Somatic clones heterozygous for recessive disease alleles of BMPR1A exhibit unexpected phenotypes in Drosophila

The majority of mutations studied in animal models are designated as recessive based on the absence of visible phenotypes in germline heterozygotes. Accordingly, genetic studies primarily rely on homozygous loss-of-function to determine gene requirements, and a conceptually-related ‘two-hit model’ remains the central paradigm in cancer genetics. Here we investigate pathogenesis due to somatic mutation in epithelial tissues, a process that predominantly generates heterozygous cell clones. To study somatic mutation in Drosophila, we generated inducible alleles that mimic human Juvenile polyposis-associated BMPR1A mutations. Unexpectedly, four of these mutations had no phenotype in heterozygous carriers but exhibited clear tissue-level effects when present in somatic clones of heterozygous cells. We conclude that these alleles are indeed recessive when present in the germline, but nevertheless deleterious when present in heterozygous clones. This unforeseen effect, deleterious heteromosaicism, suggests a ‘one-hit’ mechanism for disease initiation that may explain some instances of pathogenesis associated with spontaneous mutation.


Introduction
Genomes are inherently unstable, an attribute that drives evolution at the population level but simultaneously underlies the incidence of spontaneous disease (Veltman and Brunner, 2012;Frank, 2014;Campbell et al., 2015;Acuna-Hidalgo et al., 2016;Forsberg et al., 2017;Stenson et al., 2017). In humans, DNA replication error is estimated to induce mutations at a frequency of 10 À9 per replicative cycle such that newborns already exhibit substantial somatic mosaicism at birth (Lynch, 2010;Frank, 2014;Campbell et al., 2015;Fernández et al., 2016;Forsberg et al., 2017;Machiela and Chanock, 2017). Additional somatic mutations accumulate over time, resulting in the age-dependent incidence of various pathologies, including cancer, X-linked disorders, and neurodevelopmental disease (Veltman and Brunner, 2012;Deng et al., 2014;Frank, 2014;Fernández et al., 2016;Machiela and Chanock, 2017). Indeed, it is estimated that spontaneous DNA replication error accounts for two-thirds of cancer mutations (Tomasetti et al., 2017). Nevertheless, little is known about the underlying cellular mechanisms by which somatic mutations trigger disease, primarily because of the technical difficulty of inducing specific mutations in endogenous loci during development of model animals.

Results and discussion
To investigate the phenotypic consequences of disease-associated alleles in vivo, previous studies have frequently employed methods that drive aberrant protein expression under the control of nonendogenous promoter elements. The Drosophila Gal4/UAS system, for example, provides a highly flexible platform to drive gene expression in vivo (Brand and Perrimon, 1993). This methodology can effectively model tissue mosaicism for putative disease alleles, but results can be difficult to interpret due to the reliance on gene overexpression through exogenous regulatory elements. Indeed, even wild-type alleles expressed at non-physiological levels can produce aberrant phenotypes ( Figure 1-figure supplement 1) (Rørth, 1996;Grieder et al., 2007;Akiyama et al., 2008).
To more rigorously investigate the effects of disease-associated mutations in vivo, we used CRISPR/Cas9-dependent genome editing to establish a Flippase (FLP) recombination-dependent allele switch system. Using human disease-associated alleles of the Drosophila Bone Morphogenetic Protein Receptor 1A (BMPR1A) homologue thickveins (tkv) as a model, our approach permits the inducible conversion of the wild-type locus to a specific mutant allele ( Figure 1A; Figure 1-figure supplement 2A). In Drosophila, signaling through BMPR1A/Tkv is required to control growth and direct cell fates in the developing wing imaginal disc. The primary ligand, Decapentaplegic (DPP), is expressed in a narrow band of cells at the anterior-posterior compartment boundary, leading to a downstream phospho-Mothers against dpp (p-Mad) activity gradient in the center of the disc that functions to establish presumptive wing vein regions ( Figure 1B-D) (Lecuit et al., 1996;Nellen et al., 1996;Affolter and Basler, 2007;Restrepo et al., 2014;Akiyama and Gibson, 2015b;Akiyama and Gibson, 2015a).
We first engineered a switchable control allele, tkv GFP>>mCherry , which carries tandem duplications of the wild-type tkv coding sequence. Each tkv copy is tagged with a distinctive fluorophore such that Green Fluorescent Protein (GFP) labels the 5' allele and mCherry labels the 3' copy. FRT sites flank the tkv-GFP coding exons to allow recombinogenic removal of tkv-GFP and thus a switch to expression of tkv-mCherry within the endogenous regulatory architecture ( Figure 1A To further validate the allele switch methodology, we generated a transgenic line that permits inducible in-locus expression of Tkv Q199D , a constitutively active form of the receptor (Hoodless et al., 1996). As expected, w; tkv GFP>>Q199D-mCherry /nub-Gal4; UAS-FLP/+ wing discs exhibited ectopic BMP/DPP activity throughout the Nub expression domain, resulting in both wing disc and adult wing abnormalities ( Figure 1H-J). Taken together, these results demonstrate the utility of an in-locus allele switch system for elucidating the in vivo consequences of spontaneous disease-linked mutations.
In humans, BMPR1A is known as a causative gene for Juvenile polyposis syndrome (JPS), a condition which predisposes patients to the development of gastrointestinal cancer ( Figure 2A) (Howe et al., 2001;Sayed et al., 2002;Greenman et al., 2007;Hardwick et al., 2008). Although several BMPR1A point mutations have been identified in JPS tissue samples, precisely how each lesion influences receptor activity remains unclear. To investigate this, we generated allele switch transgenic strains for five JPS-linked point mutations at sites that are conserved in Drosophila Tkv ( Figure 2A). In order to uniformly induce either hetero-or homozygosity throughout all cells of developing wing discs, larvae of tkv allele switch transgenics carrying hs-flp were subjected to a prolonged 1 hr heat shock at 72 hr after egg laying (AEL) (Figure 2B-L; Figure 2-figure supplement 1A and B). We found that these mutations fell into three functional classes: no effect, loss-of-function, and context-dependence. First, although the BMPR1A R443C mutation was detected in JPS patients (Sayed et al., 2002;Greenman et al., 2007), the corresponding tkv mutation produced no obvious phenotypes in either the hetero-or homozygous condition (Figure 2-figure supplement 1B-H). In a second allele class, we observed the expected loss-of-function effects. While . Similarly, residual wild-type clones within tkv C97R-mCherry mutant wing discs showed high levels of BMP activity and frequently produced distinctly rounded outgrowths (n = 6/8 discs; Figure 2N-Q).
In a third and final allele class, we found that the tkv C90R-mCherry variant possessed distinct signaling capabilities depending on the wing disc location. While tkv C90R-mCherry homozygosity resulted in weaker p-Mad expression in the wing pouch, the same mutation activated BMP signaling in the hinge region in both hetero-and homozygous discs ( Figure 2G,L and M; n = 35 and 32 discs, respectively). Altogether, these results indicate that JPS-associated tkv mutations present a variety of wing disc growth phenotypes linked to distinct effects on signaling activity. In addition, we identified an intriguing new allele, tkv C90R , which either promotes or represses signaling activity in a position-dependent manner.
Human disease alleles, such as those identified in JPS patients, can be categorized according to either spontaneous or inherited origins. Leveraging the allele switch system, we next sought to model the cellular etiology associated with somatic mutations in JPS-associated alleles ( . Strikingly, however, when we induced sporadic tkv C97R heterozygous clones at 72 hr AEL, experimental animals showed aberrant p-Mad expression in wing discs and patterning defects in adult wings ( Figure 3E-K). At the cellular level, clones of tkv C97R heterozygous cells exhibited strongly reduced BMP/DPP signaling activity and increased p-Mad levels in adjacent wild-type cells ( Figure 3I and J). These results demonstrate the unexpectedly detrimental effect of mosaic heterozygosity for a classically recessive allele, an effect we define as deleterious heteromosaicism.
To confirm that mosaicism itself was the cause of abnormal wing vein patterning, we more uniformly induced tkv C97R-mCherry heterozygous cells throughout the wing disc by increasing the duration of heat shock from 10 to 60 min. Indeed, longer duration heat shock increased the number of heterozygous cells present in developing discs, and rescued the wing vein phenotypes associated with heterozygous mosaicism ( Figure 3K). This indicates that deleterious heteromosaic effects derive from the clonal confrontation between heterozygous and wild-type cells.
Cell autonomous abrogation of BMP signaling typically leads to clone extrusion and apoptosis (Adachi-Yamada and O'Connor, 2002;Gibson and Perrimon, 2005;Shen and Dahmann, 2005). Consistent with these observations, the majority of homozygous tkv C97R clones were eliminated from the wing epithelia ( Figure 4A; Figure 3-figure supplement 1). However, neither form of cell elimination was observed in wing discs carrying tkv C97R heterozygous clones ( Figure 3F,G,I and J; Figure 4B; Figure 3-figure supplement 2). We infer that BMP signal reduction in heterozygous cells was adequate to cause deleterious effects but insufficient to trigger cell removal. Further, (M) Averaged p-Mad intensity plot profiles for wild-type tkv mCherry (n = 16), tkv C90R-mCherry (n = 15) and tkv M442T-mCherry (n = 16) homozygous wing discs. Figure 2 continued on next page consistent with a general phenomenon not restricted to tkv C97R-mCherry , heterozygous cell clones for other recessive tkv alleles elicited similar phenotypes (Figure 3-figure supplement 3). In agreement with previous results ( Figure 2G and L), we also found that clones heterozygous for tkv C90R-mCherry disrupted p-Mad activity in the developing wing blade region but ectopically activated BMP/ DPP signaling in the presumptive hinge territory (n = 26/27 discs; Figure 3-figure supplement 3G-L). Lastly, we also examined for the effects of deleterious heteromosaicism in other tissues. Although tkv C97R heterozygous clones influenced BMP/DPP activity in both eye and haltere discs at the level of Mad phosphorylation, the adult structures developed normally with no obvious defect ( Figure 3-figure supplement 4). These findings suggest a tissue-specific susceptibility to the same recessive mutations in heterozygous cell clones.
Recent advances in genome-wide association studies report a substantial accumulation of somatic mutations within individuals, resulting in genetic heterogeneity (Frank, 2014;Campbell et al., 2015;Fernández et al., 2016;Forsberg et al., 2017;Machiela and Chanock, 2017). This mutational load is associated with the initiation and progression of a number of diseases, including cancer and X-linked disorders (Veltman and Brunner, 2012;Deng et al., 2014;Fernández et al., 2016;Tomasetti et al., 2017). Here, we have established an inducible allele switch system that allows us to study specific disease-associated point mutations within endogenous loci (Figures 1 and 2 Since animals uniformly heterozygous for the same alleles are phenotypically normal, heteromosaic phenotypes likely emerge from local disparities between wild-type and heterozygous cells. Cellular heterogeneity associated with homozygous mutant clones is known to perturb tissue integrity by causing cyst formation, cell elimination and apoptosis (Adachi-Yamada and O'Connor, 2002;Gibson and Perrimon, 2005;Shen and Dahmann, 2005;Hogan et al., 2009;Bielmeier et al., 2016). In contrast, tkv deleterious heteromosaic clones showed no evidence of abnormal tissue architecture or cell death (Figures 3 and 4; Figure 3-figure supplements 2 and 3). Thus, although tkv heterozygous cells disrupted normal wing pattern formation, they did not strongly influence growth and appear to escape from homeostatic surveillance mechanisms such as cell competition (Di Gregorio et al., 2016).
A prevailing model in cancer genetics, the two-hit hypothesis requires that two independent mutations arise in a single tumor suppressor gene (Nordling, 1953;Knudson, 1971). Assuming mutagenesis by random DNA replication error, the probability of two independent mutations hitting the same gene within the same cell lineage is relatively low. Here, we report that a single somatic mutation in a putative tumor suppressor causes abnormal development by disrupting the homogeneity of BMPR1A-mediated cellular communication at the tissue level (Figures 3 and 4; Figure 3- figure supplements 2 and 3). Given that a number of disease-linked mutations are known in major signaling pathway components, we speculate that deleterious heteromosaicism is not limited to the BMP signaling pathway. Further, we propose that investigation into the unique facets of tissue

Generation of tkv allele switch transgenic lines
Two tkv sgRNA DNA constructs were generated using the primers listed in Supplementary file 1. tkv sgRNA DNA construct 1 (primers 1 and 2), and tkv sgRNA DNA construct 2 (primers 3 and 4), were annealed and cloned into the BbsI site of pBFv-U6.2 (Kondo and Ueda, 2013). A ubi-mCherry selection marker was generated as previously described (Akiyama and Gibson, 2015a). To obtain a donor DNA construct for generating a tkv allele switch founder transgenic line, six PCR fragments were prepared using primers 5 to 16 in Supplementary file 1. The PCR products were combined via Gibson assembly (NEB).
A DNA mixture containing two tkv sgRNA DNAs and the donor plasmid (250 ng/ml for each) was injected into the posterior side of embryos expressing Cas9 controlled by the nos promoter (Kondo and Ueda, 2013). Transgenic flies were selected by mCherry expression and further confirmed by DNA sequencing. Finally, the selection cassette was removed via Cre/loxP-mediated recombination, resulting in the tkv allele switch founder (see Figure 1-figure supplement 2A).
To generate tkv allele cassettes, a tkv-mCherry DNA fragment with 5' EcoRI and 3' AscI sites was constructed by Gibson assembly using primers 17 to 22 in Supplementary file 1. After Gibson assembly, a second round of PCR was conducted using primers 17 and 22 (Supplementary file 1). The resulting PCR product was cloned into pCRII blunt TOPO (Thermo Fisher Scientific).
To establish tkv* allele switch transgenic lines, w + attB tkv*-mCherry DNA constructs were injected into the posterior regions of embryos obtained from a cross of nos-phiC31 int. NLS (on X, Bloomington #34770) and the w; tkv allele switch founder. Transformants were screened by the presence of red eye color (see

Image analysis
To generate p-Mad intensity plot profiles, all images were collected at the same confocal setting and analyzed using the RGB profiler of FIJI. The 'Measure' function of FIJI was used to analyze sizes of wing discs. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.