AbstractMesenchymal stromal cells (MSC) support acute myeloid leukemia (AML) cell survival in the bone marrow (BM) microenvironment. Protein expression profiles of AML-derived MSC are unknown. Reverse phase protein array analysis was performed to compare expression of 151 proteins from AML-MSC (n=106) with MSC from healthy donors (n=71). Protein expression differed significantly between the two groups with 19 proteins over-expressed in leukemia stromal cells and 9 over-expressed in normal stromal cells. Unbiased hierarchical clustering analysis of the samples using these 28 proteins revealed three protein constellations whose variation in expression defined four MSC protein expression signatures: Class 1, Class 2, Class 3, and Class 4. These cell populations appear to have clinical relevance. Specifically, patients with Class 3 cells have longer survival and remission duration compared to other groups. Comparison of leukemia MSC at first diagnosis with those obtained at salvage (i.e. relapse/refractory) showed differential expression of 9 proteins reflecting a shift toward osteogenic differentiation. Leukemia MSC are more senescent compared to their normal counterparts, possibly due to the overexpressed p53/p21 axis as confirmed by high β-galactosidase staining. In addition, overexpression of BCL-XL in leukemia MSC might give survival advantage under conditions of senescence or stress and overexpressed galectin-3 exerts profound immunosuppression. Together, our findings suggest that the identification of specific populations of MSC in AML patients may be an important determinant of therapeutic response.
There is growing evidence to support the importance of the leukemia bone marrow (BM) niche in the process of acute myeloid leukemia (AML) chemoresistance.21 Hence, optimal therapeutic strategies should also address neighboring cells in the tumor microenvironment. The critical support cells in the leukemia BM microenvironment are mesenchymal stromal cells (MSC).83 Depending on the type, MSC can act either to support or suppress tumors.1584 Our group and others have found that MSC support leukemia cell survival by diverse mechanisms that include secretion of cytokines and chemokines, activation of survival signaling in tumor cells, and blocking immune surveillance by suppressing natural killer (NK) and T cells.1352
Mesenchymal stromal cells are essential for human hematopoiesis, particularly as a source of SDF-1, which regulates homing, proliferation, and differentiation.18161096 Moreover, studies from our group and others have demonstrated that MSC protect leukemia cells from 0chemotherapy.23196 We have recently found that there is reciprocal activation of NFkB signaling between MSC and AML and acute lymphoblastic leukemia (ALL) cells that likely contribute to the effectiveness of the microenvironment to protect malignant cells.7 Medyouf et al. recently demonstrated that blast cells from myelodysplastic syndrome (MDS) patients induce changes in MSC reflecting reprograming of the stromal cells.24 MSC may also influence hematopoietic precursors to promote leukemogenesis as evidenced by the development of AML and MDS in mice where the MSC osteo-progenitors were engineered to lack Dicer, a key regulator of microRNA (miR) processing.2 Furthermore, a recent study from Zhao et al. reported that p21 could be critical for inducing senescence in MSC from MDS patients with concomitant induction of interleukin- 6 (IL6) and transforming growth factor β (TGF-β).25 This study is consistent with findings that support the role of Dicer in regulating MSC biology and also establish a possible mechanism of aberrant survival functions in malignant MSC that may be associated with p21 and senescence. The ability of malignant MSC to withstand senescence may depend on the expression of the anti-apoptotic molecule BCL-XL.2726
The cellular composition of stromal cells in a cancer microenvironment, such as the leukemic BM niche, is likely markedly different from that of the normal BM. We, therefore, set out to study the protein expression and activation in leukemic MSC (AML-MSC) and compared and contracted these to normal MSC (NL-MSC) to determine if and how they are functionally different. Reverse phase protein array analysis (RPPA, pioneered in our laboratory)3228 was used to examine expression of 151 proteins in MSC derived from AML BM (n=106) with those derived from healthy donors (n=71). The results presented here identify 28 that were differentially expressed between the two. Importantly, the 28 proteins identified as differentially expressed in the AML versus normal MSC could be grouped into four protein constellation (PC) expression signatures with different biological properties and clinical implications regarding patient response to therapy.
Bone marrow was obtained from AML patients (n=106) undergoing diagnostic BM aspiration and from healthy donors (n=71) who were undergoing BM harvest for use in allogeneic BM transplantation. Samples were acquired in accordance with the regulations and protocols approved by the Investigational Review Board of MD Anderson Cancer Center. Informed consent was obtained in accordance with the Declaration of Helsinki. Samples were analyzed under an Institutional Review Board-approved laboratory protocol. Patients’ characteristics are presented in Table 1. Details of isolation of MSC are available in the Online Supplementary Methods.
Proteomic profiling was carried out on MSC samples from patients with AML and healthy donors using RPPA. The RPPA method and sample validation technique are described fully elsewhere.3228 Antibodies against 151 proteins were used for analysis. (A list and the source of the antibodies and the concentrations utilized is provided in the Online Supplementary Table S1). The sources of antibodies have been been reported previously.30 An IgG subtype-specific secondary antibody was used to amplify the signal, and finally a stable dye was precipitated. The stained slides were analyzed using the Microvigene software (version 3.0, Vigene Tech, Carlise, MA, USA) to produce quantified data. Statistical analyses are described in the Online Supplementary Methods.
Cell senescence assessment
Microscopy assessment of β-galactosidase staining was used to detect cell senescence using a detection kit from Cell Signal Technology (Boston, MA, USA). Early passage cells (passage 2) were imaged using a Nikon Coolpix 950 camera attached to a Nikon TMS light microscope (Nikon Instruments Inc.). AML-MSC (n=4) and NL-MSC (n=5) were lysed in kit buffer. Measurement of β-galactosidase was performed using an in vitro fluorometric assay with fluorescein di-β-D-galactopyranoside (FDG) as substrate. Incubation time was 2 hours (h). Fluorescence was measured using an Optima Fluorometer (Durham, NC, USA). Activity is presented as fluorescence units/1000 cells/minute.
String software (String 10.1; available from: http://string-db.org)33 was used to determine protein associations. Pathway analysis to identify canonical pathways, upstream regulators, and protein networks was performed using Ingenuity Pathway software (Qiagen).
Proteins are differentially expressed in AML versus healthy MSC
We have routinely utilized RPPA to analyze protein expression from clinical samples from many hematologic malignancies.3228 We examined protein expression in blasts from newly diagnosed AML patients (n=85), CD34 cells from normal donors (n=10), MSC from healthy donors (n=71), and MSC from newly diagnosed AML patients (n=54). Both normal MSC and AML-MSC expressed MSC defining lineage markers CD73, CD90, and CD 105 as determined by flow cytometry (Online Supplementary Figure S1). MSC from salvage samples (i.e. relapse/refractory) were also studied (n=46). The RPPA was probed with 151 antibodies targeting 119 different proteins (114 targeting total protein with 32 paired antibodies targeting phosphoepitopes on 26 proteins, and 5 with only a phosphoepitope but not total protein epitope) covering a wide variety of cellular functions and pathways (Online Supplementary Table S1). Protein expression in AML-MSC, NL-MSC, AML blasts and normal CD34 cells was compared using principle component analysis (Online Supplementary Figure S1) and unbiased hierarchical clustering (Online Supplementary Figure S2B). NL-MSC and AML- MSC formed a cluster distinct from AML blasts and NL- CD34 cells with the vast majority of the 151 proteins tested showing statistically significant differential expression (143 of 151, P=0.01; 124 of 151, P≤10) between MSC and the blast/CD34 cells. This unsurprising observation is consistent with a previous report that gene expression profiles are distinct between blood cells and MSC.34 Principal component analysis (PCA) also shows that protein expression in NL-CD34 cells is distinct from that of AML blasts. These findings were identical to those observed when the analysis was restricted to samples from newly diagnosed patients alone.
Next we investigated whether protein expression in AML-MSC was different from that of NL-MSC. In PCA, the NL-MSC occupied a distinct space from that of the AML-MSC (Figure 1A). Unbiased hierarchical clustering comparing AML-MSC and NL-MSC revealed differential expression of 28 of those proteins (P<0.001; Q=0.0059). The Q-value, a measure of the false discovery rate determined by a β-uniform mixture model highlights that these differences are very unlikely to be random and suggest that there are significant differences in the protein expression patterns of AML-MSC relative to those of NL-MSC. The proteins clustered into three PCs (from top to bottom in Figure 1B). Nineteen of these proteins had generally higher expression in AML-MSC, including P1 [STAT1, p- PDK1 (S241)], CCND1, CDKN1A (p21), ITGA2, PARP1, PPP2R2A/B/C/D, the PP2A B regulatory subunit family B55, BAK1, CSNK2A1, CDK4, GSK3A/B) and PC2 [STAT5A/B, BCL2L1 (BCL-XL)], DIABLO, TP53 (p53), NOTCH 1 (cleaved 1744), SPP1, p-EGFR (Y992), and ERBB2). Expression of 17 of the 19 proteins was validated by immunoblot analysis (Online Supplementary Figure S3). Although expression of EGFR and ERBB2 expression could not be confirmed by western blot analysis, this may reflect the enhanced sensitivity of RPPA over standard immunoblot technology. The remaining nine proteins in PC3 were elevated in healthy donor MSC compared to AML-MSC: SMAD1, CREB1.p133 STMN1, SIRT1, CREB1 SMAD4, p-Foxo1/3 (S32), HSP90AA1/B1, and EIF2S1.
Based on various combinations of proteins expressed within these three groups, we observed that normal or leukemic MSC samples clustered into four distinct PCs. The first (aqua in Figure 1A; top row of annotations above the heat map in Figure 1B), was comprised predominantly of NL-MSC (42 of 46 members), characterized by lower expression of proteins in constellations 1 and 2, and higher expression of constellation 3 proteins. This was considered to represent the protein expression pattern of NL- MSC. AML cases in this signature are hereafter called Class 2.
A second signature (blue in Figure 1A and B) was comprised predominantly of AML-MSC (62 of 72 members), and was characterized by the opposite protein expression of the NL-MSC signature, with high expression of PC1 and 2 proteins and lower expression of PC3 proteins. This signature is called Class 4. Two groups with expression signatures between that of the Class 2 MSC and Class 4 MSC were also recognized. One signature (yellow in Figure 1A and top row of Figure 1B), comprised of 12 AML-MSC and 11 NL-MSC, mirrored the expression of Class 2 for protein constellation 1, but that of Class 4 for PC2 and 3. AML cases in this group are hereafter referred to as Class 1. The final signature (orange in Figure 1A and top row of Figure 1B), consisting of 16 AML-MSC and 8 NL-MSC, mirrored the expression of Class 4 for PC1 and 3 but had low expression of PC2, similar to Class 2. AML cases from this signature are referred to as Class 3.
Among the AML cases, there were very few clinical or laboratory features that showed a statistically significant correlation with the four MSC PCs. Specifically there was no association for continuous variables: age, percent BM or blood blasts, hemoglobin, platelet count, serum lactate dehydrogenase, albumin, creatinine, bilirubin fibrinogen, CD7, CD10, CD13, CD14, CD20, CD33 or CD34, or discrete variables disease status (new vs. salvage), race, French-American-British (FAB) Classification, performance status, history of an antecedent hematologic disorder, presence of an FLT3-ITD or presence of an NPM1 mutation (Table 1). However, a few notable associations between MSC protein signature and clinical features were observed. There was a statistical difference in presence of FLT3 D835 mutation (P=0.05). While no Class 1 or Class 2 samples contained FLT3 D835 mutation, 9.4% of Class 4 MSC and 5.6% of Class 3 MSC had the mutation (Table 1). Interestingly, there was a statistical difference in MSC population types between men and women (P=0.03) (Table 1). Men had a higher percentage of Class 4 and Class 3 MSC while women had a higher percentage of Class 1 and Class 2 MSC. At present it is unclear if sex influences MSC biology in the leukemic niche. Sex-specific effects in the microenvironment in leukemia is not unprecedented as integrin-mediated adhesion-triggered activation of GSK3B occurs exclusively in leukemia progenitor cells derived from female AML patients.35 Total GSK3 and ITGA2 are present in protein constellation 1 and the proteins in this constellation are elevated in Class 3 and Class 4 while they are lower in Class 1 and Class 2. Perhaps the integrin/GSK3 axis is also important in AML- MSC.
Finally, there were statistically significant biases with different cytogenetic groups associating with different MSC protein signatures (P=0.026) (Table 1). Specifically, although few patients had favorable cytogenetics in this dataset, these were found exclusively in the Class 1 and Class 2 signatures. Conversely, patients with unfavorable cytogenetics were not observed in the Class 2 signature, but they were found in relatively equal proportions in the three AML specific signatures (i.e. Classes 1, 3, and 4).
Having determined that leukemic MSC divide AML patients into discrete protein signatures, we next examined whether signature membership affected outcomes. There was no significant difference in median overall survival between the four MSC populations, although the median survival of 105 weeks (wk) in the Class 3 signature was longer than the other three groups at 25, 48 and 58 wk (Figure 2A); this suggested that patients with Class 3 MSC may have a better disease prognosis. Despite the small sample size, relapse rates among the newly diagnosed cases were different, with 4 of 5 Class 1 MSC patients and 13 of 16 Class 4 MSC patients relapsing compared to only 2 of 7 Class 3 MSC cases (P=0.076). Furthermore the median remission duration was different between patients with Class 3 MSC and the three other signatures (101 wk for Class 3 MSC vs. 42.2 wk for Class 4 MSC; P=0.05) (Figure 2B). Other comparisons were not made due to the small sample size. These findings suggest that MSC can influence patient survival, although at present the mechanism is unknown.
Effect of age on protein expression in NL-MSC and AML-MSC
As AML is a disease of the elderly, many of the AML- MSC samples were obtained from individuals aged over 60 years (y). Of the 106 AML-MSC samples, 50 samples were from individuals under 60 y. Of the 71 MSC samples from healthy donors, 68 samples were from individuals under 60 y. To assess the impact of age on the differences in protein expression between the AML-MSC and NLMSC groups, the protein expression of 24 of the differentially-expressed proteins was reassessed in age-matched sets. The three groups were: a) under 30 y (AML n=13; NL n=37); b) 30–49 y (AML n=17; NL n=25); and c) 50 to 59 y (AML n=20; NL n=6). As shown in Online Supplementary Table S2 and Online Supplementary Figure S4A, of the 19 proteins elevated in AML-MSC compared to NL-MSC, only two (i.e. BCL-XL and PPP2R2A/B/C/D) were significantly higher in all age groups. This suggests that increased expression of these 2 proteins is dependent on the disease state and not on age. Six other proteins including p53 and CDKN1A (p21) were also elevated in two of the three age groups suggesting expression of these proteins is also dependent on the disease state rather than on age (Online Supplementary Table S2).
Analysis of p53 and CDKN1A (p21) protein expression was performed by western blot analysis on age-matched AML and NL samples (age 40–49 y; n=3 for both groups). Protein expression of both p53 and CDKN1A (p21) was at least 2- to 3-fold higher in AML-MSC compared to the healthy donor samples (Online Supplementary Figure 4B). These findings suggest that the disease state dictates expression of at least some of the identified proteins in MSC.
Evidence for dysregulated signaling in AML-MSC
The observed differences in protein expression between AML-MSC and NL-MSC suggest that pathway utilization may be dysregulated or non-canonical in the AML-MSC. To look for evidence of abnormal utilization, we searched for statistically significant (R>0.2; P< 0.0001) protein-protein correlations that were reversed in the direction of correlation in the AML-MSC compared to the pattern in the NL-MSC. A representative analysis of STAT5 expression with the other proteins in NL-MSC revealed that this transcription factor is negatively correlated with hnRPK and positively correlated with STAT1 (Online Supplementary Figure S5B). However, in AML-MSC the correlations are reversed (Online Supplementary Figure S5B).
A list of all protein correlations that are reversed in AML-MSC compared to NL-MSC is provided in Table 2. Of note, there are significant changes in protein correlations involving β-catenin. As shown in Table 2, BCL2, BCL2L11, FOXO1.3 p24.p32, FOXO3, KIT, ITGAL, PSMD9 p10, RPS6KB1 p389, SFN, SRC, and SRC p416 are negatively correlated with β-catenin in NL-MSC but are positively correlated with the protein in AML-MSC. ARC (NOL3), HSP90AA1/B1, MAPK9, PTGS2, and YWHAZ were positively correlated with β-catenin in NL-MSC but are negatively correlated with the protein in AML-MSC. Ingenuity Pathway Analysis (IPA) was performed using software on the proteins identified as differentially correlated with β-catenin in the MSC sets and β-catenin itself (i.e. CTNNB1; BCL2; BCL2L11; FOXO1; FOXO3; KIT; ITGAL; PSMD9; RPS6KB1; SFN; SRC; NOL3; HSP90AA1; HSP90B1; MAPK9; PTGS2; YWHAZ). IPA revealed these proteins were highly associated with PI3K/AKT signaling (top canonical pathway; P=7.35E-19) (Online Supplementary Figure S5). IPA also identified the top upstream regulator of this set of proteins as p53 (P=1.04E-11).
Proteins differentially expressed in AML-MSC share interactomes
To assess the relationship among the proteins identified in the RPPA analysis, protein association network analysis was performed using STRING 10.533 on proteins identified as significantly different in the AML-MSC and NL- MSC (Figure 1B). Blalock et al. used a previous version of String software to map the nuclear interactome.36 In cases where a family of proteins was identified (e.g. PPP2R2 set and HSP90 set), a representative member was included in the analysis. With the exception of PDK-1 all proteins are interconnected at least through one association (Figure 3). This finding suggests that there is an interconnection between the various proteins that are distinctly expressed between the NL-MSC and AML-MSC groups.
Pathway analysis suggests NL-MSC have prominent adipogenic signaling while all AML-MSC populations have prominent PI3K/AKT signaling
Pathway analysis was performed using IPA. Two separate data sets were created: one with proteins from PC3 (elevated in normal MSC) and one with proteins from PC1 and 2 (elevated in AML-MSC). Proteins in group 3 are associated with adipogenesis (ninth top canonical pathway; P=2.19E-05) (Online Supplementy Figure S6). PC3 proteins are elevated in MSC that were presumably normal but reduced in AML-MSC suggesting differences in differentiation potential of MSC between NL-MSC and AML- MSC. Of the 7 proteins identified, SIRT1, FOXO1, and SMAD1 each can activate PPARβ which is a critical regulator of adipocyte differentiation.3937 PC3 also displayed the strongest association with AMPK signaling (P=6.84E-07). The top upstream regulator identified in the set of PC3 proteins was PDGFB (P=1.01E-06). PI3K/AKT pathway was highly associated with PC1 and 2 proteins, which are elevated in AML-MSC (top canonical pathways; P=7.16E-17) (Online Supplementary Figure S7). AKT was identified as one of the top three upstream regulators (P=1.22E-14), suggesting that signaling mediated by this survival kinase is prominent in AML-MSC.
AML-MSC are senescent compared to NL-MSC
The p21 protein appears to be critical for senescence in myelodysplastic syndrome (MDS)-MSC22 and AML- MSC similar to MDS-MSC have elevated p21 (Figure 1B). This finding suggests that AML-MSC might also be more senescent than NL-MSC, as was the case in MDS-MSC.25 Senescence was observed in normal donor MSC- and AML-derived MSC using β-galactosidase staining. AML- MSC were more senescent than MSC derived from healthy donors in this representative example (Figure 4A). To account for age effects on senescence, MSC were taken from donors under 58 y. Average age of the AML patients (n=4) was 52 y and the average age of normal donors (n=5) was 47 y. Also, MSC of similar cell passage (passage 2 or 3) were used, so effects of cell passage were unlikely. As shown in Figure 4B, β-galactosidase activity was significantly higher (almost 2-fold) in AML-MSC compared to NL-MSC (P=0.032). These findings suggest that AML- MSC tend toward senescence.
Therapy alters proteins expression in AML-MSC
Protein expression in AML-MSC might change between diagnosis and relapse, perhaps as a result of relapse, or in response to acquired changes in AML blasts. To determine if AML-MSC protein expression was changing between diagnosis and relapse we compared expression between the samples obtained at first diagnosis (n=53) to those of AML-MSC collected from primary refractory or relapsed patients (n=54). Nine proteins are differentially expressed between the two MSC groups (Figure 5). Phosphorylated β-catenin, phosphorylated RPS6, and galectin-3 are expressed at higher levels in MSC in the salvage set. SMAD6, TCF4, LYN, integrin-β3, phosphorylated EIF4BP1, and phosphorylated ELK1 are higher in MSC at first diagnosis compared to MSC from salvage AML patients. IPA reveals that, for canonical pathways, a set associated with osteoblast differentiation was found to be the pathway most associated with the proteins differentially expressed at diagnosis compared to salvage (Online Supplementary Figure S8).
This study presents the first systematic study of protein expression differences between NL-MSC and AML-MSC. There were several notable observations. There were clear differences between the protein expression of MSC (whether from healthy donor or AML patient) and AML blast cells. This result was not surprising as one would predict different proteins would be prominent in mesenchymal cells and cells of hematopoietic/myeloid lineage. The major observation of this research was the discovery that AML-MSC have significantly different protein expression patterns compared to normal MSC, with 28 of 151 analyzed proteins being highly significantly different (FDR<0.006). These changes assumed four signatures in AML, 81% of which were very different from that of normal MSC, while 6% had an identical signature to NL- MSC, and another 13% were more like the normal signature than the leukemia patterns. Signature membership showed an association with cytogenetics, with ‘favorable’ cytogenetics being limited to the more NL-MSC-like pattern and ‘unfavorable’ cytogenetics not occurring in AML with an NL-MSC-like pattern. There was a difference in the distribution of the MSC population between men and women. The significance of these differences is not clear, but women tended to have higher percentages of Class 1 and Class 2 MSC compared to men. In leukemia progenitor cells, GSK3B is activated via an integrin-mediated mechanism in response to adhesion to a stromal cell exclusively in women patients.35 As ITGA2 and GSK3 are members of a protein constellation (i.e. constellation 1) that is differentially expressed in Class 1 and Class 2 (lower levels) compared to Class 3 and Class 4 (higher levels), it is tempting to speculate that integrin/GSK3 axis may contribute to sex-specific effects in MSC.
Furthermore, these signatures influence outcomes including response rates, remission duration and, perhaps, survival. Patients with Class 3 MSC fare much better than patients with Class 4 MSC as demonstrated by significantly longer remission duration and a trend toward longer OS. Changes in protein expression were often characterized by protein-protein correlations that were reversed from those seen in normal MSC, providing insight into the nature of this dysregulation and potentially providing therapeutic targets. In NL-MSC, the signature proteins were associated with adipocyte differentiation. That normal MSC, but not AML-MSC, possess protein pathways important in adipogenesis is consistent with other studies from our group and another group that found AML-MSC are primed toward osteoblastic differentiation and not adipocytic differentiation.41409 We reported that AML-MSC express higher levels of osteogenic markers, including Tissue Non-specific Alkaline Phosphatase (TNAP), RUNX2, Osterix, and Ostepontin, compared to MSC from age- matched heathy donors.40 In addition, in that study we found that AML-MSC readily differentiate along the osteogenic lineage pathway but are unable to differentiate into adipocytes. This differentiation potential of the MSC may be influenced by the leukemia cells themselves, as exposure of healthy donor MSC to AML cell lines such as OCI-AML3 induces gene expression of RUNX2, TNAP, and other osteogenic genes, and induces osteogenic differentiation of the stromal cells.40 The protein networks prevalent in NL-MSC that we identify here are consistent with signaling that skews toward adipocytic differentiation in the normal cells. Diaz de la Guardia et al. found that MSC from a high- risk AML group failed to differentiate into adipcocytes.41 The IPA data on canonical pathways in the AML-MSC compared at first diagnosis with refractory and relapse samples suggests the importance of MSC of the osteoblastic lineage in the AML niche, as many of these proteins are associated with osteoblast survival and differentiation.
PI3K/AKT is very prevalent in group 1 proteins which are associated with Class 4 MSC by IPA. We have previously reported that leukemia cells in co-culture with MCS induce activation of AKT and other survival kinases.42 It is plausible that the observed activation of AKT signaling in AML-MSC is due to the presence of leukemia cells in the niche or at least that the malignant cells may contribute to activation. In MSC, AKT has been implicated in positive regulation of Cyclin D1 (CCND1) and CDK4.4443 The presence of the PP2A B55 a subunit in group 1 suggests that either the protein phosphatase is not active against AKT in those cells or that AKT is less active in the AML-Like MSC compared to Normal-Like MSC. Despite its tumor suppressor role in suppression of AKT signaling, the PP2A subunit does support β-catenin expression by dephosphorylating serine and threonine residues that are required for destruction of the transcription factor (e.g. serine 33, threonine 41).4745 It is interesting that when expression of the 151 proteins surveyed by RPPA are correlated in the normal MSC and in the AML-MSC, β-catenin exhibits the greatest difference in correlation pattern between normal MSC and AML-MSC compared to the other proteins. It is plausible that the PP2A B55 a subunit may be a factor in this phenomenon, though further investigation will be required to verify this potential mechanism. It is interesting to note that suppression of PP2A (albeit via the catalytic core subunit Ca) in MSC cell line or pre-adipocyte cells results in differentiation to adipocytes via a mechanism involving loss of β-catenin.48 As pathway analysis of proteins associated with normal MSC (i.e. Constellation 3) pathways identifies adipogenesis as a key pathway, it will be important to determine if AML-MSC would be less likely to skew toward adipocyte differentiation because of a PP2A B55 a subunit/β-catenin axis.
BCL-XL is necessary for MSC survival during differentiation.27 AML-MSC expressed higher BCL-XL and Cyclin D1 protein. These findings may be attributed to STAT5 as this transcription factor is a regulator of both BCL-XL and Cyclin D1.49 We found by qRT-PCR that gene expression of both BCL-XL and Cyclin D1 was higher in AML-MSC (n=9) compared to normal MSC (n=10) (Online Supplementary Figure S9). These findings suggest that elevation of BCL-XL and Cyclin D1 protein in AML-MSC can be attributed at least in part to a transcriptional mechanism possibly involving STAT5.
Two prominent proteins identified as elevated in the AML-MSC group are p21 and p53 (Figure 1B). Elevated expression of p21 and increased senescence of MSC is consistent with the study on MDS-MSC that showed a role for p21 in IL-6 and TGF-β production.25 However, a recent study from Desbourdes et al. found p21 and p53 levels were similar between AML-MSC and healthy donor-derived MSC.50 The reason for the difference between our results and their results is not clear. It should be noted that the Desbourdes et al. study used less than 5 samples each of AML-derived MSC and healthy donor- derived MSC to determine p21 and p53 levels, so perhaps Class 1 or Class 2 MSC (which would have lower levels of p21 and p53) are over-represented in their samples.50 In addition, the average age of the AML patients in the Desbourdes et al. study was 49 y while the average age of the healthy donors was near 60 y, so perhaps the p21 and p53 levels in the healthy donor MSC are skewed higher as the donors are older than the patients. For p21 and p53 expression, an age match comparison of p21 and p53 in the AML-MSC and NL-MSC shows levels of these proteins are higher in the AML-MSC in at least 2 ages for each (Online Supplementary Table S2).
In conclusion, proteomic analysis identified a distinct set of proteins that distinguish normal MSC from AML- MSC. Our RPPA studies identified four major signatures of MSC in AML patients that may impact their function in the tumor microenvironment.
- ↵* SMK and PPR contributed equally to this work.
- Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/5/810
- Funding This work was supported in part by the research funding from the National Institutes of Health (P01CA49639 and MD Anderson Cancer Center Support Grant CA016672) and the Paul and Mary Haas Chair in Genetics (to MA).
- Received May 9, 2017.
- Accepted February 1, 2018.
- Konopleva M, Tabe Y, Zeng Z, Andreeff M. Therapeutic targeting of microenvironmental interactions in leukemia: mechanisms and approaches. Drug Resist Updat. 2009; 12(4–5):103-113. PubMedhttps://doi.org/10.1016/j.drup.2009.06.001Google Scholar
- Tabe Y, Konopleva M. Advances in understanding the leukaemia microenvironment. Br J Haematol. 2014; 164(6):767-778. PubMedhttps://doi.org/10.1111/bjh.12725Google Scholar
- Barcellos-de-Souza P, Gori V, Bambi F, Chiarugi P. Tumor microenvironment: bone marrow-mesenchymal stem cells as key players. Biochim Biophys Acta. 2013; 1836(2):321-335. Google Scholar
- da Silva Meirelles L, Chagastelles PC, Nardi NB. Mesenchymal stem cells reside in virtually all post-natal organs and tissues. J Cell Sci. 2006; 119(Pt 11):2204-2213. PubMedhttps://doi.org/10.1242/jcs.02932Google Scholar
- Eggenhofer E, Luk F, Dahlke MH, Hoogduijn MJ. The life and fate of mesenchymal stem cells. Front Immunol. 2014; 5:148. PubMedGoogle Scholar
- Greenbaum A, Hsu YM, Day RB. CXCL12 in early mesenchymal progenitors is required for haematopoietic stem-cell maintenance. Nature. 2013; 495(7440):227-230. PubMedhttps://doi.org/10.1038/nature11926Google Scholar
- Jacamo R, Chen Y, Wang Z. Reciprocal leukemia-stroma VCAM-1/VLA-4-dependent activation of NF-kappaB mediates chemoresistance. Blood. 2014; 123(17):2691-2702. PubMedhttps://doi.org/10.1182/blood-2013-06-511527Google Scholar
- Studeny M, Marini FC, Dembinski JL. Mesenchymal stem cells: potential precursors for tumor stroma and targeted-delivery vehicles for anticancer agents. J Natl Cancer Inst. 2004; 96(21):1593-1603. PubMedhttps://doi.org/10.1093/jnci/djh299Google Scholar
- Battula VL, Chen Y, Cabreira Mda G. Connective tissue growth factor regulates adipocyte differentiation of mesenchymal stromal cells and facilitates leukemia bone marrow engraftment. Blood. 2013; 122(3):357-366. PubMedhttps://doi.org/10.1182/blood-2012-06-437988Google Scholar
- Chen Y, Jacamo R, Shi YX. Human extramedullary bone marrow in mice: a novel in vivo model of genetically controlled hematopoietic microenvironment. Blood. 2012; 119(21):4971-4980. PubMedhttps://doi.org/10.1182/blood-2011-11-389957Google Scholar
- Kidd S, Spaeth E, Watson K. Origins of the tumor microenvironment: quantitative assessment of adipose-derived and bone marrow-derived stroma. PLoS One. 2012; 7(2):e30563. PubMedhttps://doi.org/10.1371/journal.pone.0030563Google Scholar
- Ling X, Marini F, Konopleva M. Mesenchymal Stem Cells Overexpressing IFN-beta Inhibit Breast Cancer Growth and Metastases through Stat3 Signaling in a Syngeneic Tumor Model. Cancer Microenviron. 2010; 3(1):83-95. PubMedhttps://doi.org/10.1007/s12307-010-0041-8Google Scholar
- Poggi A, Giuliani M. Mesenchymal Stromal Cells Can Regulate the Immune Response in the Tumor Microenvironment. Vaccines (Basel). 2016; 4(4)Google Scholar
- Spaeth E, Klopp A, Dembinski J, Andreeff M, Marini F. Inflammation and tumor microenvironments: defining the migratory itinerary of mesenchymal stem cells. Gene Ther. 2008; 15(10):730-738. PubMedhttps://doi.org/10.1038/gt.2008.39Google Scholar
- Spaeth EL, Labaff AM, Toole BP, Klopp A, Andreeff M, Marini FC. Mesenchymal CD44 expression contributes to the acquisition of an activated fibroblast phenotype via TWIST activation in the tumor microenvironment. Cancer Res. 2013; 73(17):5347-5359. PubMedhttps://doi.org/10.1158/1538-7445.AM2013-5347Google Scholar
- Kim JA, Shim JS, Lee GY. Microenvironmental remodeling as a parameter and prognostic factor of heterogeneous leukemogenesis in acute myelogenous leukemia. Cancer Res. 2015; 75(11):2222-2231. PubMedhttps://doi.org/10.1158/0008-5472.CAN-14-3379Google Scholar
- Raaijmakers MH, Mukherjee S, Guo S. Bone progenitor dysfunction induces myelodysplasia and secondary leukaemia. Nature. 2010; 464(7290):852-857. PubMedhttps://doi.org/10.1038/nature08851Google Scholar
- Shi C. Recent progress toward understanding the physiological function of bone marrow mesenchymal stem cells. Immunology. 2012; 136(2):133-138. PubMedhttps://doi.org/10.1111/j.1365-2567.2012.03567.xGoogle Scholar
- Balakrishnan K, Burger JA, Quiroga MP. Influence of bone marrow stromal microenvironment on forodesine-induced responses in CLL primary cells. Blood. 2010; 116(7):1083-1091. PubMedhttps://doi.org/10.1182/blood-2009-10-246199Google Scholar
- Kojima K, McQueen T, Chen Y. p53 activation of mesenchymal stromal cells partially abrogates microenvironment-mediated resistance to FLT3 inhibition in AML through HIF-1alpha-mediated down-regulation of CXCL12. Blood. 2011; 118(16):4431-4439. PubMedhttps://doi.org/10.1182/blood-2011-02-334136Google Scholar
- Marquez ME, Hernandez-Uzcategui O, Cornejo A, Vargas P, Da Costa O. Bone marrow stromal mesenchymal cells induce down regulation of CD20 expression on B-CLL: implications for rituximab resistance in CLL. Br J Haematol. 2015; 169(2):211-218. PubMedhttps://doi.org/10.1111/bjh.13286Google Scholar
- Xia B, Tian C, Guo S. c-Myc plays part in drug resistance mediated by bone marrow stromal cells in acute myeloid leukemia. Leuk Res. 2015; 39(1):92-99. PubMedhttps://doi.org/10.1016/j.leukres.2014.11.004Google Scholar
- Zeng Z, Shi YX, Samudio IJ. Targeting the leukemia microenvironment by CXCR4 inhibition overcomes resistance to kinase inhibitors and chemotherapy in AML. Blood. 2009; 113(24):6215-6224. PubMedhttps://doi.org/10.1182/blood-2008-05-158311Google Scholar
- Medyouf H, Mossner M, Jann JC. Myelodysplastic cells in patients reprogram mesenchymal stromal cells to establish a transplantable stem cell niche disease unit. Cell Stem Cell. 2014; 14(6):824-837. PubMedhttps://doi.org/10.1016/j.stem.2014.02.014Google Scholar
- Zhao Y, Wu D, Fei C. Down-regulation of Dicer1 promotes cellular senescence and decreases the differentiation and stem cell- supporting capacities of mesenchymal stromal cells in patients with myelodysplastic syndrome. Haematologica. 2015; 100(2):194-204. PubMedhttps://doi.org/10.3324/haematol.2014.109769Google Scholar
- Oliver L, Hue E, Priault M, Vallette FM. Basal autophagy decreased during the differentiation of human adult mesenchymal stem cells. Stem Cells Dev. 2012; 21(15):2779-2788. PubMedhttps://doi.org/10.1089/scd.2012.0124Google Scholar
- Oliver L, Hue E, Rossignol J. Distinct roles of Bcl-2 and Bcl-Xl in the apoptosis of human bone marrow mesenchymal stem cells during differentiation. PLoS One. 2011; 6(5):e19820. PubMedGoogle Scholar
- Kornblau SM, Singh N, Qiu Y, Chen W, Zhang N, Coombes KR. Highly phosphorylated FOXO3A is an adverse prognostic factor in acute myeloid leukemia. Clin Cancer Res. 2010; 16(6):1865-1874. PubMedhttps://doi.org/10.1158/1078-0432.CCR-09-2551Google Scholar
- Kornblau SM, Womble M, Qiu YH. Simultaneous activation of multiple signal transduction pathways confers poor prognosis in acute myelogenous leukemia. Blood. 2006; 108(7):2358-2365. PubMedhttps://doi.org/10.1182/blood-2006-02-003475Google Scholar
- Ruvolo PP, Qiu Y, Coombes KR. Phosphorylation of GSK3alpha/beta correlates with activation of AKT and is prognostic for poor overall survival in acute myeloid leukemia patients. BBA Clin. 2015; 4:59-68. Google Scholar
- Ruvolo PP, Qui YH, Coombes KR. Low expression of PP2A regulatory subunit B55alpha is associated with T308 phosphorylation of AKT and shorter complete remission duration in acute myeloid leukemia patients. Leukemia. 2011; 25(11):1711-1717. PubMedhttps://doi.org/10.1038/leu.2011.146Google Scholar
- Tibes R, Qiu Y, Lu Y. Reverse phase protein array: validation of a novel proteomic technology and utility for analysis of primary leukemia specimens and hematopoietic stem cells. Mol Cancer Ther. 2006; 5(10):2512-2521. PubMedhttps://doi.org/10.1158/1535-7163.MCT-06-0334Google Scholar
- Szklarczyk D, Franceschini A, Wyder S. STRING v10: protein-protein interaction networks, integrated over the tree of life. Nucleic Acids Res. 2015; 43(Database issue):D447-452. PubMedhttps://doi.org/10.1093/nar/gku1003Google Scholar
- Huang JC, Basu SK, Zhao X. Mesenchymal stromal cells derived from acute myeloid leukemia bone marrow exhibit aberrant cytogenetics and cytokine elaboration. Blood Cancer J. 2015; 5:e302. Google Scholar
- Bertrand J, Despeaux M, Joly S. Sex differences in the GSK3beta-mediated survival of adherent leukemic progenitors. Oncogene. 2012; 31(6):694-705. PubMedhttps://doi.org/10.1038/onc.2011.258Google Scholar
- Blalock WL, Piazzi M, Bavelloni A. Identification of the PKR nuclear interactome reveals roles in ribosome biogenesis, mRNA processing and cell division. J Cell Physiol. 2014; 229(8):1047-1060. PubMedhttps://doi.org/10.1002/jcp.24529Google Scholar
- Ali AT, Hochfeld WE, Myburgh R, Pepper MS. Adipocyte and adipogenesis. Eur J Cell Biol. 2013; 92(6–7):229-236. PubMedhttps://doi.org/10.1016/j.ejcb.2013.06.001Google Scholar
- Jin W, Takagi T, Kanesashi SN. Schnurri-2 controls BMP-dependent adipogenesis via interaction with Smad proteins. Dev Cell. 2006; 10(4):461-471. PubMedhttps://doi.org/10.1016/j.devcel.2006.02.016Google Scholar
- Kauppinen A, Suuronen T, Ojala J, Kaarniranta K, Salminen A. Antagonistic crosstalk between NF-kappaB and SIRT1 in the regulation of inflammation and metabolic disorders. Cell Signal. 2013; 25(10):1939-948. PubMedhttps://doi.org/10.1016/j.cellsig.2013.06.007Google Scholar
- Battula VL, Le PM, Sun JC. AML- induced osteogenic differentiation in mesenchymal stromal cells supports leukemia growth. JCI Insight. 2017; 2(13)Google Scholar
- Diaz de la Guardia R, Lopez-Millan B, Lavoie JR. Detailed Characterization of Mesenchymal Stem/Stromal Cells from a Large Cohort of AML Patients Demonstrates a Definitive Link to Treatment Outcomes. Stem Cell Reports. 2017; 8(6):1573-1586. Google Scholar
- Tabe Y, Jin L, Tsutsumi-Ishii Y. Activation of integrin-linked kinase is a critical prosurvival pathway induced in leukemic cells by bone marrow-derived stromal cells. Cancer Res. 2007; 67(2):684-694. PubMedhttps://doi.org/10.1158/0008-5472.CAN-06-3166Google Scholar
- Gharibi B, Ghuman MS, Hughes FJ. Akt- and Erk-mediated regulation of proliferation and differentiation during PDGFRbeta- induced MSC self-renewal. J Cell Mol Med. 2012; 16(11):2789-2801. PubMedGoogle Scholar
- Lai VK, Ashraf M, Jiang S, Haider K. MicroRNA-143 is a critical regulator of cell cycle activity in stem cells with co-overexpression of Akt and angiopoietin-1 via transcriptional regulation of Erk5/cyclin D1 signaling. Cell Cycle. 2012; 11(4):767-777. PubMedhttps://doi.org/10.4161/cc.11.4.19211Google Scholar
- Hein AL, Seshacharyulu P, Rachagani S. PR55alpha Subunit of Protein Phosphatase 2A Supports the Tumorigenic and Metastatic Potential of Pancreatic Cancer Cells by Sustaining Hyperactive Oncogenic Signaling. Cancer Res. 2016; 76(8):2243-2253. PubMedhttps://doi.org/10.1158/0008-5472.CAN-15-2119Google Scholar
- Stamos J, Weis W. The β-catenin destruction complex. Cold Spring Harb Perspect Biol. 2013; 5:a007898. PubMedhttps://doi.org/10.1101/cshperspect.a007898Google Scholar
- Zhang W, Yang J, Liu Y. PR55 alpha, a regulatory subunit of PP2A, specifically regulates PP2A-mediated beta-catenin dephosphorylation. J Biol Chem. 2009; 284(34):22649-22656. PubMedhttps://doi.org/10.1074/jbc.M109.013698Google Scholar
- Okamura H, Yang D, Yoshida K, Teramachi J, Haneji T. Reduction of PP2A Calpha stimulates adipogenesis by regulating the Wnt/GSK-3beta/beta-catenin pathway and PPARgamma expression. Biochim Biophys Acta. 2014; 1843(11):2376-2384. Google Scholar
- Grad JM, Zeng XR, Boise LH. Regulation of Bcl-xL: a little bit of this and a little bit of STAT. Curr Opin Oncol. 2000; 12(6):543-549. PubMedhttps://doi.org/10.1097/00001622-200011000-00006Google Scholar
- Desbourdes L, Javary J, Charbonnier T. Alteration Analysis of Bone Marrow Mesenchymal Stromal Cells from De Novo Acute Myeloid Leukemia Patients at Diagnosis. Stem Cells Dev. 2017; 26(10):709-722. Google Scholar