Abstract
Linkage and association studies have mapped thousands of genomic regions that contribute to phenotypic variation, but narrowing these regions to the underlying causal genes and variants has proven much more challenging. Resolution of genetic mapping is limited by the recombination rate. We developed a method that uses CRISPR to build mapping panels with targeted recombination events. We tested the method by generating a panel with recombination events spaced along a yeast chromosome arm, mapping trait variation, and then targeting a high density of recombination events to the region of interest. Using this approach, we fine-mapped manganese sensitivity to a single polymorphism in the transporter Pmr1. Targeting recombination events to regions of interest allows us to rapidly and systematically identify causal variants underlying trait differences.
Identification of DNA sequence differences that underlie trait variation is a central goal of modern genetic research. The primary tools for connecting genotype and phenotype are linkage and association studies. In these studies, co-inheritance of genetic markers with the trait of interest in large panels of individuals is used to localize variants that influence the trait to specific regions of the genome. The localization relies on meiotic recombination events that break up linkage between markers on a chromosome. Therefore, the spatial resolution of genetic mapping is limited by the recombination rate. In practice, the recombination rate in most settings is too low to resolve the mapped regions to individual genes, much less to specific variants within genes. Increasing mapping resolution requires construction of ever-larger panels of individuals and/or additional generations of recombination, and these approaches are laborious to the point of often being impractical. As a consequence, the genes and variants underlying trait variation have yet to be identified for the vast majority of regions implicated by linkage or association mapping.
To address this problem, we have devised a new method for genetic mapping that precisely targets recombination events to regions of interest. The method uses recombination events that occur during mitosis rather than meiosis. Rare mitotic recombination events occur naturally when a chromosomal double strand break (DSB) is repaired by homologous recombination (HR) that leads to the formation of a recombined chromosome (Yin and Petes 2013). In a heterozygous individual, mitotic cell division can then generate daughter cells with a new genotype that is completely homozygous from the recombination site to the telomere and unchanged heterozygous everywhere else (Fig. 1A); such an event is termed “loss of heterozygosity” (LOH). Individuals with LOH events at various locations in the genome have been used to construct a genetic map (Henson et al. 1991), and this and related approaches (Laureau et al. 2016) can, in principle, be used to map the genetic basis of trait variation (Fig. 1B). However, this approach has been limited in practice by the very low frequency of natural mitotic recombination events.
We have leveraged the CRISPR-Cas9 system to produce targeted mitotic recombination events at high frequency and at any desired location, allowing facile construction of LOH-based mapping panels. In the CRISPR (clustered, regularly interspaced, short palindromic repeats) system, the bacterial endonuclease Cas9 is directed to create a DSB at a site specified by a variable targeting sequence of a bound guide RNA (gRNA) (Doudna and Charpentier 2014). Successful cutting also requires the targeted sequence to be followed by an invariant protospacer-adjacent motif (PAM). An LOH event in a heterozygous diploid individual can be generated by cutting only one chromosome, leaving its homolog intact to serve as a template for repair by HR. This is accomplished by using polymorphic heterozygous PAM sites that are present on only one of the two chromosomes.
In order to demonstrate that LOH events can be targeted to precise genomic loci using CRISPR, we designed 95 gRNAs targeting Streptococcus pyogenes Cas9 to sites distributed across the left arm of S. cerevisiae chromosome 7 (Chr7L). The gRNAs targeted heterozygous sites in a diploid yeast strain generated by crossing a lab strain (BY) and a vineyard strain (RM), using PAMs polymorphic between the two strains. After cutting, repair, and mitosis, cells in which the DSB repair led to an LOH event were isolated by fluorescence-activated cell sorting (FACS) through their loss of a telomere-proximal green fluorescent protein (GFP) gene. We picked approximately four GFP(-) lines per targeted site, for a total of 384 lines. Genotyping by low-coverage whole-genome sequencing demonstrated that CRISPR-induced recombination was highly effective, with LOH events in more than 95% of lines and few off-target effects (Supplementary Methods). Most LOH recombination events occurred within 20 kb of the targeted cut site (Fig. 2), consistent with previous measurements of LOH gene conversion tract length (St. Charles and Petes 2013). LOH events were successfully generated at sites across the entire length of the targeted chromosome arm (Fig. 2), demonstrating that our method is not limited to certain genomic contexts.
We next used the LOH panel to map quantitative traits to loci on Chr 7L. We measured growth of each of the 384 LOH lines in 12 different conditions, chosen because we previously mapped quantitative trait loci (QTLs) for growth in these conditions to Chr 7L. In parallel, we measured growth of 768 segregants from a cross between BY and RM (Bloom et al. 2013). One of the traits, growth on 10 mM manganese sulfate, mapped to a large-effect QTL with a maximum LOD score of 109.4 in the LOH panel (Fig. 3). The confidence interval obtained with the 384 LOH lines overlapped with and was narrower (2.9 kb) than that obtained with 768 segregants (3.9 kb). The LOH-based interval contained two genes and 12 polymorphisms between BY and RM. We identified concordant QTLs of smaller effect for eight other traits in the two panels (fig. S1). Two traits mapped to a QTL of small effect size in just one panel, likely due to low statistical power (fig. S2). One trait lacked a Chr 7L QTL in both panels.
To rapidly fine-map the causal variant for manganese sensitivity, we generated a second panel of LOH lines whose recombination events were all targeted to the mapped manganese sensitivity interval. We took advantage of the fact that LOH gene conversion tracts vary in length, which means that in different individuals, DSBs generated by the same gRNA can lead to slightly different LOH crossover sites, typically within 10 kb of the DSB (St. Charles and Petes 2013). We isolated 358 GFP(-) lines generated with three gRNAs targeting sites near the mapped interval. We genotyped this panel by sequencing and observed that 46 of the lines (13.1%) had a recombination event within the 2.9 kb QTL interval; together, the recombination events separated almost all the variants in the interval (Fig. 4A). In contrast, only 0.7% of segregants had recombination events in the interval (Bloom et al. 2013). To obtain a comparable number of recombination events at this locus by random meiotic segregation, a segregant panel would require more than 7,500 lines. Thus, with targeted LOH events, we can in effect generate unnaturally strong recombination hotspots at any region of interest (fig. S3).
We measured manganese sensitivity in this fine-mapping panel (Fig. 4B). Comparison of the panel phenotypes with the breakpoint locations pinpointed a single polymorphism as responsible for increased sensitivity in BY. The polymorphism changes a phenylalanine in BY to a leucine in RM at position 548 of Pmr1, a manganese transporter. Six lines had recombination events between Pmr1-F548L and the closest polymorphism to the right, 402 bp away, and were either fully sensitive or resistant to manganese, depending on which Pmr1-F548L allele was homozygous in the line. One line had a recombination between Pmr1-F548L and the closest polymorphism to the left, 125 bp away, and showed the intermediate manganese sensitivity phenotype expected for a heterozygote at the causal variant. LOD score analysis of the fine-mapping panel also identified a support interval containing only Pmr1-F548L (Fig. 4B).
To directly test the effect of Pmr1 variants on manganese sensitivity, we individually engineered the RM alleles of Pmr1-F548L, the two neighboring polymorphisms, as well as the two remaining nonsynonymous Pmr1 polymorphisms into BY. Variant replacements were carried out in a single step by CRISPR-directed DSB formation combined with repair off a supplied template carrying the desired allele. As expected based on the LOH fine-mapping, changing phenylalanine-548 to leucine made BY significantly more resistant to manganese, whereas none of the other four polymorphisms queried had a significant effect on manganese sensitivity (Fig. 5).
PMR1 encodes an ion pump that transports manganese and calcium into the Golgi (Culotta et al. 2005). Pmr1 is a member of the P-type ATPase family of ion and lipid pumps that are found in all branches of life, and many other P-type ATPases have a conserved leucine at the position homologous to phenylalanine-548 of Pmr1. The NMR structure of the rat sodium pump (Hilge et al. 2003) and the crystal structure of the rabbit calcium pump (Toyoshima and Mizutani 2004) show a leucine at the homologous position making direct contact with ATP (fig. S4). Furthermore, mutating the homologous leucine of the rabbit calcium pump to phenylalanine decreases function by affecting ATP binding (Clausen et al. 2003). Thus, the F548L polymorphism is expected to reduce the ability of Pmr1BY to transport manganese into the Golgi, relative to Pmr1RM, consistent with the observed manganese sensitivity of BY.
Pmr1 leucine-548 is conserved across fungi, with a few species having an isoleucine or valine at the homologous position, and none having phenylalanine (fig. S5). In the S. cerevisiae population, almost all sequenced PMR1 alleles have leucine-548, with phenylalanine-548 found only in BY and other laboratory yeast strains (Liti et al. 2009; Song et al. 2015) whose PMR1 alleles are likely directly related to BY (Schacherer et al. 2009). This region of BY Chr 7 is inherited from EM93, a diploid yeast strain isolated from a fig in California in 1938 (Mortimer and Johnston 1986). Sequencing of PMR1 in EM93 revealed that EM93 is heterozygous for Pmr1-F548L (fig. S6), suggesting either that the mutation is not laboratory-derived or that it occurred between the isolation of and its entry into a stock collection.
Decades of mapping studies have uncovered loci for myriad traits, but identification of the underlying genes and variants has lagged. We developed a new CRISPR-assisted mapping approach that promises to close this gap. Our approach uses CRISPR to generate targeted recombination events densely covering a region of interest of any size. We first generated CRISPR-directed LOH events that span a yeast chromosome arm, and used the resulting lines to map trait variation. We then saturated a QTL interval with recombination events to rapidly identify a single causal polymorphism responsible for manganese sensitivity in laboratory strains of S. cerevisiae. We confirmed the causal role of this polymorphism by direct CRISPR-assisted engineering of the resistance allele into a sensitive strain. In contrast to previous strategies, our method generates a much higher density of recombination events, is easily targetable to any region of the genome, and does not require extra time-consuming generations of crossing to increase recombination frequency.
We anticipate that trait mapping with targeted LOH panels will greatly aid efforts to understand the genetic basis of trait variation. In addition to many applications in single-celled organisms, LOH panels could be generated from cultured cells, enabling in vitro genetic dissection of human traits with cellular phenotypes. Mapping resolution in multicellular organisms could be enhanced with CRISPR-directed meiotic recombination events. Indeed, the mutagenic chain reaction system developed in vivo in fruit flies (Gantz and Bier 2015) and mosquitos (Gantz et al. 2015; Hammond et al. 2015) uses CRISPR to generate gene conversion events in meiosis with extremely high efficiency. Another approach would be to stimulate LOH during early development, generating chimeric individuals. Another approach would be to stimulate LOH early in development, generating chimeric individuals. The targeted LOH method also has the exciting potential to be applied to viable interspecies hybrids that cannot produce offspring, allowing trait variation between species to be studied genetically beyond the few systems where it is currently possible (Orr and Presgraves 2000; Woodruff et al. 2010).
Recently, several groups have reported using CRISPR to create chromosomal rearrangements (Choi and Meyerson 2014; Maddalo et al. 2014; Li et al. 2015). To the best of our knowledge, ours is the first report of using CRISPR to generate LOH events. In addition to their research applications, targetable endonucleases hold promise for gene therapy (Hsu et al. 2014; Tebas et al. 2014). Certain disease alleles may be very difficult to directly target by CRISPR because of their sequence complexity; one example is provided by the expanded trinucleotide repeats that underlie Huntington’s disease. In these cases, directing a DSB to occur in the vicinity of a pathogenic allele so that it is replaced with its nonpathogenic counterpart by LOH may represent a more feasible alternative.
Acknowledgement
We thank the members of the Kruglyak laboratory for helpful discussion and critical reading of the manuscript, and George Church for plasmids. Funding was provided by the Howard Hughes Medical Institute and NIH grant R01 GM102308 (L.K.).