Phosphoproteomics profiling reveals a kinase network conferring acute myeloid leukaemia intrinsic chemoresistance and indicates HMGA1 phosphorylation as a potential influencer

Dear Editor, The underlying mechanisms of cancer intrinsic drug resistance1 remain elusive. Herein, we report findings relevant to phosphoproteomics of acute myeloid leukaemia (AML) specimens. Specifically, we profiled phosphoproteins of cells fromAMLpatients undergoing chemo-failure comparedwith those achieving remission, and identify signatures associated with AML refractoriness. We collected bone marrow specimens at initial diagnosis from patients with comparable clinical characteristics (Tables S1 and S2); they exhibited either treatment failure (F) or reached complete remission (R) following ‘7 + 3’ induction therapy. We performed quantitative phosphoproteomics and total proteomics (Figure 1A), and found 9181 phosphorylation sites corresponding to 3001 phosphoproteins from phosphor-proteome, 4648 proteins from total-proteome (Figure 1B). Due to phosphopeptide distribution and technical consideration (Figure 1C), we focused on phosphor-serine/threonine. We confirmed high accuracy of the analysis (mass-errors/peptide length) (Figure S1A-C) and observed good reproducibility of phosphorylated and total peptides (Figure 1D; Figure S1D). A total of 20% (630) of 3001 phosphoproteins (of which 627 were up-regulated) and 3% (146) of the 4648 proteins were differentially seen inF relative to theR group (Figure 1E,F). Differential proteins/phosphoprotein criteria was applied as fold-change [F/R] > 1.5 or < 0.67 and p < 0.05). Unsupervised hierarchical clustering analysis to separate failure from remission samples indicated that chemofailure cases exhibited a distinct phosphoprotein signature (Figure 1G). Ingenuity pathway analysis (IPA) identified that DNA damage response pathway was top-ranked (Figure 2A).


L E T T E R T O E D I T O R
Phosphoproteomics profiling reveals a kinase network conferring acute myeloid leukaemia intrinsic chemoresistance and indicates HMGA1 phosphorylation as a potential influencer Dear Editor, The underlying mechanisms of cancer intrinsic drug resistance 1 remain elusive. Herein, we report findings relevant to phosphoproteomics of acute myeloid leukaemia (AML) specimens. Specifically, we profiled phosphoproteins of cells from AML patients undergoing chemo-failure compared with those achieving remission, and identify signatures associated with AML refractoriness.
We collected bone marrow specimens at initial diagnosis from patients with comparable clinical characteristics (Tables S1 and S2); they exhibited either treatment failure (F) or reached complete remission (R) following '7 + 3' induction therapy. We performed quantitative phosphoproteomics and total proteomics ( Figure 1A), and found 9181 phosphorylation sites corresponding to 3001 phosphoproteins from phosphor-proteome, 4648 proteins from total-proteome ( Figure 1B). Due to phosphopeptide distribution and technical consideration ( Figure 1C), we focused on phosphor-serine/threonine. We confirmed high accuracy of the analysis (mass-errors/peptide length) (Figure S1A-C) and observed good reproducibility of phosphorylated and total peptides ( Figure 1D; Figure S1D). A total of 20% (630) of 3001 phosphoproteins (of which 627 were up-regulated) and 3% (146) of the 4648 proteins were differentially seen in F relative to the R group ( Figure 1E,F). Differential proteins/phosphoprotein criteria was applied as fold-change [F/R] > 1.5 or < 0.67 and p < 0.05). Unsupervised hierarchical clustering analysis to separate failure from remission samples indicated that chemofailure cases exhibited a distinct phosphoprotein signature ( Figure 1G).
Ingenuity pathway analysis (IPA) identified that DNA damage response pathway was top-ranked ( Figure 2A).
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. ATM signalling was up-regulated in chemo-failure specimens, consistent with others. 2 FLT3, ERK/MAPK and Rho kinase signalling pathways were also enriched in F samples, as reported 3 ( Figure S2A). We next focused on upregulated phosphoproteins (627) to identify corresponding kinases. NetworKIN analysis showed that in F group, 250 up-regulated phosphor-proteins (confidence score > 2) 4 are potentially phosphorylated by 54 kinases. Those 250 substrates were more abundant than the remaining phosphoproteins, as evidenced by a shift in intensity ( Figure S2B). We further analysed the top 24 kinases (Table S3), of which each was responsible for >1% of total phosphorylation sites ( Figure 2B). Those top 24 kinases all function in cell cycle regulation ( Figure S2C). In addition to NetworKIN, KEA2 analysis revealed that activity of Casein Kinase II (CK2, gene name: CSNK2A1) or CDK family members was higher in F relative to R specimens ( Figure 2C; Table S4). Moreover, DEPMAP analysis revealed that depleting CDK family members in AML cell lines decreased cell viability (a gene with a score < -1 indicates an essential gene) ( Figure 2D). Not only identified as a top kinase from NetworKIN analysis (Table S3), CK2 was also the most enriched upstream kinase in chemo-failure samples through our IPA analysis ( Figure 2E). Accordingly, CK2 inhibition by CX-4945 significantly enhanced cytarabine-induced cell death in chemo-failure AML cells ( Figure 2F). Thus, we next focused on CK2 substrates in context of haematopoiesis.
We observed an overlap of 27 differentially phosphorylated proteins between the top two IPA terms relevant to haematopoiesis ( Figure 3A; Table S5). We then analysed the top 10 most abundant proteins of the 27 ( Figure 3B); among them, KIT, BCR, LAIR1 and RB1 are leukemic oncoproteins. 5 Interestingly, modified forms of HMGA1 phosphorylated at S99, S102 or S103 were among the top most abundant phosphoproteins in chemofailure AML samples ( Figure 3C; Table S6). HMGA1 S99/102/103 is highly conserved across species ( Figure 3D). Our mass spectrometry analysis in another set of samples verified HMGA1 hyperphosphorylation in chemofailure AML specimens ( Figure 3E,F; Table S7). Although HMGA1 function in leukemogenesis has been unclear, phosphorylation of HMGA1 S102 is reportedly catalysed by CK2. 6 We verify the activity by treating primary AML cells with CK2 inhibitor ( Figure S3A). It is also noteworthy that as a central hub of the CK2-substrate network ( Figure 3G), HMGA1 interacts with other proteins, such as SP1, which is a critical transcription factor responsible for aberrant expression of many genes which regulate cancer progression. 7 IPA analysis suggested that HMGA1 phosphorylation may promote cell survival ( Figure S3B). To test this, we used shRNA to knockdown (KD) HMGA1 in AML lines and observed markedly decreased cell growth ( Figure 4A; Figure S4A) and induction of G0/G1 cell cycle arrest ( Figure S4B,C). We also observed remarkable loss of phosphor-serine signals after mutating HMGA1 residues S99/S102/S103 (to phosphorylation-deficient S3A), confirming that they are the primary HMGA1 phosphorresidues ( Figure 4B). We then ectopically expressed HMGA1 constructs mimicking constitutive phosphorylation (S3D) or phospho-deficiency (S3A) form in MLL-AF9/FLT3-ITD (MA9/ITD) murine bone marrow cells to assess AML growth regulation ( Figure S4D). Enforced S3D expression enhanced CFC of MA9/ITD, while expression of S3A in cells did not ( Figure 4C; Figure S4E). Furthermore, serum-starved S3D-expressing cells showed enhanced survival relative to similarly treated MOCK cells, while S3A overexpression conferred no survival advantage ( Figure 4D). Finally, when we treated MA9/ITD cells with cytarabine (AraC), S3D cells showed some resistance relative to cells expressing S3A or MOCK ( Figure 4E-G).
Overall, we reveal that HMGA1 phosphorylation promotes intrinsic resistance and that blocking CK2-mediated HMGA1 phosphorylation may enhance cytarabine-based chemo-therapy. We also reveal a kinase signature predictive of chemoresistance and illustrate the importance of proteomics technology for better understanding cancer resistance.  Figure 3A, the rank is based on fold-change. (C) Phospho-proteomics screen for all the protein phosphorylation sites in eight AML specimens; pHMGA1 S99, S102 and S103 are marked in red.