Sry-Box Transcription Factor 9 ( SOX9)-Driven Mesenchymal Stromal Cell Differentiation Signature As a Determinant of Induction Chemotherapy Responsiveness in Acute Myeloid Leukemia Core Bone Marrow Biopsies

Leukemic clones in acute myeloid leukemia (AML) alter the bone marrow (BM) microenvironment so that the normal functions of hematopoietic stem cells (HSCs) are interrupted, aiding malignant progression. Accumulating evidence points to a protective role that the microenvironment provides for the AML clone (Colmone et al., Science, 2008). It is therefore crucial to detail the mechanisms by which the BM stroma promotes leukemia progression to inform the development of more effective targeted therapies. Most studies focused on detailing the role of BM stroma in AML etiology use BM aspirates, which have a limited ability to represent the physical structure of the BM microenvironment. To accurately capture BM composition, our group used the core bone marrow biopsies (cBMBs) of 15 AML patients who responded to and 14 AML patients who did not respond to 7+3 induction chemotherapy (Treaba et al., British Journal of Haematology, 2022). RNA-sequencing data of these biopsies (GEO GSE213210) indicated significant differences in the composition of the stromal component of responders and non-responders. Responder cBMBs were enriched in undifferentiated CXCL12-positive mesenchymal stem cells (MSCs), with elevated expression of platelet derived growth factor beta (PDGFRB), snail family transcriptional repressor 2 (SNAI2), or vimentin (VIM). CXCL12-positive MSCs in non-responder cBMBs had increased expression of osteo-committed markers such as osteopontin (SPP1), osterix (SP7), and collagens (Figure 1B, osteogenesis). Based on these findings, we formulated the hypothesis that the differentiation state of the BM stroma at diagnosis may indicate treatment outcomes, specifically at the initial 14-day post-induction therapy assessment. First, we determined the adipocyte- and chondrocyte-commitment states of BM MSCs in responder and non-responder AML cBMBs before and after the 14-day induction therapy. We found significant increases in transcripts linked to adipogenesis in responder cBMBs, suggesting an increase in frequencies of adipo-committed MSCs that represent the early differentiation state of these cells. This observation was consistent with our already reported findings (Treaba et al., British Journal of Haematology, 2022) that less differentiated MSCs are prevalent in responder cBMBs post-treatment. Intriguingly, we also noted significant differences between expression of SOX9-regulated transcripts in responder compared to non-responder cBMBs. As a master regulator of chondrogenesis, SOX9 is required for MSCs to differentiate towards a chondrogenic fate. SOX9 expression was significantly upregulated in non-responder post-treatment versus diagnostic cBMBs, with a log2 fold change of 3.6. Collagen 2 alpha 1 (COL2a1) and collagen 9 alpha 1 (COL9a1), two targets of SOX9 important in chondrogenesis, were also notably elevated in non-responder cBMBs with a log2 fold change of 6.1 and 7, respectively. In contrast, SOX6, which also regulates chondrogenesis, and COL2a1 were downregulated in responder post-treatment versus diagnostic cBMBs, with a log2 fold change of -2.7 and -4.5, respectively. Another striking change was seen when directly comparing the transcriptomes of responder to non-responder cBMBs at both diagnosis and post-treatment stages. Non-responder log2 expression of COL2a1 relative to responder log2 expression increased from -4.7 at diagnosis to 5.9 at post-treatment. Importantly, SOX9-regulated or -interacting genes were the most downregulated chondrogenic genes in responder cBMBs and the most upregulated in non-responder cBMBs, which is consistent with the contrasting differentiated state of MSCs between responders and non-responders initially observed with osteogenic genes. In summary, for the first time, we report that expression levels of SOX9 and SOX9-regulated targets in AML patients are indicative of the differentiated state of MSCs in AML patients who responded to or did not respond to 7+3 induction therapy. These data represent compelling support for our hypothesis correlating mesenchymal differentiation capacity with treatment response and warrant in-depth mechanistic studies.