Proteomic analysis reveals heat shock protein 70 has a key role in polycythemia Vera

Sixty-seven patients diagnosed with MPN (28 PV, 25 ET JAK2^V617F, and 14 ET JAK2^WT) were included in this study, in addition to 26 healthy donors. A diagnosis of MPN was based on the World Health Organization criteria 2001/2008, or the Polycythemia Vera Southern Study Group (according to the standard criteria at the moment of diagnosis). Mutational Screening for JAK2 V617F was performed using real time PCR on DNA from whole peripheral blood.

The study was approved by the 12 Octubre Hospital ethics committee and written informed consent was obtained from all patients, according to the Declaration of Helsinki. A flow diagram of the patients is shown in Figure 1.

Peripheral venous blood was collected in ethylenediaminetetraacetic acid or heparin lithium and processed immediately. Leukocytes, granulocytes, and mononuclear cells were isolated by Ficoll-Paque density gradient centrifugation (Pharmacia, Freiburg, Germany). Erythrocytes were eliminated using a commercial red blood cell lysis buffer (Roche Applied Sciences, Manheim, Germany), with more than 90% granulocytes. Lymphocyte contamination was assessed in five samples by flow cytometry, and was less than 2% of the total cell count.

Protein cytosolic fractions of granulocytes were extracted using Proteoextract subcellular proteome extraction (Calbiochem, Darmstaz, Germany).

To ensure equal protein loading on both 2D-PAGE and, the protein concentration was determined using a non-interfering assay (RC-DC Protein Assay Kit; BIO-RAD, NY, USA).

Protein cytosolic fractions from peripheral blood granulocytes from 10 ET, 10 PV, and 10 healthy donors as controls, were initially pooled for two-dimensional difference gel electrophoresis ( 2D-DIGE). We performed three analysis, ET versus PV pools, ET versus control pools, and PV versus control pools by duplicate (6 gels), changing labels with different dye (Cy3 or Cy5), for each pool in each analysis. Patients and clinical data of the 2D-DIGE study are presented in Table 1.

The PV and ET cytosolic protein pools (20 micrograms each) were minimally labeled with 160 pmol of the N-hydroxysuccinimide esters of Cy3 or Cy5 fluorescent cyanine dyes (GE Healthcare, Piscataway, NJ, USA). An internal standard pool was generated by mixing equal amounts of proteins obtained from all the samples and labeled with 160 pmol of Cy2 dye. PV and ET labeled protein pools and the internal standard protein samples, were combined in pairs, diluted In rehydration buffer (7 M urea, 2 M thiourea, 4% CHAPS, 10 mM DTT, 0.5% carrier ampholites pH 3–11 NL), and applied by cup loading to 18 cm IPG strips pH 3–11 NL (GE Healthcare, Piscataway, NJ, USA), previously rehydrated with 340 μl of rehydration buffer containing 1.2% DeStreak. The first dimension was run at 0.05 mA/IPG strip in an IPGphor IEF System (GE Healthcare, Piscataway, NJ, USA) following a voltage increase until 43000 Vhrs were reached. Strips were then reduced and alkylated in the dark in SDS equilibration buffer (75 mM Tris pH 8,8, 6 M urea, 30% (v/v) glycerol, 2% (w/v) SDS, and traces of bromophenol blue) containing 1% (w/v) DTT or 4% (w/v) iodoacetamide. Finally, the proteins were separated using 12.5% tris-glycine gels in an Ettan Dalt Six device (GE Healthcare, Piscataway, NJ, USA) at 20°C.

Following electrophoresis, the 2D-gels were scanned in a Typhoon 9400 scanner (GE Healthcare, Piscataway, NJ, USA) at 100 μm resolution, and with the appropriate wavelengths and filters for Cy2, Cy3 and Cy5 dyes. Relative protein quantification was performed using DeCyder software v7.0. Background subtraction, quantification, and normalization were automatically applied with low experimental variation. Differences were calculated as average ratios for each spot, and average ratios > or = 1.5 or < or = −1.5. The student’s t-test was used to compare average ratios for each spot between PV and ET samples. P values less than 0.05 were considered significant. Individual coordinates corresponding to the spots of interest were automatically calculated and automatic spot pick-up was carried out using a Spot Picking Robot (GE Healthcare, Piscataway, NJ, USA).

For MALDI-TOF/TOF analysis, samples were automatically acquired in an ABi 4800 MALDI TOF/TOF mass spectrometer (AB Sciex, Framingham, MA, USA) in positive ion reflector mode (ion acceleration voltage was 25 kV for MS acquisition and 1 kV for MSMS). PMF and MSMS fragment ion spectra were smoothed, corrected to zero baseline and internally calibrated with the mass signals of trypsin autolysis ions to reach a typical mass measurement accuracy of < 25 ppm. Known trypsin and keratin mass signals, as well as potential sodium and potassium adducts (+21 Da and +39 Da) were removed from the peak list. To submit the combined PMF and MS/MS data to MASCOT software v.2.1 (Matrix Science, London, UK), GPS Explorer v4.9 was used, searching in the non-redundant UniProt/SwissProt protein database (taxonomy: Homo sapiens).

Twelve PV, 10 ET JAK2 positive, 13 ET JAK2 negative and 11 controls from formalin fixed and paraffin-embedded bone marrow biopsies were collected. Non haematological diseases patients or patients with secondary thrombocytosis and/or erythrocytosis, both with free-infiltrate bone marrow were used as negative MPN controls. They were used to validate the DIGE/MS results. Patients and clinical data of the IHC study are presented in Table 2.

We performed immunohistochemical (IHC) staining in four micron-thick tissue sections from all cohorts for HSP70 (clone MAB1663 anti-h/m/r/rHSP70, 1/8000 dilution; R&D, Minneapolis, MN, USA), SERPINB1 (clone HPA018871 anti-SERPINB1, 1/12000 dilution, SIGMA, Steinheim, Germany), and LTA4H (clone HPA00399 anti-LTA4H, 1/50 dilution, SIGMA, Steinheim, Germany). After incubation, immunodetection was done with the DAKO EnVision visualization method (Dako, Glostrup, Denmark), with diaminobenzidine chromogen as the substrate. Sections were counter stained with hematoxylin. Immunostaining was evaluated by two different pathologists, using granulocyte percent and stain intensity criteria. Only distinct and intense cytoplasmic staining was considered positive.

Colony assays were performed using Methocult TM GF_H4535 (StemCell Technologies, Vancouver, BC, Canada, http://​www.​stemcell.​com).

In brief, a 0.5 mL cell suspension, containing 5×10⁵ peripheral blood mononuclear cells from four PV and four ET patients, and three healthy donors as controls, were each mixed in 500 μl of methylcellulose solution consisting of methocult, 20 ng/mL interleukin 3 (IL-3), and 50 ng/mL stem-cell factor (SCF) and 3 U/mL erythropoietin (all fromStemCell Technologies, Vancouver, BC, Canada) in 3.5-cm culture dishes. We cultured the cells with (3 U/mL erythropoietin) and without EPO (endogenous growth).

Additionally, the burst formation unit-erythroid (BFU-E) assay was performed with 2×10³ CD34+ bone marrow cells per well from two PV, two ET, and two cord blood samples as controls.

HSP70 was inhibited by 100 μM, 50 μM, and 10 μM KNK437 (Calbiochem, Darmstadt, Germany). For experimental controls we excluded KNK437, and all samples were assayed in duplicate. After 2 weeks, the colonies were counted. Colony morphology was also observed using an inverted light microscope.

Cells from the BFU-E were extracted, washed and resuspended in 10 mL PBS. Ten microlitres aliquots of cells were used to test viability using trypan blue (1:1). Cells were analyzed by flow cytometry after the addition of 10 μl (5 μl of APC antibodies) of markers; CD71-FITC, CD45-PerCP, CD44-PE, annexin V-APC, CD41a-FITC, and CD34-APC (BD Biosciences Europe, Oxford, UK), incubated for 30 minutes at 4°C, and washed with PBS or binding buffer 1X (BD Biosciences Europe, Oxford, UK) (in the case of annexin) before analysis. Samples were analyzed using a flow cytometer FACSCalibur (Beckman Coulter, Fullerton, CA). Cell suspensions with IgG isotype control antibodies (BD Biosciences Europe, Oxford, UK) were used as negative controls.

DNA from BFU-E cultured cells was extracted using the Maxwell 16 SEV automated extraction system (Promega, Manheim, Germany).

Protein from cells was extracted with the CBA extraction kit (BD Biosciences Europe, Oxford, UK) according to the manufacturer’s instructions but with the addition of phosphatase and protease inhibitors. CBA technology is a set of microspheres (beads) with different sizes and fluorescent intensities and each bead binds a specific protein (similar to the ELISA technique). Each CBA assay includes seven principal steps: preparation of beads, preparation of Phycoerythrin reagent, setting standard curve, preparation of samples, cytometer calibration, acquisition of samples, and file analysis. We analyzed four phosphorylated, and their respective native, proteins: AKT, p-ATK, P38, p-P38, MEK, p-MEK, STAT1, and p-STAT1. These proteins represent the most important pathways downstream of the JAK2 signaling pathway. Protein concentrations were analyzed using concentration ratios of phosphoproteins normalized with non-phosphoproteins and total protein.

To confirm the above CBA results, we analyzed JAK-STAT and MAPK activation after KNK437 treatment, a specific pharmacological HSP70 inhibitor, in HEL and Ba/F3 JAK2 V617F EPOR cell lines that were kindly transferred by Dr A. Quintas-Cardama for MD Anderson, and cultured as previously described [25]. We used these cell lines as MPN model due to its JAK2 mutational status. HEL cells were obtained from the DSMZ collection (cells were obtained on 01/08/2008; JAK2 V617F mutation status was tested by QRT-PCR on 01/06/2012, finding JAK2 V617F mutation in homocigosis) and cultured in RPMI-1640 medium containing 10% fetal calf serum, with L-glutamine and NaHCO₃ in a humidified 5% CO₂ atmosphere. For the inhibition assay, subconfluent cells in 9.5 cm² wells (P6) were treated with KNK437 (50 μM) for 24 hours. Results were analyzed with the trypan blue viability test. Cells were washed twice in PBS and protein was extracted with the Cytobuster protein extraction reagent (Novagen, EMD Bioscences Inc., Madision, WI). The protein concentration was determined using a non-interfering assay (RC-DC Protein Assay Kit; BIO-RAD, NY, USA) and Western Blot was performed using rabbit anti-actin primary antibody (protein control), anti-p-MEK (ser217/221), anti-ERK, anti-p-ERK (thr202/tyr204), anti-p-P38 (thr180/tyr182), anti-JAK2, anti-p-JAK2 (tyr1007/1008) , anti-STAT5, anti-p-STAT5 (ser456) (Cell signaling, Beverly, MA, USA), and mouse anti-HSP90 and anti-HSP70 (R&D, Minneapolis, MN, USA). The membranes were then incubated with the respective secondary antibodies for 1 h and antigens were detected by using the ECL Advance Western Blotting Detection Kit (GE Healthcare, Piscataway, NJ, USA).

In order to confirm the specificity of KNK437 over HSP70, we analyze the effect of the interference on HSP70 molecule through a specific siRNA. HEL cell line was transfected using the Amansa Electronucleofector 2b and Cell Line Nucleofector kit V (Lonza, NJ, USA). Anti HSP70 siRNA Trilencer 27 was acquired from Origene (Rockvile, MD, USA). Cells were incubated 8 h. Pmax gfp (Lonza USA, NJ) was used as a fluorescent control, showing a transfection efficacy greater than 80%.

DIGE and MS were used to identify differences in the whole cytosolic proteome between PV and ET groups. Figure 2A, show three representative spots from the proteomic analyses of samples from ET and PV patients. We found 112 spots representing proteins with differential expression between both diseases. Identification of the spots yielded 65 proteins. Three proteins were especially interesting in the context of our model and selected for further studies by doing a literature search on their biological function. These three differentially expressed proteins includes LTA4H, SERPINB1 and HSP70 (Figure 2A). Of note, HSP70 is a chaperone related to GATA-1 and erythroid differentiation. Most of the other spots corresponded to a large group of proteins implicated in metabolic and biochemical processes, for example, glycogen phosphorylase, pyruvate kinase, and lactotransferrin. Healthy donors also showed differences when compared with PV samples. There were 174 spots and 19 proteins identified (HSP70 included). Samples from controls and ET showed differences in 97 spots, and six proteins were identified. Most of the proteins identified were implicated in metabolic and biochemical pathways, similarly to those observed when ET and PV were compared. A full list of the differentially expressed proteins is summarized in Additional file 1: Table S1, Additional file 2: Table S2 and Additional file 3: Table S3.

Bone marrow IHC analysis of HSP70, SERPINB1, and LTA4H was performed to confirm and identify the expression pattern found by 2D-DIGE/MS. Twelve bone marrow biopsies from PV patients, 10 ET JAK2 V617F positive, 13 ET JAK2 V617F negative, and 11 controls were studied. HSP70 was markedly over-expressed in PV bone marrow (median of 80% positive granulocytes, range 2-95%, SD 0.35%) vs. ET (median of 23% positive granulocytes, range 1-95%, SD 34.25%) (Figure 2B-C). A full list of the HSP70 expression sample per sample is summarized in Additional file 4: Table S4.

SERPINB1 was expressed in the nucleus and cytoplasm of the granulocytes. We found a differential expression pattern between the bone marrow of controls (median of 20% positive granulocytes, range 2-99%, SD 39.05%) and PV (median of 95% positive granulocytes, range 65-99%, SD 16.72%), ET JAK2 V617F (median of 67% positive granulocytes, range 5-99%, SD 36.04%), and ET JAK2 wild-type (median of 98% positive granulocytes, range 3-98%, SD 38.31%).

LTA4H also was expressed in granulocytes and in a small percent of the megakaryocytic lineages. It was expressed in 90-100% of the bone marrow studied and in 90-100% of the granulocytes, and no differences were observed between the different groups (data not shown).

BFU-E cultures of CD34+ cells with KNK437 showed a decrease of colony formation and erythroid precursor viability. This KNK437-mediated decrease of viability reached an IC₅₀ of 20.05 μM in PV samples. Erythroid precursor cell viability in cord blood samples and ET patient cells was higher (IC₅₀ = 38.87 μM) than in the PV patients (Figure 3A-B). KNK437 also decreased cell viability in the HEL and Ba/F3 JAK2 V617F cell lines (Figure 3C). However, statistical significance among the groups was not found.

Flow cytometry results of BFU-E colonies showed differences in apoptosis among the erythroid population (that express CD71) in untreated PV cultures (10% annexin positive cells) vs. treated (KNK437, 50 μM) cultures (34% annexin positive cells). However, the same differences were not seen when treated ET samples (KNK437, 50 μM) (21% annexin positive cells) were compared with untreated cells (10% annexin positive cells). Figure 4A-B shows the flow cytometry results for the CD34+ BFU-E cultures.

CBA analysis showed an important decrease of phospho-STAT1 in PV samples patients (ratio of pSTAT1 in PV without treatment: 29.2; ratio of pSTAT1 in PV samples with KNK437, 50 μM: 21.55), however, we found no significant differences in phospho-STAT1 with and without KNK437 treatment in ET patients. We define ratios as concentration ratios of phosphoproteins normalized with non-phosphoproteins as total protein numeric value. Additionally, phospho-MEK showed under-expression after KNK437 treatment, and this was more pronounced in samples from PV patients (ratio in PV without treatment: 63.17; ratio in PV with KNK437, 50 μM: 23.58) vs. ET patients (ratio in ET without treatment: 66.19; ratio in ET with KNK437, 50 μM: 59.82). Moreover, the other MAPK phospho-protein, phospho-p38, was differentially expressed with and without KNK437 treatment in samples from PV patients (ratio in PV without treatment: 7.30; ratio in PV with KNK437, 50 μM: 4.18), but was unchanged in ET patients. Phospho-AKT showed no decrease with treatment. A full list of the proteins and phospho-proteins expression sample per sample is summarized in Additional file 5: Table S5.

To confirm the molecular mechanism of the HSP70 inhibitor in the JAK2/STAT and MAPK pathways, we performed Western blot on HEL and Ba/F3 JAK2 V617F cell lines proteomes, with and without KNK437 (10-50-100 μM) treatment (Figure 3D). This showed a reduction of the phospho-JAK2 and phospho-STAT5 protein with treatment (Figure 3E and F), but no reduction of phospho-ERK and phospho-p38. ImageJ quantification confirmed these results and showed a 50% reduction in the expression of phospho-JAK2 in the HEL cell line following treatment with KNK437. Additionally, HSP70, HSP90 and detection by Western Blot showed a slight decrease of HSP70 expression after KNK437 treatment, but no significant difference in HSP90 expression in the HEL cell line, with and without KNK437 treatment (Figure 3D). KNK437 decreased the activation of JAK2 as well as its expression. This decrease in JAK2 expression resulted in the inhibition of leading proliferative pathways related to JAK2 (Figure 3E) GATA1 also showed no differential expression with the HSP70 inhibitor treatment (Additional file 6: Figure S1). Similarly to the primary BFU-E, incubation with the HSP70 inhibitor KNK437 (50 μM) in HEL and Ba/F3 JAK2 V617F caused a reduction of 20-50% in the cell viability (Figure 3C).

In order to validate the KNK437 inhibition on HSP70, and check the specificity of this treatment, additional HSP70 interference was performed with specific a siRNA (Figure 3F). The results showed a proper interference, decreasing the protein levels of HSP70, but not HSP90. Besides, HSP70 interference assay produces the decrease of the expression of JAK2, and the inhibition of JAK-STAT signaling due to the decrease of phospho-STAT5.