Could Somatic Gene Rescue Explain Frequency of Compound Heterozygous and Concurrent Receptor Tyrosine Kinase Mutations in Hematological Diseases?

CSF3R plays a key role in regulation of myelopoiesis. In general, CSF3R mutations may be sub-grouped into two classes by receptor distribution with class I mainly consisting of missense mutations in the transmembrane domain (TMD) and seen in CNL and atypical CML. Class II are, instead mainly nonsense mutations in the cytoplasmic domain (CD), resulting in loss of carboxy-terminal domain, often associated with transformation to AML in the context of severe congenital neutropenia (SCN). Somatic (SM) gain of function (GOF) mutations following a hypomorphic mutation may represent a form of maladaptive somatic genetic rescue (SGR). By analogy to this “lesson of nature” we argue that in adult patients affected by SM CSFR3 mutations, a similar mechanism may be observed. The rescued germline (GL) alteration may be less pathogenic or recessive and thus their penetrance might be weaker with a longer disease anticipation, explaining their presentation later in life and in conjunction with more cryptic clinical presentation. We investigated this theory for GL and SM CSF3R variants and other receptor tyrosine kinase (RTK) variants.

We analyzed a total of 7312 cases from multiple institutions, and after exclusion of benign variants we found a total of 54 adult patients with assumed GL suspicious VUS and likely pathogenic (LP) CSF3R variants but confirmed pathogenic GL CSF3R variants were not found. We identified 65 patients with a total of 76 confirmed SM CSF3R variants, of which 36 were in the cytoplasmic domain (47%), 27 in the transmembrane domain (36%), and 13 in the extracellular domain (ECD) (17%). 52% of CD variants were found in AML and 14% in MDS, while 48% of TMD variants were found in MPN or MDS/MPN and 16% in MDS.

We investigated whether any of these SM mutations were associated with hypomorphic GL variants typical of SCN and variants of other RTK (FLT3, CSF1R, CSF2RA, CSF2RB, KIT). We found 3 cases of compound heterozygous GL CSF3R^T640I with CSF3R^R583H, and one case each of GL CSF3R^P782Qfs*6 with CSF3R^T618I, GL CSF3R^H599N with CSF3R^T618I, GL CSF2RB^V890I with CSF3R^T618I, GL CSF2RB^P508H with CSF3R^{Q741*, T618I} and GL ELANE^R81P with CSF3R^Q739*.

In total, 8/65 (12%) of adult patients with somatic CSF3R mutations had concurrent GL RTK or hypomorphic alterations. Of patients with compound heterozygous GL and SM CSF3R mutations, the most common somatic variants included R583H and T618I (4.5%, 3% of SM variants). Of note is that CSF3R^R583H was always in a compound heterozygous configuration.

GL CSF3R variants were located in ECD, (n=17), CD (n=26), and TMD (n=11). Variants with high likelihood of effect on receptor function (T640I, L723V, G687S) were more frequently observed in our cohort (12/7312), vs general population databases (gnomAD, 104/251,194, p=0.0001). Combined allelic burden of GL VUS/LP CSF3R variants in our cohort was similar to the general population (0.37% vs 0.44%, p=0.24) suggesting some variants may only affect myelopoiesis marginally. Higher frequency of cytopenia without overt myeloproliferation or dysplasia was observed with SM CSF3R alterations in CD compared to TMD (18% vs 0%, p=0.032), while lower frequency of cytopenia was observed with GL CSF3R alterations in CD compared to TMD (11% vs 46%, p=0.031). As compared to the rest of our cohort, GL CSF3R variants had more somatic mutations in SETBP1, CEBPA, NF1, WT1, while SF3B1 and TP53 mutations were less common (all p<0.05). Overall, cases of GL CSF3R concurrent with SM RTK alterations were enriched for AML/high risk MDS compared to GL CSF3R alterations alone (60% vs 7%, p=0.0017).

In total, 13/54 (24%) GL CSF3R variants were found to be associated with SM CSF3R (n=5), CSF2RB (n=3), CSF2RA (n=1), FLT3 (n=4), KIT (n=1) mutations. GL VUS variants of CSF2RB (n=7), CSF2RA (n=5), CSF1R (n=4), KIT (n=1) and FLT3 (n=1) were found, GL CSF2RB^P508H was concurrent with SM FLT3^V592A. Of note is that, 2 patients with GL CSF3R^R698C and CSF2RB^L276V acquired -7, none of which acquired any SM variants of RTK.

In summary, we present our findings showing that alterations in GL hypomorphic and RTK variants may present in patients who acquired other SM RTK mutations by analogy to SCN whereby the acquisition of concurrent mutations represents a maladaptive SGR. Patients with GL RTK mutations may constitute a suitable genetic milieu for selection of acquired hits that can reverse the underlying hypomorphic phenotype and proceed to develop myeloproliferative or dysplastic disorders.