Evolved Bmp6 enhancer alleles drive spatial shifts in gene expression during tooth development in sticklebacks

Abstract Mutations in enhancers have been shown to often underlie natural variation but the evolved differences in enhancer activity can be difficult to identify in vivo. Threespine sticklebacks (Gasterosteus aculeatus) are a robust system for studying enhancer evolution due to abundant natural genetic variation, a diversity of evolved phenotypes between ancestral marine and derived freshwater forms, and the tractability of transgenic techniques. Previous work identified a series of polymorphisms within an intronic enhancer of the Bone morphogenetic protein 6 (Bmp6) gene that are associated with evolved tooth gain, a derived increase in freshwater tooth number that arises late in development. Here, we use a bicistronic reporter construct containing a genetic insulator and a pair of reciprocal two-color transgenic reporter lines to compare enhancer activity of marine and freshwater alleles of this enhancer. In older fish, the two alleles drive partially overlapping expression in both mesenchyme and epithelium of developing teeth, but the freshwater enhancer drives a reduced mesenchymal domain and a larger epithelial domain relative to the marine enhancer. In younger fish, these spatial shifts in enhancer activity are less pronounced. Comparing Bmp6 expression by in situ hybridization in developing teeth of marine and freshwater fish reveals similar evolved spatial shifts in gene expression. Together, these data support a model in which the polymorphisms within this enhancer underlie evolved tooth gain by shifting the spatial expression of Bmp6 during tooth development, and provide a general strategy to identify spatial differences in enhancer activity in vivo.

Figures S1-S9 Table S1-S4 Supplemental Methods Supplemental Results Figure S1. Sequence alignment of marine and freshwater alleles of Bmp6 tooth enhancer Six core single nucleotide polymorphisms (green) concordant with the presence or absence of a large effect tooth number QTL lie upstream of a ~511 bp minimal Bmp6 tooth enhancer (start and end in yellow). Other polymorphisms (white) are not concordant with the presence or absence of the tooth QTL (Cleves et al. 2018).

Figure S2. A Col2a1 enhancer drives reporter expression in craniofacial cartilage and notochord in developing stickleback embryos. (A)
In a ten day post-fertilization embryo, reporter expression was observed in the notochord (n) and Meckel's cartilage (m) and (B) all other craniofacial cartilages including the palatoquadrate (pq), ceratohyal (ch), interhyal (ih), hyosympletic (hs), and ceratobranchials (cbs). Expression was also seen in the scapulocoracoid (sc), and otic vesicle (ov). The lens positive control domain driven by the Hsp70l promoter is marked with an asterisk. Scale bars = 500µm. n > 10 fish.

Figure S3. Insulator scoring scale in F0 injected fish
Examples of each insulator efficiency score. Fish were injected with a bicistronic construct (Col2a1a enhancer driving mCherry, Bmp6 intron 4 enhancer driving eGFP, separated by the mouse tyrosinase insulator) and domains were scored for insulator activity. A score of 0 was assigned for a domain in which both fluorophores were present (white arrowhead) throughout the same extent of the domain, indicating a lack of insulation. A score of 1 was assigned for a domain in which fluorophores were co-expressed in only a portion of the observed domain (white arrowhead), while there were also regions in which only a single fluorophore was observed (black arrowhead), indicating partial insulation activity. A score of 2 was assigned for a domain in which the predicted fluorophore was the only signal present (black arrowhead). White arrow heads indicate regions in which both fluorophores were observed, black arrowheads indicate regions in which only the predicted reporter was observed. n = 92 embryos. Figure S4. Marine and freshwater Bmp6 enhancers drive different spatial patterns in dorsal pharyngeal teeth. Dorsal pharyngeal tooth plates from fish doubly transgenic for two alleles of the Bmp6 intron 4 enhancer driving two different reporter genes (A,D): the marine enhancer driving mCherry with the freshwater enhancer driving eGFP (B,C) and the marine enhancer driving eGFP with the freshwater enhancing driving mCherry (E,F). Unilateral dorsal pharyngeal tooth plates (B,E) are shown, next to representative teeth from three stages (C,F): early, middle, and late highlighted by white boxes in B,E. (C,F) Early: freshwater enhancer drove expression in the epithelium (black arrowheads), with concentrated expression in the tip (asterisk), while the marine enhancer did not reliably drive expression in the epithelium, but was observed in the distal tip (F) in some instances. Both enhancers also drove expression in the mesenchyme (solid white arrowhead) with a larger expression domain of the marine allele (yellow dotted line) compared to the freshwater allele (orange dotted line). Middle: freshwater allele still drove expression in the epithelium while the marine allele was restricted to the distal tip. The marine allele drove more robust mesenchymal expression compared to the freshwater allele. Late: marine allele drives robust expression in the mesenchyme compared to freshwater allele in mineralized tooth (dashed line). Diagram: summary of tooth epithelial and mesenchymal domains. The relative sizes of green and magenta hatched lines correspond to the approximate relative strength of expression in the epithelium. Overlapping mesenchyme domain is gray, and expanded marine mesenchyme is marked with white arrowhead. Scale bars = 100µm (B,E), 50µm (C, F). n = 3 fish per genotype (6 total fish), >25 teeth per fish (298 total teeth).

Figure S5.
(A,B) Area of reporter gene expression in the mesenchyme (A) and epithelium (B) of teeth at three different tooth stages (early, middle, late). In both mesenchyme and epithelium, no significant differences were observed between any tooth stage for the marine:mCherry; freshwater:eGFP and marine:eGFP; freshwater:mCherry genotypes ("n.s." = not significant, P > 0.05, Wilcoxon rank sum two-tailed test)(C-D) Pooling all stages in a genotype also displayed the same trend with greater freshwater reductions in mesenchyme (C) and expansions in the epithelium (D) at late (> 20mm) fish stage than early (< 20mm) fish in both reciprocal genotypes. In C and D, three teeth per fish were included, so no statistical tests were performed as samples are not all independent. n = 3 fish per genotype per fish stage (12 total fish), 3 teeth per fish (36 total teeth). Figure S6. Differences in enhancer activity vary based on dorsal vs. ventral tooth field, fish total length, and epithelial vs. mesenchymal domain. (A) In < 20mm total length (pre-tooth number divergence) fish, the marine and freshwater alleles were expressed in the epithelium of all developing tooth germs regardless of genotype, while in > 20 mm total length (post-tooth number divergence) fish epithelial expression differences were consistent across tooth plates and genotypes. The freshwater allele consistently drove expression in all tooth germs scored, while the marine allele did not. Error bars show 95% C.I.s (B) The proportion of erupted teeth that demonstrated an observed mesenchymal bias of an expanded marine enhancer domain differed across dorsal and ventral tooth plates (dorsal and ventral, respectively), with more bias ventrally than dorsally. (C) Examples of erupted teeth (white dashed lines) from both dorsal and ventral tooth plates that were scored as either having a marine bias in the mesenchyme [if the freshwater enhancer mesenchymal domain (orange dotted line) was more restricted compared to the marine enhancer domain (yellow dotted line)], or no bias if the freshwater enhancer mesenchymal domain was equivalent to the marine enhancer domain. Scale bars = 50µm (C). n = 3 fish per genotype per fish size class (12 total fish), >50 teeth per fish (1108 total teeth, see Table S3). (A) In young, pre-hatching fish (6 dpf) the marine and freshwater enhancers drive expression in identical patterns in the developing fin margins of the pectoral fins (solid white arrowhead) and median fin (black arrowhead). (B) In adult caudal fins the more basal intersegmental joints were observed to have activity from both the marine and freshwater alleles (solid white arrowhead) while more distal joints were observed to only have freshwater enhancer activity (black arrowhead). The pattern was observed across both enhancer/reporter pairings. (C) Left pectoral fins from adults were observed to have activity from both enhancers in more basal intersegmental joints (solid white arrowheads) while only the freshwater allele was observed to have activity in more distal joints (empty arrowheads). Scale bars = 0.5 mm. n > 6 fish per genotype and >3 fish per stage.

Figure S8. Fin expression patterns of both alleles change over developmental time.
(A) Caudal and pectoral fins with the freshwater enhancer driving eGFP and marine enhancer driving mCherry. Only the freshwater enhancer is active in more distal joints (green arrowhead) while in more basal joints both enhancers are active (solid white arrowhead). No enhancer activity was observed in the most basal joints (black arrowhead). (B) Caudal and pectoral fins with the freshwater enhancer driving mCherry and marine enhancer driving eGFP. Similar to (A), the freshwater allele is active in more distal joints than the marine allele (purple arrowhead), more basal joints exhibit activity from both enhancers (solid white arrowhead). In the most basal joints, activity from either enhancer was not observed (black arrowhead). Scale bars = 0.5mm. n > 6 fish per genotype.

Figure S9. DAPI counterstain distinguishes between epithelial and mesenchymal tissues on thin sections.
Inner four columns show brightfield in situ hybridization (ISH) images for Bmp6 expression on marine (left) and freshwater (right) backgrounds, innermost columns with no annotations, adjacent to the same images with annotations (as presented in Figure 7). The outermost four columns show DAPI counterstains of the same sections, again shown both with and without annotations. The first row shows a cap stage tooth, the second row shows an early bell stage tooth, and the third row shows a late bell stage tooth. All dotted lines (black in brightfield images, white in DAPI images) demarcate the basalmost layer of epithelium in the tooth field, which is contiguous with the inner and outer dental epithelia of tooth germs. Regions where differences in expression were detected are marked with arrowheads: white arrowheads mark expanded mesenchymal expression in marine relative to freshwater, while black arrowheads mark expanded epithelial expression in freshwater relative to marine (as shown in Figure 7). Scale bar = 20 μm and applies to all panels. n = 6 fish per population, >10 teeth per fish. For each reporter positive domain in F0 fish with Col2a1a:mCh;Bmp6 tooth enhancer:eGFP transgene, a score of 0-2 was given for observed non, partial, or complete insulation. See Figure  S3 for examples of 0-2 scores. For each reporter positive domain in F0 fish with Col2a1a:mCh;Bmp6 tooth enhancer:eGFP transgene, a score of 0-2 was given for observed non, partial, or complete insulation. See Figure  S3 for examples of 0-2 scores.  For each tooth field (dorsal or ventral pharyngeal tooth plate, DTP or VTP), stage (predivergence = <20 mm fish length, post-divergence = >20 mm fish length, tooth stage [early or middle (mid), see Methods], the number (N), percentage (%) of detected mesenchymal bias in expression are listed, along with total number of teeth and genotype of transgene.

Supplemental methods
Multiple fluorescent reporter transgenes were assembled using the methods and primers as Primers used to amplify components of the Freshwater Bmp6 tooth enhancer:eGFP;marine Bmp6 tooth enhancer:mCherry insulator containing bicistronic construct

Scoring effectiveness of insulators
To assess insulator effectiveness, all surviving injected fish were raised to 7

Insulator effectiveness in bicistronic constructs
Insulator scores were not significantly different across injection clutches for the Col2a1a R2:mCherry; Bmp6 tooth enhancer:eGFP construct (Kruskal-Wallis left pectoral fin P = 0.075, right pectoral fin P = 0.52, median fin fold P = 0.116, Wilcoxon rank sum notochord P = 0.25), nor the Col2a1a R2:eGFP; Bmp6 tooth enhancer:mCherry construct (Wilcoxon rank sum left pectoral fin P = 0.144, right pectoral fin P = 0.134, median fin fold P = 0.211), suggesting that the inter-clutch variation did not have a significant impact on insulation scores. The left pectoral fin (P = 0.036) and the median fin fold (P = 0.016) were found to be significantly different between the two constructs while the right pectoral fin (P = 0.68) and notochord (P = 0.29) were not significantly different.

Marine enhancer activity in the epithelium differs across tooth stage and fish size
In post-tooth number divergence fish activity of the freshwater enhancer was observed in the epithelium in both ventral and dorsal tooth plates in all pre-eruption teeth

Ventral bias of evolved enhancer shifts
Quantification of epithelial and mesenchymal expression, and bias towards enhancer activity was scored for three tooth plates of each type (ventral and dorsal) at pre and post tooth number divergence. In post divergence fish, activity of the freshwater enhancer was observed in the epithelium in both ventral and dorsal tooth plates in nearly all pre-eruption teeth ( Figure S6A & Table S3). The marine allele was detected in the epithelium of only a subset of pre-eruption teeth, from approximately 70-90% of pre-eruption teeth in pooled tooth plate data ( Figure S6A).
When combining tooth plate data for each genotype the marine enhancer was active in the epithelium in a higher percentage of early stage germs compared to middle stage while there does appear to be a stage effect, variation also exists within stages. Overall, the freshwater enhancer drove expression more frequently and more robustly in the epithelium of early and middle stage teeth compared to the marine allele in post divergence fish. However, in pre-divergence fish, the epithelium of all pre-eruption teeth exhibited robust expression of both enhancers, across both genotypes and tooth plates ( Figure S6A).
A bias towards the marine allele in the mesenchyme was observed in nearly every early or middle stage tooth germ, while the lack of bias, or entirely overlapping mesenchymal expression, was almost exclusively observed in late stage (erupted) tooth germs ( . In early and middle stage teeth, we observed a consistent marine bias in the mesenchyme of both the ventral and dorsal tooth plates. In fully formed erupted teeth, a difference between the tooth plates became apparent. A larger proportion of erupted teeth were observed to have a marine bias in the mesenchyme in the ventral tooth plate compared to the dorsal tooth plate ( Figure S6B-C).
There was a reduction in the proportion of erupted teeth with a marine bias when comparing post to pre divergence fish for all integrations and tooth plates (pre-divergence