Impact of mesenchymal stromal cell–derived vesicular cargo on B-cell acute lymphoblastic leukemia progression

B-cell acute lymphoblastic leukemia (B-ALL) is a hematological malignancy affecting children and adults.¹ BCR-ABL1+ B-ALL is a subtype of B-ALL accounting for 3% to 5% of pediatric and 25% of B-ALL cases in adults,² conferring a poor prognosis.³ BCR-ABL1⁺ B-ALL is characterized by the chromosomal translocation t(9;22), giving rise to the constitutively active tyrosine kinase BCR-ABL1.² Despite overall improved clinical outcomes, especially in pediatric B-ALL,¹ the incidence of relapse in BCR-ABL1⁺, also BCR-ABL1⁻ B-ALL, remains a concern, highlighting the need for the discovery of novel and potentially targetable mechanisms that accelerate the progression of B-ALL.⁴ 

The bone marrow (BM) microenvironment (BMM) supports the progression of B-ALL and other leukemias and promotes their resistance to established therapies.^5,6 Reciprocal interactions between leukemia cells and the BMM are mediated by BMM constituents, such as mesenchymal stromal cells (MSCs), osteoblasts, and other factors.⁵ Although osteoclasts are considered components of the BMM, knowledge of their role in hematopoiesis has been limited to their influence on hematopoietic stem cell mobilization.⁷ Only few reports have described the effects of B-ALL on bone physiology,^8,9 and overall, insight into the role of osteoclasts in B-ALL is limited.

Pleckstrin homology domain family M member 1 (PLEKHM1) is a protein that facilitates the trafficking of secretory lysosomes in osteoclasts. Mutations in PLEKHM1 may lead to dysfunctional osteoclasts and osteopetrotic phenotypes in rats and humans.¹⁰ In a related function, PLEKHM1 is an autophagy adapter protein mediating the fusion of autophagosomes with lysosomes^11,12 and the trafficking of endosomes,^13-15 which include multivesicular bodies (MVBs). These MVBs contain intraluminal vesicles, which, upon fusion of the MVB with the plasma membrane, are released as small extracellular vesicles (sEVs) or exosomes.¹⁶ Therefore, we hypothesized that PLEKHM1 may influence B-ALL progression via osteoclast-mediated defects as found in osteopetrosis¹⁰ or altered trafficking of vesicles released from osteoclasts or other cells.

Here, we identify a circular model in which TNF-α secreted from BCR-ABL1⁺ B-ALL cells establishes a favorable niche for B-ALL progression via a reduction of PLEKHM1 in MSC and hence increased loading of MVB-derived sEVs with syntenin and syndecan-1, which are released into their extracellular environment. Uptake of these sEVs by B-ALL cells increases phosphorylation of protein kinase B (AKT) and focal adhesion kinase (FAK), promoting B-ALL cell proliferation, migration, and BCR-ABL1⁺ B-ALL advancement. The discovered mechanism connects the endolysosomal system as a quality control pathway in cells of the tumor microenvironment with leukemia progression and may also be relevant in other nonleukemic cancer types.

To induce BCR-ABL1⁺ B-ALL, BM cells were harvested from wild-type (WT) mice, transduced once with MSCV IRES green fluorescent protein (GFP) BCR-ABL1–expressing retrovirus and transplanted into sublethally irradiated (2 × 450 cGy) recipient mice.¹⁷ 

sEVs were obtained by differential centrifugation.¹⁸ 

Data are represented as mean ± SD and were considered statistically significant if P ≤ .05. GraphPad Prism Software (Prism 9.0) was used for the generation of all graphs and for statistical analyses. Statistical significance was calculated using the Student t test, one-way or two-way analysis of variance (ANOVA). For post hoc analysis, Tukey, Sidak, or Dunnett tests were used to determine statistical significance between multiple comparisons. For survival analyses by Kaplan-Meier curves, the log-rank (Mantel-Cox) test was used.

To study the role of PLEKHM1 in the BMM in B-ALL, we transplanted WT murine BM, transduced with BCR-ABL1–expressing retrovirus, into Plekhm1 KO (Knockout) or WT recipient mice. Comparison of the hematological profile in normal Plekhm1 KO vs WT mice did not show differences (supplemental Figure 1A-E). Upon induction of B-ALL, a significantly increased number of leukocytes (P = .04; Figure 1A) and GFP⁺ (BCR-ABL1⁺) BP-1⁺ pre-B cells were observed in Plekhm1 KO compared to WT recipient mice (P = .02; Figure 1B). This was accompanied by significant disease acceleration in Plekhm1 KO recipients (P = .04; Figure 1C). This phenotype was not due to increased homing of the BCR-ABL1⁺ (GFP⁺) BP-1⁺ leukemia-initiating cells (LICs) to the BM or spleen of Plekhm1 KO compared with WT recipient mice (Figure 1D). Nonirradiation of recipients before transplantation of B-ALL LIC¹⁹ confirmed the increased percentage of GFP⁺ (BCR-ABL1⁺) cells in Plekhm1 KO recipients (supplemental Figure 2A-B). Transplantation of 5-fluorouracil–pretreated BM, transduced with BCR-ABL1 or MLL-AF9–expressing retrovirus, to induce chronic myeloid leukemia (CML)–like myeloproliferative neoplasia or acute myeloid leukemia (AML), in contrast, revealed that Plekhm1 KO recipient mice with CML-like myeloproliferative neoplasia showed a decreased GFP⁺ (BCR-ABL1⁺) CD11b⁺ myeloid tumor burden in PB (supplemental Figure 3A) and a modest prolongation of survival compared with WT recipients (supplemental Figure 3B). No differences were observed between the 2 cohorts with regard to tumor load (supplemental Figure 3C) or overall survival in the AML model (supplemental Figure 3D).

Focusing on BCR-ABL1⁺ B-ALL, we showed reduced early apoptosis in GFP⁺ (BCR-ABL1⁺) BP-1⁺ cells from Plekhm1 KO compared to those from WT recipient mice (P = .02; Figure 1E; supplemental Figure 3E). In addition, expression of the B-cell–specific markers CD19 (P = .02; Figure 1F) and B220 (P = .017; Figure 1G) was significantly reduced in GFP⁺ (BCR-ABL1⁺) BP-1⁺ cells derived from Plekhm1 KO compared to those from WT mice. Consistently, expression levels of the paired box 5 (Pax5) gene (Bsap), which encodes for a transcription factor with a key role in B-lymphocyte development²⁰ and CD19 expression,²¹ were reduced in BM cells from Plekhm1 KO mice (P = .03; Figure 1H). Proteomic analysis of GFP⁺ (BCR-ABL1⁺) BP-1⁺ cells from Plekhm1 KO mice with B-ALL revealed significantly decreased expression of the immunoglobulin heavy constant alpha 1 and immunoglobulin kappa constant proteins, both of which are components of the B-cell receptor of mature B lymphocytes,²² compared with WT mice (supplemental Figure 3F). This may suggest a more immature phenotype of B-ALL cells in a Plekhm1 KO than in a WT BMM. To assess the number or function of LICs, BM cells from WT or Plekhm1 KO mice with established B-ALL (supplemental Figure 3G) were transplanted into WT secondary recipient mice. This led to an increased tumor load in the PB (P = .046; Figure 1I) and a strikingly reduced overall survival (P < .0001; Figure 1J) of mice that had received B-ALL cells from Plekhm1 KO compared with WT donor mice.

Mice transplanted with B-ALL cells from a Plekhm1 KO BMM had an increased percentage of GFP⁺ (BCR-ABL1⁺) BP-1⁺ cells in the BM (P = .02; Figure 1K). The percentage of these cells in the S phase of the cell cycle (P = .006; Figure 1L) was increased, and the fraction of these cells undergoing early (P = .04; Figure 1M) and late (P = .01; Figure 1N) apoptosis was decreased. In contrast, transplantation of PLEKHM1-deficient B-ALL–initiating cells led to a trend toward survival prolongation compared with recipients of WT LICs (supplemental Figure 3H).

Taken together, these data indicate that PLEKHM1 deficiency in the BMM increases B-ALL tumor load and shortens survival. This effect is more pronounced in the setting of secondary transplantation, suggesting that a Plekhm1 KO BMM increases the number and/or function of LICs in B-ALL.

As deficiency of PLEKHM1 leads to reduced resorptive activity in osteoclasts,¹⁰ we initially hypothesized that the shortened survival observed in Plekhm1 KO mice with B-ALL may be due to defective osteoclasts. Therefore, we isolated monocytes from the BM of WT or Plekhm1 KO mice and differentiated them into osteoclasts.²³ Indeed, Plekhm1 KO osteoclasts showed reduced resorptive activity (supplemental Figure 4A). However, no differences in the proliferation of BCR-ABL1⁺ BA/F3 cells, a frequently used in vitro model for B-ALL,²⁴ cultured on WT or Plekhm1 KO osteoclasts were observed (supplemental Figure 4B), suggesting that accelerated B-ALL progression in Plekhm1 KO mice was likely not attributable to osteoclasts. Because PLEKHM1 is also expressed in MSCs,²⁵ which have a known role in B-ALL,^24,26 we subsequently hypothesized that MSCs may be involved in the establishment of the observed phenotype. To address this, we isolated CD45⁻ Ter-119⁻ PDGFRα⁺ Sca-1⁺ MSC²⁷ from the BM of healthy WT or Plekhm1 KO mice, which did not differ significantly with regard to the immunophenotype (supplemental Figure 4C). Testing the function and differentiation of MSCs, we found that the colony-forming ability (P = .0002; Figure 2A) and osteogenic differentiation potential of Plekhm1 KO MSC were significantly higher than in WT MSCs (P = .01; Figure 2B). Conversely, the adipogenic differentiation potential was significantly lower (P = .01; Figure 2C). Given that PLEKHM1 plays a role in vesicular trafficking and fusion of late endosomes with lysosomes,¹⁴ we further hypothesized that the reduced differentiation ability of Plekhm1 KO MSCs into adipocytes may be associated with impairment of lipid droplet fusion. Therefore, we investigated whether the number of MVBs, a subset of specialized endosomes,²⁸ that is, late endosomes, differed between WT and Plekhm1 KO MSC. Immunofluorescence staining for CD63, a marker of MVBs,²⁹ showed significantly increased MVBs in Plekhm1 KO than in WT MSCs (P = .002; Figure 2D). Immunofluorescence for LC3-positive (autophagosomes) or LAMP1-positive vesicles (lysosomes) showed no significant differences in their number or colocalization between WT and Plekhm1 KO MSCs (supplemental Figure 5A-D), indicative of overall similar basal levels of autophagy between WT and Plekhm1 KO MSCs. Taken together, these results suggest that Plekhm1 KO MSCs have an increased capacity to form colonies and a decreased capacity to differentiate into adipocytes while harboring an increased number of CD63⁺ MVBs.

MVB-derived exosomes³⁰ are critical mediators in the communication between leukemia cells and their environment.^31,32 Given the increased number of CD63⁺ MVBs in Plekhm1 KO MSCs, we hypothesized that deficiency of PLEKHM1 may affect the release or content of exosomes. Using differential centrifugation, sEVs, known to be enriched in exosomes and found in the 100k fraction,^18,33 were isolated from the conditioned medium of WT or Plekhm1 KO MSCs. We confirmed the successful isolation of sEVs by the expression of the exosomal markers CD81 and flotillin-1 in the sEV preparation, as well as by the absence of expression of the Golgi matrix protein 130 kD (GM130; Golgin A2)³⁴ (Figure 3A). Gradient centrifugation, in contrast, led to significant loss of material and, therefore, could not be used for this study, but revealed differential enrichment of CD81 and flotillin-1 in the 10% and 15% gradient for exosomes from WT vs the 25% gradient for exosomes from Plekhm1 KO MSCs (supplemental Figure 6A). The amount of CD81 protein (Figure 3A; supplemental Figure 6B) and the number of CD81⁺ EVs being detected by imaging flow cytometric analysis at the single EV level³⁵ (Figure 3B; supplemental Figure 6C) were similar, regardless of whether the EVs were derived from MSCs from WT or Plekhm1 KO mice (hereafter called WT and Plekhm1 KO sEVs, respectively). Consistently, molecules of equivalent soluble fluorochrome, describing the amount of antibody bound to CD81 per sEV, were also similar between sEVs derived from WT and Plekhm1 KO MSCs (supplemental Figure 6D). In addition, the distribution in respective cell cycle stages was similar between WT and Plekhm1 KO MSCs (supplemental Figure 6E). Proteomic analysis of prepared sEV samples showed increased levels of vacuolar protein sorting 23 (TSG101) and vacuolar protein sorting 28 in sEV preparations from Plekhm1 KO compared with WT MSCs (Figure 3C). Both identified proteins belong to the endosomal sorting complex required for transport (ESCRT) complex I (ESCRT-I)³⁶ and are responsible for the formation and sorting of cargo in intraluminal vesicles. Similarly, syntenin, which is involved in ESCRT-dependent exosomal biogenesis,³⁷ was also found to be significantly increased in Plekhm1 KO sEV preparations (Figure 3C). Immunoblot analysis confirmed the enrichment of syntenin in Plekhm1 KO sEV preparations (Figure 3D). Syntenin directly interacts with syndecan-1, a heparan sulfate proteoglycan and integral membrane protein³⁷ present in both healthy and tumor cells, including leukemia cells.³⁸ Consistent with the enrichment of syntenin, we also observed increased syndecan-1 protein levels in Plekhm1 KO sEV preparations (Figure 3D; supplemental Figure 6F). Immunoblot analysis of MSCs, that is, the donor cells of sEVs, similarly revealed an enrichment of syntenin and syndecan-1 in Plekhm1 KO compared with WT MSCs (P = .049 [syntenin], P = .043 [syndecan-1]; Figure 3E). Increased levels of syndecan-1 were also confirmed in Plekhm1 KO MSCs (P = .01; Figure 3F). These findings are consistent with reports that the overexpression of the cytoplasmic adapter molecule syntenin and syndecan-1 in donor cells may enhance their exosomal expression.³⁷ Our data suggest that deficiency of PLEKHM1 leads to increased levels of syntenin and syndecan-1 in MSCs and MSC-derived sEVs, while leaving numbers of released sEVs unaffected.

Given the differences in the cargo of sEVs from Plekhm1 KO and WT MSCs, we tested the effect of the obtained sEV preparations on B-ALL target cells. Indeed, exposure of BCR-ABL1⁺ BA/F3 cells to Plekhm1 KO sEV preparations led to a significant increase in the levels of intracellular syntenin and syndecan-1 compared with BCR-ABL1⁺ BA/F3 cells exposed to sEVs prepared from WT MSCs (P = .02 [syntenin], P = .04, [syndecan-1]; Figure 4A). Prior treatment of BCR-ABL1⁺ BA/F3 cells exposed to sEVs from Plekhm1 KO vs WT MSCs with bortezomib and cycloheximide to inhibit protein degradation and synthesis, respectively, confirmed that the elevated levels of syntenin after exposure to Plekhm1 KO sEVs were due to uptake of the sEV-derived cargo (supplemental Figure 7A). In addition, exposure of BCR-ABL1⁺ BA/F3 cells to Plekhm1 KO sEV preparations significantly increased the phosphorylation of AKT on the Ser473 residue compared with sEVs prepared from WT MSCs (P = .03; Figure 4B). AKT phosphorylation on Ser473 is required for maximal AKT activation controlling cell cycle, proliferation, and survival.^39,40 This effect was reversed after treatment of BA/F3 cells with dynasore to inhibit endocytosis⁴¹ (Figure 4B; supplemental Figure 7B). In addition, combination treatment with a syntenin inhibitor, PDZ1i,⁴² and the BCR-ABL1–targeting tyrosine kinase inhibitor dasatinib reduced the viability of BCR-ABL1⁺ BA/F3 cells more efficiently than dasatinib alone (P = .008; supplemental Figure 7C). Furthermore, we found a trend toward higher expression of syntenin in BM cells of Plekhm1 KO compared with WT mice with B-ALL but not CML (supplemental Figure 7D-E).

Syndecan-1 forms complexes with growth factor receptors and/or integrins on the cell surface, resulting in their stabilization and enhancement of their downstream signaling pathways.^43,44 In agreement with this, we observed a significant increase in phosphorylation of FAK on residue Tyr397 in BCR-ABL1⁺ BA/F3 cells after exposure to Plekhm1 KO compared with sEVs prepared from WT MSCs, whereby FAK is known to serve as a mediator of activated integrin signal transduction⁴⁵ (P = .03; Figure 4C). Expression of integrins αV, β3, α5, and α2b on the surface of B-ALL cells isolated from WT or Plekhm1 KO mice, however, did not reveal any differences (supplemental Figure 8A-B). In line with the known role of FAK,^46,47 syndecan,³⁸ and syntenin⁴⁸ for cell mobility, the migration of BCR-ABL1⁺ BA/F3 cells treated with sEVs derived from Plekhm1 KO compared with WT MSCs was significantly increased (P = .02; Figure 4D). To test directly whether sEV preparations from Plekhm1 KO MSCs promoted the acceleration of B-ALL, we treated BCR-ABL1–transduced WT BM cells with WT or Plekhm1 KO sEVs and transplanted these pretreated BM cells into WT recipient mice. This did not lead to differences in the tumor load in the form of GFP⁺ (BCR-ABL1⁺) BP-1⁺ cells in PB of the recipient mice (supplemental Figure 8C). However, a strong trend toward shortened survival of mice that had received BCR-ABL1⁺ BM cells previously exposed to Plekhm1 KO compared with WT sEVs was observed (P = .06; Figure 4E). Previous exposure of LICs to Plekhm1 KO but not WT sEVs also led to a trend toward an increase of pAKT Ser473 in GFP⁺ (BCR-ABL1)⁺ BP-1⁺ cells from recipient mice with B-ALL (Figure 4F; supplemental Figure 8D).

Testing possible differences between BCR-ABL1⁺ and BCR-ABL1⁻ B-ALL, we showed that transplantation of the human BCR-ABL1⁺ B-ALL cell line SUPB15, previously exposed to sEVs from Plekhm1 KO compared with WT MSCs, but not the equally treated BCR-ABL1⁻ B-ALL cell line NALM6 into NOD SCID IL-2 receptor γ KO mice (supplemental Figure 7H-I), led to a strong trend toward shortened survival (P = .057; Figure 4G; supplemental Figure 8E-G). In summary, uptake of sEVs derived from Plekhm1 KO compared with WT MSCs led to an increase in the levels of syntenin, syndecan-1, pFAK, and pAKT Ser473 in B-ALL cells, an increase in migration of B-ALL cells, and likely acceleration of B-ALL progression. This effect seemed more pronounced in BCR-ABL1⁺ lymphoid rather than in BCR-ABL1⁺ myeloid or BCR-ABL1⁻ lymphoid leukemias.

Because B-ALL cells modify MSCs via the generation of tumor necrosis factor (TNF) α²⁴ and TNFα/TNF levels were higher in murine B-ALL vs AML and CML (supplemental Figure 9A), as well as in BCR-ABL1⁺ SUPB15 compared with BCR-ABL1⁻ NALM6 human B-ALL cell lines (supplemental Figure 9B-C), we tested whether the secretion of TNF-α by B-ALL cells may reproduce the phenotype observed in Plekhm1 KO MSCs. Indeed, the treatment of WT MSCs with TNF-α led to increased phosphorylation of the NF-κB subunit p65 (supplemental Figure 9D). In addition, treatment with TNF-α or coculture of WT MSCs with BCR-ABL1⁺ BA/F3 cells, separated by a membrane, decreased expression of PLEKHM1 in WT MSCs with concomitant upregulation of syntenin and syndecan-1 (Figure 5A; supplemental Figure 9E-F). Adalimumab, an antibody inhibiting TNF-α, rescued PLEKHM1 reduction in MSCs after coculture with BCR-ABL1⁺ BA/F3 cells (Figure 5B; supplemental Figure 9G).

The decrease of PLEKHM1 after TNF-α treatment was already detectable in MSCs (supplemental Figure 10A) and in the stromal cell line MS5 (supplemental Figure 10B) after 30 minutes. In addition, the migration of BCR-ABL1⁺ BA/F3 cells, which had been exposed to sEVs from MSCs (P = .004; Figure 5C) or MS5 (supplemental Figure 10C) cells, pretreated with TNF-α, was increased. Testing the qualitative effect of TNF-α on sEV production, we found that sEVs from WT MSCs, which had been treated with TNF-α, led to an upregulation of pFAK in sEV-exposed BCR-ABL1⁺ BA/F3 cells (P = .04; Figure 5D).

With relevance to the human setting, we identified that exposure of HS5 cells, a human BM stroma cell line, to SUPB15 cells (P = .04; Figure 5E) and treatment of HS5 cells (P = .001; Figure 5F) and primary human MSCs (P = .03; Figure 5G) with human TNF-α also led to a reduction of PLEKHM1. Treatment of HS5 cells with TNF-α resulted in the release of sEVs with increased levels of their cargo, syntenin (P = .05; Figure 5H). Furthermore, TNF-α decreased levels of PLEKHM1 in MC3T3 cells, an osteoblast precursor cell line, but not in H5V cells or human umbilical vein endothelial cells, murine and human endothelial cell lines, respectively (supplemental Figure 11A-C).

Finally, in an effort to test whether the effect was specific to TNF-α, we demonstrated that treatment of MSCs (supplemental Figure 12A) or MS5 cells (supplemental Figure 12B) with IL-1β, another proinflammatory cytokine, caused a significant decrease in PLEKHM1 levels after 24 hours. Similar to TNF-α, Il1b was more highly expressed in SUPB15 compared with NALM6 cells (supplemental Figure 12C). In summary, these data demonstrate that secretion of TNF-α and other inflammatory cytokines such as IL-1β reduces expression of PLEKHM1 in murine and human MSCs and, in the case of TNF-α, alters the quality of sEVs secreted from MSCs. The altered sEV cargo affected migration and likely pFAK signaling in target leukemia cells, thereby maintaining a circuit that may perpetuate leukemia progression. A reduction of PLEKHM1 in response to TNF-α was also identified in osteoblastic but not endothelial cell lines.

Testing the relevance of our findings in the human setting, we found the percentage of PLEKHM1⁺ CD271⁺ MSCs to be decreased in trephine biopsies of patients with B-ALL compared to CML (P = .03; Figure 6A-B; supplemental Figure 13A). The percentage of CD271⁺ cells was significantly higher in sections from healthy individuals compared with patients with certain hematological malignancies (supplemental Figure 13B), whereas the percentage of PLEKHM1⁺ cells was lowest in patients with B-ALL (supplemental Figure 13C). Although the number of analyzed patients is low, the data suggest that PLEKHM1 expression may also be decreased in human MSCs in B-ALL compared to CML, similar to our findings in mice.

In the context of cancer, most studies have focused on cancer cell–derived exosomes or sEVs, which have the capacity to transform the tumor microenvironment.^31,32,49 Few studies and, to our knowledge, none in B-ALL have investigated the role of sEVs derived from the microenvironment.^50-52 Here, we demonstrate that deletion of Plekhm1, a central protein in the endolysosomal system, modulates the cargo of MSC-derived, that is, cancer cell–extrinsic, sEVs by increasing their loading with syntenin and syndecan-1. Uptake of this cargo decreases apoptosis and increases leukemia cell cycling. This promotes B-ALL progression and leukemia-initiating capacity via increased phosphorylation of FAK and AKT, specifically in the secondary transplantation setting, in which exposure of LICs to a Plekhm1 KO BMM is prolonged. In turn, TNF-α and IL-1β, secreted by B-ALL and other leukemias,⁵³ decrease PLEKHM1 in murine and human MSCs and osteoblastic but not endothelial cells. However, given our observed differences in the phenotype of CML, AML vs B-ALL in a Plekhm1 KO BMM and the predominance of the effect in BCR-ABL1⁺ vs BCR-ABL1⁻ B-ALL, we hypothesize that syndecan-1 and/or syntenin may have leukemia lineage- and/or oncogene-specific effects, as has been suspected.^38,54 These effects may contribute to the worse prognosis in BCR-ABL1⁺ B-ALL.

Consistent with other reports on the involvement of vesicular trafficking–associated proteins for the regulation of EV cargo,⁵⁵ we discovered a potential role for PLEKHM1 in the loading of intraluminal vesicles with proteins such as syntenin and syndecan-1. In contrast, the levels of both proteins did not differ in our proteomics study of B-ALL cells from a WT vs a Plekhm1 KO BMM, suggesting that their intracellular presence may be short but trigger the phosphorylation of AKT and FAK.

Syntenin and syndecan-1 have been suggested to stabilize growth factor receptors and/or integrins on the cell surface.^43,56,57 Consistently, we observed increased expression of pFAK⁵⁸ and pAKT downstream of integrin receptors in leukemia cells exposed to sEVs from a Plekhm1 KO compared with a WT BMM, although expression levels of integrins did not differ. An effect of stabilization of growth factor receptors due to syntenin- and syndecan-1–loaded sEVs on AKT phosphorylation cannot be excluded. Consistent with our data, syndecan-1 was recently identified as an important mediator of CML progression in blast crisis and to promote the migration and localization of CML progression in blast crisis cells while augmenting the activity of integrin β₇.³⁸ 

Inflammatory cytokines, specifically TNF-α, have previously been implicated in the remodeling of the BMM²⁴ or the tumor microenvironment,^59,60 which, in turn, promoted tumor growth. Therefore, the net effect is similar to our study, although the mechanisms differ. TNF-α is known to promote sEV release and to modify their cargo,^61,62 but the described machinery did not involve proteins of the endosome generation and trafficking pathways.

Finally, autophagy inhibitors, such as hydroxychloroquine, have been used in combination with targeted therapies in clinical trials,⁶³ but caution must be exercised with regard to the development of novel autophagy inhibitors^64,65 because these treatments may also affect the release and quality of tumor-modifying exosomal cargo by cells of the microenvironment, endangering treatment success or negatively influencing cancer progression.

Taken together, our data suggest that quality control processes in cells of the tumor microenvironment are usurped by cancer cells to promote cancer progression. Although human patient numbers are small, our data suggest that PLEKHM1⁺ MSCs may indeed be reduced in B-ALL but not in CML, likely because of higher TNF-α levels in B-ALL, explaining observed survival differences in Plekhm1 KO mice with B-ALL vs CML. How the discovered pathways can be exploited therapeutically will be studied in depth in the future.

The authors thank the Quantitative Proteomics Unit (IBC2, Goethe University Frankfurt) for support on liquid chromatography–mass spectrometry (LC-MS) instrumentation and the Deutsche Forschungsgemeinschaft (DFG) for funding the Orbitrap Lumos LC-MS system used in this study (FuG 403765277).

This work was supported by grant KR2072/6-1 from the DFG (to D.S.K.) and by the LOEWE Center for Cell and Gene Therapy Frankfurt (CGT) and institutional funds of the Georg-Speyer-Haus (to D.S.K.). The Georg-Speyer-Haus is funded jointly by the German Federal Ministry of Health and the Ministry of Higher Education, Research and the Arts of the State of Hessen (HMWK). CGT is funded by HMWK, reference number: III L 4 to 518/17.004 (2010).

Contribution: C.K. and D.S.K. designed the experiments; C.K. and V.R.M. performed the in vitro and in vivo experiments; R.K., C.Z., J.B., R.S.P., and A.C.-P. assisted with experiments and critically discussed data; G.T. and C.M. performed the proteomics experiments; T.T. and B.G. performed the experiment involving AMNIS technology; K.B. performed the staining of human bone sections; D.G.M. and I.D. critically discussed data; L.-M. P. and H.B. provided human MSC; P.D.-P. helped with osteoclast differentiation; C.K. analyzed the data and wrote a first draft of the manuscript; D.S.K. supervised the research, analyzed data, and wrote the manuscript; and all authors critically read and approved the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Daniela S. Krause, Goethe University, c/o Georg-Speyer-Haus, Institute for Tumor Biology and Experimental Therapy, Paul-Ehrlich-Str. 42-44, 60596 Frankfurt am Main, Germany; e-mail: krause@gsh.uni-frankfurt.de.