPA. To this aim, we have studied the expression of enzymes of glucose metabolism and of oxidative phosphorylation in a cohort of thirty-two muscle biopsies including samples from control and PM, DM and sIBM affected patients using validated monoclonal antibodies. The final purpose of the study is to translate the “signature” of energy metabolism to bed-side application of patients affected with IMs.

To explore the potential applicability of a signature of metabolism in the diagnosis of IMs we selected for study representative proteins from glycogenolysis (Glycogen phosphorylase, PYGM), glycolysis (glyceraldehyde-3-phosphate dehydrogenase, GAPDH, pyruvate kinase M2, PKM2 and lactate dehydrogenase A, LDHA), oxidative phosphorylation (β-subunit of the H⁺-ATP synthase, β-F1-ATPase and ATPase Inhibitory Factor 1, IF1), and electron shuttling of glycolytic NADH to mitochondria (glycerol-phosphate dehydrogenase 1, GPD1). In addition, β-actin and heat shock protein 60 kDa (Hsp60) were also studied as structural markers of the cell and mitochondria, respectively. The major limitation of quantitative RPPA is the availability of specific monoclonal antibodies (mAbs) against the proteins being studied. Figure 1 shows that the mAbs used in this study only recognized a single protein band at the expected molecular weight in human muscle extracts validating their utilization in RPPA techniques [11‐13].

A representative protein microarray illustrating the printing protocol of human muscle biopsies developed with antibodies against the glycolytic PKM2 is shown in Fig. 2a. Microarrays developed with the other antibodies are shown below (Fig. 2a). Protein extracts from muscle biopsies of control (green boxed in Fig. 2a), PM, DM and IBM (yellow, red and blue boxed in Fig. 2a, respectively) were prepared and spotted onto RPPA in triplicate from left to right (Fig. 2a). Increasing amounts of BSA and murine IgGs (black and brown boxed in Fig. 2a, respectively) were spotted in the array as control of the background of the assay and positive control of antibody recognition, respectively. Increasing protein amounts of cellular extracts derived from HCT116 cells were also spotted (magenta boxed in Fig. 2a). The HCT116 extracts revealed a linear increase in fluorescent intensity as the amount of protein increases (Fig. 2b), providing the linear plot of the assay (Fig. 2b, see also Additional file 1: Figure S1 for the rest of the mAbs tested). The arrays illustrate the specific recognition of the corresponding antigen in minute amounts of protein of HCT116 extracts as well as in the biopsies (Fig. 2a). As expected BSA did not show any fluorescent signal, confirming the absence of non-specific binding of the primary and secondary antibodies onto the spotted proteins (Fig. 2a). In contrast, a linear increase in signal was observed in murine IgGs (Fig. 2a), confirming that the secondary antibody works properly. The quantification of the expression of each marker in control (n = 6) and patient (n = 26) biopsies was calculated by interpolating the fluorescent intensity signal obtained in the sample in the standard curve of HCT116 cells (Fig. 2b and Additional file 1: Figure S1). The box plots in Fig. 3 display the results of protein expression in PM, DM and sIBM when compared to controls revealing the inter-individual variation within each group.

Interestingly, only DM and sIBM patients showed significant alterations of the expression level of the studied markers when compared to controls (Fig. 3). Muscle biopsies from patients affected with DM showed an increase in the expression of Hsp60 and β-actin concurrent with a similar increase in the expression of PKM2 and the mitochondrial ATPase inhibitor factor IF1 (Fig. 3). These changes occurred in the absence of relevant changes for the expression of other markers and with a significant reduction in PYGM expression (Fig. 3). In contrast, biopsies from sIBM patients showed a significant reduction in the expression of the cytoplasmic GAPDH, LDH-A, PYGM, GPD1 and mitochondrial Hsp60 (Fig. 3). Concurrently, a significant increase in β-actin and IF1 expression (Fig. 3) was observed in sIBM. Overall, and from the point of view of a potential biomarker to distinguish between DM from normal biopsies and the rest of the other IMs stands the sharp increase in PKM2, IF1 and Hsp60 expression (Fig. 3). In fact, PKM2 alone achieved a sensitivity of 96.1% and specificity of 100% (AUC of 0.988) (Fig. 4). Details of the sensitivity (ROC) for IF1 and Hsp60 are provided in Additional file 2: Table S1). In the same line, the down-regulation of glycolytic markers distinguishes sIBM from control biopsies and other IMs (Fig. 3; and see Additional file 2: Table S1).

Representative western blot analysis of the three glycolytic markers investigated in RPPAs (Additional file 3: Figure S2) confirmed the higher expression of PKM2 in DM samples and the downregulation of both GAPDH and LDH-A in sIBM biopsies when compared to control or PM samples. Interestingly, PKM2 expression in DM was as high as in the HCT116 carcinoma cell line (Additional file 3: Figure S2). A helpful biomarker that informs of the relative activity of energy provision pathways during development, differentiation and in cancer is the bioenergetic signature [10, 14, 15]. The bioenergetic signature is calculated by the ratio between the catalytic subunit of the H⁺-ATP synthase (β-F1-ATPase) relative to the expression of a glycolytic enzyme [14]. Remarkably, the β-F1-ATPase/PKM2 ratio was significantly diminished in DM providing an excellent bioenergetic marker in order to discriminate this disorder from controls or any other IM (Fig. 5a; Table 1). Likewise, unsupervised hierarchical clustering of the biopsies using the expression of 1, 2 or 3 proteins for aggregation further illustrated the potential of metabolic biomarkers to discriminate normal biopsies from DM (Fig. 5b; Table 1) and from sIBM (Fig. 5c; Table 1) with very high sensitivity and specificity (Fig. 5; Table 1).

To further explore the potential bed-side translation of PKM2 as a non-invasive biomarker of DM we investigated whether the inflammation and tissue remodeling that accompanies the disease could be reflected on the plasma levels of PKM2. To this aim we studied the expression of PKM2 in plasma samples of control and IM patients by ELISA and by a modified RPPA technique. Unfortunately, the results obtained revealed that IMs do not modify plasma PKM2 levels to a significant extent when compared to controls (Additional file 4: Figure S3).

In an effort to provide some basic information regarding the expression and regulation of PKM2 in muscle of DM patients we also studied the expression of the PKM1 isoform. We found that expression of PKM1 is not altered in the muscle of DM patients when compared to control samples (Fig. 6a). Due to the very large induction of PKM2 observed in muscle of DM patients (Figs. 3, 6a) the results support that rather than a switch in the expression of PK isoforms DM triggers the specific induction of PKM2. Likewise, analysis of the phosphorylation status of PKM2 in muscle of control and DM patients using the 105Tyr PKM2 antibody revealed, despite the large intra-group variability noted in the phosphorylation of the protein (Fig. 6a), no significant differences between the two groups of samples (Fig. 6a). This finding suggests that PKM2 induction in the muscle of DM patients accounts largely for the active form of the enzyme. Moreover, analysis of the subcellular localization of PKM2 in cryo-sections of muscle samples from control and DM patients by immunofluorescence microscopy revealed that the immunostaining was found in the cytoplasm preferentially clustered towards the sarcolemma and perinuclear region of the cells. We found no evidence for PKM2 translocation into the nucleus of both control and DM muscle cells (Fig. 6b).

To evaluate the potential activation of glycolysis in the muscle of DM patients as a result of the potential limitation of the oxidation of pyruvate in mitochondria we studied the expression of pyruvate dehydrogenase E1α (PDH) and of its less active phosphorylated form (p-PDH) in samples of control and DM patients (Fig. 6c). The results obtained support that glucose oxidation might be partially restricted in the muscle of DM patients when compared to controls because there is a sharp and significant increase in the expression of the less active p-PDH (Fig. 6c) in the absence of relevant changes in total PDH (Fig. 6c). Moreover, these changes were further supported by the parallel and significant concurrent increase in the expression of pyruvate dehydrogenase kinase (PDK1) in muscle samples of DM patients (Fig. 6c).

Diagnosis of IMs relies on histopathologic and immunopathologic examination of muscle biopsies by an experienced laboratory [2, 16]. Specific myositis-related autoantibodies may help diagnosis [17, 18]. Although the diagnosis of DM usually is not difficult based on clinical signs and symptoms, this is not the case with respect to PM and sIBM. The diagnosis of PM is often assessed by exclusion criteria [1], and in turn, usually more than one or two biopsies over time are required before reaching the diagnosis of sIBM. RPPA is a widely adopted high-throughput technique that allows quantitative protein profiling of cells or tissues that is especially indicated for the identification of biomarkers of diagnosis, prognosis and therapeutic response [6, 10, 19, 20]. In fact, RPPA have been recently approved for implementation in clinical trials [6]. Herein, we have interrogated a cohort of muscle biopsies from inflammatory myopathic patients using RPPA and nine validated mAbs against key proteins of energy metabolism to uncover new biomarkers that could help differential diagnosis of IMs. We report that the metabolic biomarkers studied are unable to distinguish control from PM biopsies suggesting the lack of a specific molecular fingerprint of the disease in agreement with its ambiguous clinical profile. In contrast, both DM and sIBM reveal significant differences in the expression of proteins of energy metabolism when compared to controls according to a well-defined clinical profile. Although age, type of muscle fibers and extent of cellular damage in the biopsies are likely to affect protein expression, the low dispersion of the values obtained for each marker suggests that these are not main factors contributing to the differences reported.

Consistent with previous reports [16] we observed an increased expression in β-actin in both DM and sIBM. β-actin overexpression in IMs is the likely consequence of muscle fiber regeneration, a criteria for the classification of IMs [16] also supported by their association with connective tissue disorders [21, 22]. Contrary to the increased expression of β-actin, sIBM biopsies revealed a diminished expression of the glycolytic markers GAPDH and LDHA when compared to controls. Consistently, the expression of GPD1, which forms part of the mitochondrial shuttle of the glycolytic generated NADH, was also diminished in sIBM. Moreover, the expression of PYGM, which is upstream of glycolysis, was also diminished in sIBM biopsies. These findings suggest that muscle fibers of sIBM experience a partial repression of glycolytic metabolism when compared to controls in agreement with recent proteomic findings in this regard [23].

In contrast to the findings in muscle of sIBM, the biopsies of DM patients showed an enhanced expression of the mitochondrial protein Hsp60 and IF1 and a very large increase in the glycolytic PKM2. The βF1-ATPase/LDHA ratio, which has been recently described as a potential biomarker of neuromuscular diseases [13], failed in the discrimination of IMs except for sIBM patients, which showed an increase of the ratio. However, the βF1-ATPase/PKM2 ratio offered a reliable indicator to discriminate DM from any other IM or from control biopsies. In fact, the power of PKM2 as a biomarker in IMs is well demonstrated by the high specificity and sensitivity in the stratification of patients affected with DM when compared to controls or any other group of patients affected with IMs. Overall, we support that the increased expression of PKM2 in DM confers to this protein a great value as biomarker for this particular type of IM, a finding that might benefit some patients by sparing an additional muscle biopsy [16]. Unfortunately, we found that PKM2 levels in plasma of DM patients do not provide a non-invasive biomarker of the disease.

It is generally agreed that cancer is more frequently associated with DM than with PM [2] and a clear increase of cancer related mortality has been reported in DM patients when compared to PM patients [1, 24‐26]. A distinctive feature of cancer cells is the enhanced aerobic glycolysis [14, 27]. Pyruvate kinase (PK) is the enzyme that catalyzes the final step in glycolysis, converting phosphoenolpyruvate (PEP) to pyruvate [28]. Four different isoforms of pyruvate kinase exist: type-R and type-L are generated by alternative splicing of the PKLR gene and are expressed in erythrocytes and in liver, respectively [29]. PKM1 is the isoform expressed in adult skeletal muscle while PKM2, which results from alternative splicing of the PKM gene is expressed exclusively in embryonic and proliferating tissues. Notably, PKM2 is allosterically regulated due to its possibility to switch from a dimeric low-active form to a tetrameric very high active form [30‐32]. In addition, phosphorylation of S37 and Y105 in PKM2 prevents the binding of the PKM2 cofactor fructose-1,6-bisphosphate, thus inhibiting the active tetrameric form of PKM2 which promotes aerobic glycolysis and tumor growth [33] (Fig. 5). Our findings stress the idea that DM triggers an increase in the expression of PKM2 rather than a switch in the expression of muscle isoforms. Likewise, the increased expression of PKM2 in muscle of DM patients is in its non-phosphorylated active state. Moreover, PKM2 also has a “non-metabolic” role in tumorigenesis since its translocation into the nucleus regulates gene transcription of several pathways involved in metabolic reprogramming, cell proliferation and cancer development [34‐37]. Although we did not observe any nuclear localization of PKM2 in muscle of DM patients, we cannot exclude this possibility (Fig. 7) because its nuclear translocation might represent a late event in the pro-oncogenic development of the disease.

Strikingly, and highly consistent with the increased cancer incidence observed in DM patients [1, 24‐26], we found an elevated expression of IF1 in DM biopsies. In fact, IF1 is highly overexpressed in most prevalent human carcinomas [12, 38]. Overexpression of IF1 interferes with oxidative phosphorylation by inhibiting the H⁺-ATP synthase promoting aerobic glycolysis and signaling, via reactive oxygen species (ROS), a pro-oncogenic phenotype (Fig. 7) that enhances proliferation, invasion and cell survival [12, 39, 40]. More recently, the overexpression of IF1 in human hepatocellular carcinomas, bladder and gastric cancers and in gliomas has provided a valuable biomarker of bad cancer prognosis [41‐44]. Mechanistically, the pro-oncogenic features of IF1 over-expression stem from its effects on favoring proliferation, cell death resistance, epithelial mesenchymal transition and angiogenesis [39‐41, 45]. Altogether, we suggest that the increased expression of PKM2 and IF1 could cooperate to promote a metabolic phenotype in DM that is prone for the onset of oncogenesis (Fig. 7). Consistent with this idea, our findings support that pyruvate oxidation in muscle of DM patients might be partially restricted when compared to controls by the diminished fraction of active pyruvate dehydrogenase present in mitochondria (Fig. 6c). We suggest that these findings might stimulate the development of future basic investigations and therapeutic approaches for the management of DM patients in order to prevent their higher cancer incidence.

Our study highlights the usefulness of combining RPPA and proteins of energy metabolism as a tool to identify novel biomarkers in rare myopathic diseases. We further illustrate the clinical relevance of PKM2 and IF1, two proteins that respectively act on glycolysis and oxidative phosphorylation, as novel biomarkers of dermatomyositis and potential metabolic drivers of the higher cancer incidence observed in this inflammatory myopathy.