Molecular, clinical, and prognostic implications of PTPN11 mutations in acute myeloid leukemia

Acute myeloid leukemia (AML) is the most commonly diagnosed acute leukemia in adults and is best characterized by the aberrant proliferation of clonal myeloid stem or progenitor cells with a differentiation block.¹  Although AML has a common myeloid origin, the pathogenesis is believed to be due to one or more genetic driver events such as chromosome translocations and/or gene mutations followed by the acquisition of mutations that promote the full phenotype of the disease. The complexity of the disease is further amplified by specific, age-associated disease characteristics. Recognition that AML is not one disease, but likely many, may explain why the cure rate remains very low, with similar chemotherapy given to all patients with this disease. Indeed, induction chemotherapy with an anthracycline plus cytarabine regimen followed by intensive consolidation without allogeneic stem cell transplant cures 35% to 40% of patients aged <60 years and 5% to 15% of patients aged ≥60 years.²  Despite a relatively frequent occurrence, one gene mutation in AML only recently characterized is mutation in the protein tyrosine phosphatase non-receptor type 11 (PTPN11) gene.

The PTPN11 gene encodes the protein Src homology region 2 (SH2)-containing protein tyrosine phosphatase 2 (SHP2). SHP2 is ubiquitously expressed and required for the normal development and function of hematopoietic cells.^3,4  SHP2 is composed of two SH2 domains at the N-terminal (sequentially labeled N- and C-terminal), a phosphatase (PTP) domain, and a C-terminal tail. The N-terminal SH2 (N-SH2) domain self-inhibits the PTP domain.^5,6  Upstream signaling recruits the N-SH2 domain and releases this self-inhibition to induce downstream signaling.^6,7  Oncogenic PTPN11 mutations induce prolonged SHP2 activation through the removal of self-inhibition.⁸ 

PTPN11 mutations have been found in various hematologic malignancies, including AML.^9-11  A PTPN11 mutation is found in ∼7% of patients with de novo AML and ∼12% of patients with therapy-related AML.^12-15  Given the recent emergence of primary resistance to targeted therapy such as ivosidenib, enasidenib, venetoclax, and entospletinib,^16-20  a reassessment of the associations of PTPN11 mutations with cytogenetic findings, mutations of other genes, clinical characteristics, and outcome features in AML patients treated with standard 7 + 3 chemotherapy is warranted. These analyses are necessary considering that many patients with AML still receive frontline chemotherapy, especially fit, younger patients. There is little information regarding how PTPN11 mutations affect prognosis of adult patients with AML in response to standard therapy or about associations with co-existing mutations and/or cytogenetic abnormalities. To our knowledge, ours is the largest study of PTPN11-mutated patients, in which we examine in detail the exact mutation sites and variant allele frequencies (VAFs) of PTPN11 mutations, chromosome abnormalities, co-occurring mutations in other genes, clinical features, and outcomes of adult patients with AML treated on clinical studies performed by the Cancer and Leukemia Group B (CALGB)/Alliance for Clinical Trials in Oncology (Alliance).

We analyzed the 1725 adults (≥17 years of age; age range, 17-92 years) with newly diagnosed, de novo AML (excluding acute promyelocytic leukemia) whose pretreatment bone marrow (BM) or blood samples underwent next-generation sequencing analysis.²¹  There were 1131 younger patients, defined as those aged <60 years and 594 older patients, defined as those ≥60 years of age. The patients were treated on CALGB trials with standard chemotherapy treatment as described in the supplemental Methods. CALGB is now part of the Alliance. Ninety-five percent of patients received intensive treatment, whereas 5% of patients received nonintensive treatment as described in the supplemental Methods. All patients were considered for outcome analyses, including those who experienced an early death, defined as death within 30 days of starting therapy irrespective of cause.

Patients provided written informed consent to participate in treatment studies and companion protocols. CALGB 8461 (cytogenetic studies), CALGB 9665 (leukemia tissue bank), and/or CALGB 20202 (molecular studies) involved collection of pretreatment BM and blood samples. Treatment protocols were in accordance with the Declaration of Helsinki and approved by the Institutional Review Boards at each center.

Cytogenetic analyses of pretreatment BM and/or blood samples were performed by institutional laboratories approved by CALGB/Alliance using unstimulated short-term (24- or 48-hour) cultures. Normal karyotype was determined in patients for whom at least 20 BM metaphase cells from a short-term culture were analyzed, and no clonal abnormality was found. Cytogenetic results were confirmed by central karyotype review.²² 

Viable cryopreserved BM or blood cells were stored for future analyses before starting treatment. Mononuclear cells from BM or blood were enriched by Ficoll-Hypaque gradient and cryopreserved in liquid nitrogen until thawed at 37°C for analysis. DNA extractions were performed by using the DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany). The mutational status of 80 protein-coding genes was determined centrally at The Ohio State University by targeted amplicon sequencing using the MiSeq platform (Illumina, San Diego, CA), as previously described²¹  and outlined in the supplemental Methods. Testing for the presence or absence of FLT3-ITD was performed as previously described.²³  In addition to the 80 genes analyzed by using the targeted amplicon sequencing panel, testing for CEBPA mutations was performed with Sanger sequencing as previously described,²⁴  thus resulting in a total of 81 genes whose mutational status was assessed in the current study. In accordance with the revision of the World Health Organization classification of myeloid neoplasms and acute leukemia and the European LeukemiaNet guidelines for AML,²⁵  only patients with biallelic CEBPA mutations were considered in the CEBPA-mutant category.

Definitions of clinical end points are provided in the supplemental Methods. Demographic and clinical features of any 2 patient groups were compared by using the Fisher’s exact test for categorical variables and Wilcoxon rank sum tests for continuous variables. The Kaplan-Meier method was used for estimating probabilities of overall survival (OS), disease-free survival (DFS), and event-free survival (EFS) and differences between survival distributions were tested by using the log-rank test.²⁶  We used logistic regression for modeling complete remission (CR) attainment, Cox proportional hazards regression for modeling DFS and OS for univariable and multivariable outcome analyses, and P values adjusted to control for per-family error rate. For the multivariable analysis, a limited backward selection technique was used to build the final model. Variables considered in the multivariable model were significant at the likelihood ratio test–adjusted P value < .20 from the univariable models. All statistical analyses were performed by the Alliance Statistics and Data Center, and SAS 9.4 (SAS Institute, Inc., Cary, NC) was used. The database was locked on June 9, 2020.

Of the 1725 patients with AML examined, the presence of a PTPN11 mutation was detected in 140 (8.1%) patients, which is comparable to the reported mutation frequency in other studies.^12,27  There were 98 younger patients and 42 older patients. The median follow-up of patients still alive was 9.0 years. There was a wide range of VAFs for PTPN11 mutations among patients, ranging from 0.05 to 0.54, with 59 (42%) patients having a VAF above 0.30 (Figure 1).^28,29  The majority of the mutations (61%) were localized in the N-SH2 domain, a known PTPN11 mutation hotspot location that is associated with increased SHP2 activity, whereas a minority of mutations were in other portions of the gene, such as the PTP domain (Figure 2).^27-30  With regards to pretreatment clinical characteristics, patients with mutated PTPN11 (PTPN11^mut) presented more often with higher platelet counts (median, 72 vs 54 × 10⁹/L; P < .001) and were more likely to have extramedullary involvement (33% vs 24%; P = .03) compared with PTPN11 wild-type (PTPN11^wt) patients (Table 1). All other clinical features of patients with PTPN11^mut were similar to those of patients with PTPN11^wt.

Cytogenetic findings at diagnosis are important factors affecting the outcome for patients with AML.^31,32  In our study, patients with PTPN11^mut more commonly had a normal karyotype (61% vs 45%; P < .001) or inv(3)(q21q26)/t(3;3)(q21;q26) (5% vs 1%; P = .004) than patients with PTPN11^wt. The latter finding was especially striking because as many as 26% (7 of 27) of patients with inv(3)/t(3;3) harbored a PTPN11 mutation, as previously reported.³³  Moreover, all 7 of these patients also had abnormalities in chromosome 7, including ‒7 in six and a deletion of the short arm of chromosome 7 [del(7)(p13p15)] in one patient. In contrast, PTPN11 mutations were less commonly observed in patients with a typical complex karyotype (3% vs 8%; P = .04)³⁴  and in those with core-binding factor AML. There were no PTPN11^mut patients with t(8;21)(q22;q22) (0% vs 100%; P = .005), and only 2% of patients with PTPN11^mut harbored inv(16)(p13;q22)/t(16;16)(p13;q22) compared with 7% of patients with PTPN11^wt (P = .03) (supplemental Table 1). For other cytogenetic abnormalities, there were no significant associations with PTPN11 mutations.

In addition to cytogenetic findings at diagnosis, recurrent gene mutations have emerged as important factors affecting the outcome of patients with AML.²⁵  Previous studies focusing on PTPN11-mutant AML mainly included a limited number of recurrently mutated genes, whereas 2 very recently published papers and our own study examined a broader mutation panel relevant to AML.^27,35  We noted that patients with PTPN11^mut have a higher mutation rate (median number of mutations, 4 vs 3; P < .001) than PTPN11^wt patients, albeit this finding is based on a targeted sequencing panel. An oncoprint of the 140 patients with PTPN11 mutations shows the co-occurring gene mutations (Figure 1).^28,29  Notably, patients with PTPN11^mut more frequently harbored NPM1 (61% vs 31%; P < .001), DNMT3A (39% vs 22%; P < .001), and STAG2 (6% vs 3%; P = .04) mutations than those with PTPN11^wt. In a similar fashion, patients with PTPN11^mut less frequently had double-mutated CEBPA (1% vs 8%; P = .003), KIT (1% vs 5%; P = .04), ZRSR2 (1% vs 5%; P = .04), and TP53 (4% vs 8%; P = .05) mutations (supplemental Table 2).

Because PTPN11 mutations tend to cluster in the N-SH2 and PTP domains, which are both involved in SHP2 self-inhibition, we questioned whether mutations in different domains of the PTPN11 gene resulted in comparable pretreatment clinical characteristics. Eighty-six patients had N-SH2 domain mutations, and 45 patients had PTP domain mutations. We found that the only difference at baseline was that patients with N-SH2 mutations had a higher percentage of blasts in the BM (median, 65% vs 52%; P = .03) (Table 2). There were no significant differences in distribution of cytogenetic findings between N-SH2 and PTP PTPN11-mutated patients (supplemental Table 3). Patients with N-SH2 mutations were less likely to have GATA2 (0% vs 7%; P = .04) and PLCG2 (0% vs 7%; P = .04) mutations than patients with PTP mutations (supplemental Table 4).

We compared clinical outcomes of patients with and without PTPN11 mutations both in the entire patient cohort and, separately, in younger and older patients. There were no significant differences in CR, early death rates, or DFS, OS, and EFS between patients with PTPN11^mut and PTPN11^wt in the entire cohort (supplemental Table 5). We then stratified patients into two age groups, those aged <60 years and those aged ≥60 years, because these patients were treated differently on CALGB/Alliance protocols. Although the presence of PTPN11 mutations did not associate with significant differences in early death rates, CR rates, OS, or EFS in either older or younger patients, older patients harboring a PTPN11^mut had a shorter DFS (3-year rates, 5% vs 15%; P = .05) than PTPN11^wt patients (supplemental Table 6).

We also studied whether mutations in different domains of the PTPN11 gene affected patients’ outcomes. The only difference we detected was that 20% of patients with the PTPN11 mutations located in the N-SH2 domain died early as opposed to no early deaths among patients with PTPN11 mutations in the PTP domain (P < .001) (supplemental Table 7). There were no significant differences in CR rates or DFS, OS, and EFS between the 2 groups (Figure 3). In addition, higher early death rates (P = .02), but no other significant outcome differences, were found in younger patients with N-SH2 domain PTPN11 mutations compared with those with a mutation in the PTP domain. In the older age group, there were no significant differences in outcome (supplemental Table 8).

Given that 85 (61%) of the 140 patients with PTPN11^mut also harbored an NPM1 mutation, we next sought to determine if the clinical and molecular features differed between NPM1-mutated patients with or without PTPN11 mutations. With regard to pretreatment characteristics, patients with NPM1^mut/PTPN11^mut had a higher baseline platelet counts (median, 78 vs 59 × 10⁹/L; P = .008) (supplemental Table 9). Distribution of cytogenetic aberrations was similar between the 2 groups (supplemental Table 10). Notably, patients with NPM1^mut/PTPN11^mut had a higher frequency of DNMT3A mutations (56% vs 43%; P = .03), whereas FLT3-ITD (19% vs 44%; P < .001) was less frequent in this genomic group compared with patients with NPM1^mut/PTPN11^wt (supplemental Table 11). This suggests NPM1^mut/PTPN11^mut clones are less dependent on additional signaling mutations such as FLT3-ITD. Despite these differences in baseline biology, there were no significant differences in any of the outcome end points between NPM1-mutated patients with and those without PTPN11 mutations regardless of age (Figure 4A; supplemental Tables 12 and 13).

We were also interested if PTPN11 mutations can influence outcomes of patients with NPM1^wt. A comparison of pretreatment characteristics between patients with PTPN11^wt and PTPN11^mut revealed no significant differences (supplemental Table 14). Cytogenetically, patients with NPM1^wt/PTPN11^mut were more likely to harbor prognostically unfavorable inv(3)(q21q26)/t(3;3)(q21;q26) (13% vs 2%; P < .001), other balanced rearrangements involving 3q26 (4% vs 0.2%; P = .01), and t(11;19)(q23;p13.3)/KMT2A-MLLT1 (4% vs 0.4%; P = .03) (supplemental Table 15) compared with patients with NPM1^wt/PTPN11^wt. Moreover, patients with NPM1^wt/PTPN11^mut had a higher median number of mutations (3 vs 2; P < .001) than those with NPM1^wt/PTPN11^wt and were more likely to have KMT2A (7% vs 1%; P = .006) and NF1 (21% vs 6%; P = .01) mutations (supplemental Table 16).

Among combined younger and older patients with NPM1^wt, those with PTPN11^mut had a lower CR rate (36% vs 61%; P < .001) and shorter EFS (3-year rates, 9% vs 19%; P = .003) than patients with PTPN11^wt (supplemental Table 17). Likewise, younger patients with NPM1^wt/PTPN11^mut had a lower CR rate (45% vs 71%; P = .002), OS (3-year rates, 30% vs 41%; P = .04) (Figure 4B), and EFS (3-year rates, 13% vs 27%; P = .008), but not DFS, than those with NPM1^wt/PTPN11^wt. Older patients with NPM1^wt/PTPN11^mut also had a lower CR rate (18% vs 43%; P = .04), DFS (3-year rates, 0% vs 10%; P = .02), and EFS (3-year rates, 0% vs 4%; P = .02), but not OS, compared with patients with NPM1^wt/PTPN11^wt (Table 3).

Multivariable analyses were performed to determine what other factors, including gene mutations, associated with inferior outcomes of AML patients with NPM1^wt. We could not perform separate multivariable analyses in younger and older patients because there would have been too few patients to obtain meaningful results. In the multivariable modeling for CR attainment, mutations in PTPN11, TP53, and FLT3-ITD and age remained in the final model (Table 4), indicating that PTPN11 mutations still affect the probability of CR achievement even when accounting for other variables (P < .001). However, in the multivariable analyses of OS and EFS, PTPN11 mutations did not remain significant in the final models.

Herein, we showed that PTPN11 mutations may affect clinical outcomes dependent on age group and mutation subset analyses in a retrospective study of patients with AML receiving intensive therapy in clinical trials performed by the CALGB/Alliance. Although the presence of a PTPN11 mutation in addition to an NPM1 mutation did not associate with poorer outcomes (with the exception of older patients with PTPN11^mut having a marginally reduced DFS compared with patients with wild-type PTPN11), PTPN11 mutations did associate with inferior outcomes in AML patients with NPM1^wt regardless of age. We also found that patients with PTPN11 mutations in the N-SH2 domain had higher BM blast counts and early death rate than those with PTP domain mutations. These results suggest that an N-SH2 mutation might generate a different phenotype. We hypothesize this phenotype could be immunosuppressive, explaining the higher early death rate but no difference in response to chemotherapy induction, DFS, OS, or EFS. Collectively, our study outlines the complex effects of a PTPN11 mutation in AML and provides evidence that its prognostic impact should be considered in the context of NPM1 mutation status.

Similar to others, we also found an association between PTPN11 mutations and inv(3)(q21q26)/t(3;3)(q21;q26), the later aberration being a marker of poor prognosis in AML.^25,33  We also confirmed that PTPN11 mutations are less likely to occur in patients with typical complex karyotype and those with core-binding factor AML²⁷  but are most often found together with NPM1 mutations.^27,36  We also found an association between PTPN11 mutations and mutations in DNMT3A or STAG2. Furthermore, we observed that there were few patients with PTPN11^mut who also had co-mutations in CEBPA, KIT, TP53, and ZRSR2. Our analysis of co-occurring mutations in patients with NPM1^mut/PTPN11^mut revealed that these patients had a lower frequency of co-occurring FLT3-ITD mutations, suggesting that PTPN11 and FLT3-ITD mutations result in activation of overlapping signaling pathways.

There are few published data regarding how PTPN11 mutations affect clinical outcomes. Hou et al³⁷  and Swoboda et al³⁵  have shown that patients with NPM1^wt/PTPN11^mut had reduced OS compared with patients with NPM1^wt/PTPN11^wt, and Alfayez et al²⁷  reported that PTPN11 mutations are associated with poor outcomes for both de novo and relapsed/refractory AML. Our current study validates these findings but also goes further by analyzing a larger cohort of patients, which allowed us to stratify the patients according to age. Furthermore, our study analyzed associated mutations and questioned how the location of the mutation within the PTPN11 gene affected outcome.

A limitation of our study is the time span over which these patients were treated and the fact that, on these clinical trials, patients received only intensive induction followed by consolidation chemotherapy. Supportive care for AML has clearly improved over time with the addition of more effective proton pump inhibitors, antifungal agents, and transfusion support. In addition, patients with FLT3 mutations on this study typically did not receive midostaurin. Among AML patients with NPM1^wt, 15% of patients with PTPN11^mut also harbored FLT3-ITD, raising a possibility that inferior outcomes in PTPN11-mutated patients could be associated with FLT3-ITD. However, both PTPN11 mutations and FLT3-ITD stayed in the multivariable model, suggesting that they negatively affect outcomes independently from each other. Hence, we believe our findings are relevant to the current era of AML therapy, and moving forward, it will be important to study how PTPN11 mutations affect responses to the newly approved targeted therapies, as early evidence suggests that these patients might be resistant.^16-20,38  More clinical studies and basic science research are needed to understand how SHP2 and NPM1 proteins are interacting and why PTPN11 mutations are associated with worse outcome in AML patients with NPM1^wt.

The authors thank the patients who participated in clinical trials, Christopher Manring and the CALGB/Alliance Leukemia Tissue Bank at The Ohio State University Comprehensive Cancer Center for sample processing and storage services, and Lisa J. Sterling for data management. The visual abstract was created with BioRender.com.

This work was supported by the National Cancer Institute, National Institutes of Health, under grants U10CA180821, U10CA180882, and U24CA196171 (to Alliance for Clinical Trials in Oncology), and UG1CA233180, UG1CA233191, UG1CA233331, UG1CA233338, UG1CA189850, and R35CA198183; Pelotonia; ASH Scholar Award; Leukemia & Lymphoma Society; D. Warren Brown Foundation; Harry Mangurian Foundation; and Pelotonia Fellowship Graduate Program. Support to Alliance for Clinical Trials in Oncology and Alliance Foundation Trials programs is listed at https://acknowledgements.alliancefound.org.

Contribution: S.F., E.H., and J.C.B. conceived and designed the study; S.F., K.M., E.H., and J.C.B. drafted the manuscript; J.K., H.G.O., and D.N. analyzed data; J.C.B. obtained funding for this study; E.H. and J.C.B. supervised this study; and all authors contributed to the acquisition, analysis, and interpretation of these data and were critical in manuscript revision.

Conflict-of-interest disclosure: J.C.B. is a paid consultant for Syndax, Trillium, AstraZeneca, Novartis, and Kronos; and chair of the scientific advisory board and a major stockholder in Vincerx Pharma. J.S.B. consults for AbbVie, AstraZeneca, KITE Pharma, and INNATE Pharma. R.M.S. serves on the advisory board for AbbVie, Actinium, Arog, BMS, Boston Pharmaceuticals, Janssen, Jazz, Novartis, Syros, Takeda, Elevate Bio, Syndax Pharma, Gemoab, Foghorn Thera, GSK, Aprea, and OncoNova; is a part of the Steering Committee for AbbVie and the AML Expert Council for GSK; and serves on the data safety monitoring board for Takeda and Syntrix/ACI Clinical. E.S.W. has received consulting fees from AbbVie, Astellas, BMS, Genentech, GlaxoSmithKline, Jazz, Kite, Kura Oncology, Novartis, Mana Therapeutics, Pfizer, Stemline, and Takeda; serves on the speakers bureau for Stemline, Kura, Pfizer, and Dava Oncology; and serves on the data safety monitoring committee for AbbVie, Rafael Pharmaceuticals. B.L.P. has clinical trial funding from Ambit Biosciences, Hoffmann–La Roche, Jazz Pharmaceuticals, Novartis, Pfizer, and Rafael Pharmaceuticals; and consults for Rafael Pharmaceuticals. The remaining authors declare no competing financial interests.

Correspondence: John C. Byrd, Department of Internal Medicine, University of Cincinnati College of Medicine, 231 Albert Sabin Way, ML 0551, Room 6065, Cincinnati, OH 45267-0551; e-mail: byrd2jc@ucmail.uc.edu.