In this issue of Blood, Pangallo et al compare the changes in splicing outcome in patients with rare mutations in splicing factors U2AF1 and SRSF2 to the changes due to common hotspot mutations. Many of these rare mutations phenocopy the common ones, suggesting that they have been evolutionarily selected to alter splicing and drive pathogenicity by similar mechanisms.1
Mutations in splicing factors are among the most prevalent mutations in myelodysplastic syndromes (MDSs) and acute myeloid leukemia (AML).2,3 There are as many as 300 proteins and 5 small RNAs associated with the spliceosome,4 yet only a handful are mutated in myeloid malignancies. In particular, the bulk of splicing factor mutations are found in 3 factors: SF3B1, SRSF2, and U2AF1.2,3 Additionally, within these factors, mutations are commonly found in hotspots resulting in changes to only 1 or 2 aa.
Initially, it was hypothesized that the splicing factor mutations would converge on common mechanisms to alter splicing and drive disease. Extensive molecular characterizations over the past several years revealed that splicing changes are factor specific.5 For U2AF1, which has 2 major hotspot alterations, S34F and Q157R, the splicing changes were even mutation specific.6 These findings suggest that not all mutations lead to the same outcome. Pangallo et al further explored this question by testing whether rare, nonhotspot mutations in SRSF2 and U2AF1 elicited the same or distinct splicing changes. They found that some mirrored hotspot changes, whereas others did not.
To explore the impact of rare mutations on splicing, Pangallo et al performed and analyzed RNA-sequencing data on cells expressing wild-type, hotspot, or rare mutant versions of SRSF2 and U2AF1. As the rare mutations were sometimes only observed in single MDS/AML patients, in addition to profiling primary patient material, they also generated human K562 cell lines that expressed wild-type, hotspot, or rare mutant versions of the factors.
Comparison of SRSF2 mutants unveiled remarkable overlap with most rare and hotspot mutations clustering together in their splicing pattern. Moreover, rare SRSF2 mutants alter exonic enhancer specificity, as was shown for hotspot mutations both in vivo and using biochemistry,7,8 suggesting similar splicing mechanisms. In contrast, comparison of U2AF1 containing rare and hotspot mutations showed more divergence. Some rare U2AF1 mutants alter 3′ splice-site sequence preference, as was shown for hotspot mutations,6 whereas others did not.
In many ways, the results that hotspot and rare mutations are similar are not surprising. The patients who were sequenced were selected by disease state. Thus, this in effect represents a genetic screen performed in humans by nature, with selective pressure for a highly specific phenotype. The results are reminiscent of the outcomes from many decades of yeast genetic screens for phenotypic splicing outcomes performed in the laboratory. It would be expected that mutations in splicing factors found in patients, no matter their location within those factors, would have similar outcomes on splicing, if splicing has anything to do with the disease phenotype. However, this has not been entirely clear. Indeed, the results from Pangallo et al provide a more compelling argument that it does for some splicing factors.
Although the outcomes for all of the SRSF2 mutants cluster together and support the model described herein, the more heterogeneous patterns observed with the U2AF1 mutants indicate a more complex situation. There also are “silent” mutations in both factors that do not significantly change splicing at all. It is unclear whether the clinical features of the patients with “silent” mutations are similar to those that phenocopy. If similar, these “silent” mutations could support the involvement of a nonsplicing function in disease etiology, which has been proposed by other groups.9,10
The results also raise the question of why mutations in hotspots are more frequent than those in rare positions. The answer, although not known, must be genomic context, likely the local DNA sequence and chromatin environment, both of which may influence mutation and DNA repair rates.
There are many spliceosome-associated factors whose function in splicing is still murky. As the current results demonstrate that messenger RNA changes from factor-specific mutations are quite canonical, could splicing pattern analysis in MDS/AML be used to identify factors with similar functionality as the major disease-associated splicing factors?
Overall, the work from Pangallo et al suggests that nonhotspot mutations should be considered similarly to common mutations, both in their mechanism and potential treatment.
Conflict-of-interest disclosure: The authors declare no competing financial interests.