Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2017 Jan 6.
Published in final edited form as: Cell Rep. 2017 Jan 3;18(1):275–286. doi: 10.1016/j.celrep.2016.12.014

A precise genome editing method reveals insights into the activity of eukaryotic promoters

Gregory L Elison 1,2, Ruijie Song 2,3, Murat Acar 1,2,3,4,†,ξ
PMCID: PMC5216458  NIHMSID: NIHMS835858  PMID: 28052256

Abstract

Despite the availability of whole-genome sequences for almost all model organisms, making faithful predictions of gene expression levels based solely on the corresponding promoter sequences remains a challenge. Plasmid-based approaches and methods involving selection markers are not ideal due to copy-number fluctuations and their disruptive nature. Here we present a genome editing method utilizing the CRISPR/Cas9 complex and elucidate insights into the activity of canonical promoters in live yeast cells. The method involves the introduction of a novel cut site into a specific genomic location, followed by the integration of an edited sequence into the same location in a scarless manner. Using this method to edit the GAL1 and GAL80 promoter sequences, we found that the relative positioning of promoter elements was critically important for setting promoter activity levels in single cells. The method can be extended to other organisms to decode genotype-phenotype relationships in various gene networks.

Graphical Abstract

graphic file with name nihms835858u1.jpg

INTRODUCTION

In the few years since its development, the CRISPR/Cas9 genome editing technique has been extensively used for the genetic modification of a large variety of organisms (Cong et al. 2013, DiCarlo et al. 2013, Jiang et al. 2013, Li-En Jao 2013, Mali et al. 2013, Wang et al. 2014), including the budding yeast S. cerevisiae (Bao et al. 2015, DiCarlo et al. 2013, Horwitz et al. 2015, Mans et al. 2015, Ryan et al. 2014). However, while the technique has been remarkably successful, it still fails to edit the genome in a completely scarless manner. For successful editing to take place, the final sequence must have either the protospacer adjacent motif (PAM) sequence or the targeting sequence altered (Horwitz et al. 2015, Mans et al. 2015). Recent papers and protocols acknowledge that they can only achieve scarless genome editing if the desired edit happens to disrupt the targeting or PAM sequence of the gRNA being used (Mans et al. 2015, Ryan et al. 2016). This has not been a problem for most applications involving proteins, as these edits can be made in such a way as not to change the resulting polypeptide (Mans et al. 2015). However, when investigating less well defined genomic regions, such as promoters, where the effects of minor base pair changes are unknown, there is a great need for a truly scarless version of CRISPR which introduces only desired edits with no unwanted changes across a fairly large region of DNA (Ryan et al. 2014). This problem has led to a dearth of in vivo promoter studies aiming to dissect genotype-phenotype relationships, as introducing a wide variety of scarless changes has previously been too time consuming.

Despite the existence of other editing methods, even in S. cerevisiae, a model system known for its amenability to genomic editing, there are numerous problems when attempting to make scarless edits. Homologous recombination, the well-known method to introduce novel genetic material into yeast, is dependent on the presence of a selectable marker in order to increase editing efficiency to a usable range. This marker is a large scar in the genome, which may have a variety of unknown effects on sensitive genomic functions. Additionally, the scarless method known as URA-FOA counter-selection (Boeke 1987) has such a low efficiency that finding a correctly edited yeast strain takes a significant amount of time.

To address this challenge, we have adapted the CRISPR genome editing technique in order to modify genomic regions with no undesired base pair changes. We utilize CRISPR/Cas9 to create double strand breaks flanking a region of interest in order to completely remove it from the genome and introduce a novel targeting sequence and PAM site in its place. This new site is then cut in a second round of editing, in which the original region is reintroduced along with any desired modifications. This two-step editing method allows any desired edits to be introduced across the genomic region in question, while maintaining all other base pairs as they are in the parent strain. In addition, once the intermediate strain is created, it may be reused indefinitely, thus every subsequent new edit to the same genomic region only requires one transformation instead of two.

We used this method for the in vivo editing of the canonical GAL1 promoter in yeast. We systematically removed or recoded the Gal4 binding sites from the promoter in order to study the phenotypic consequences of these changes at the single cell level. The activity from the edited promoter architectures was compared to the bimodal activity profile of the wild type GAL1 promoter and deviations from the wild type behavior were analyzed in terms of the fraction of ON cells and the expression level of the ON state. We found that the fourth binding site does not have any effect on transcriptional activity. Removing or recoding the first site prevented any activity from the promoter despite the presence of the second and third binding sites. Surprisingly, however, further removing the third Gal4 binding site (together with the first one) partially restored the wild type activity in the GAL1 promoter. Using our method to edit the GAL80 promoter, analyzing the activity of the edited GAL80 promoter architectures, and interpreting the results in the context of the results from the edited GAL1 promoters supported the conclusion that the relative positioning of promoter elements is of critical importance for determining in vivo promoter activity levels at endogenous chromosomal locations. This was then confirmed by edits which changed only the spacing between the Gal4 binding sites and the TATA box, which resulted in expression level changes only attributable to spacing changes within the promoter.

RESULTS

Scarless genome editing in live cells

To demonstrate the viability of this technique, we chose to use a strain of S. cerevisiae in which one copy of the canonical GAL1 promoter driving the yellow fluorescent protein (YFP) has been integrated into the ho locus (Acar et al. 2005, Acar et al. 2010) (Fig. 1A) (Table S1). By targeting this promoter, we were able to see the influence of any edits on the output of YFP fluorescence in environments containing various concentrations of galactose.

Figure 1. The two-step CRISPR editing method.

Figure 1

Open in a new tab

(A) Introducing a novel CRISPR cut site to the GAL1 promoter of the PGAL1-YFP construct integrated in the ho locus of the yeast genome. PGAL1 is flanked by the YFP gene downstream and genomic DNA upstream. Two gRNA/Cas9 complexes target cut sites immediately adjacent to PGAL1 resulting in its loss from the genome. The addition of a donor oligonucleotide with a novel CRISPR cut site flanked by regions of homology to the upstream genome and YFP allows repair. The repair results in a new region in which the old cut sites have been removed and a novel one is inserted. (B) Introducing an edited PGAL1 oligonucleotide back to the original location. The novel CRISPR cut site is cut by a gRNA/Cas9 complex. Repair is facilitated by the addition of a donor oligonucleotide consisting of an edited PGAL1 and homology to the upstream genome and YFP. The repair gives rise to an edited PGAL1 upstream of YFP. (C) Efficiency of CRISPR editing. Of the colonies tested for CRISPR editing, 63% contained the correct sequence after the first stage of editing for PGAL80, and 77% contained the correct sequence after the second stage of editing for PGAL1. Error bars indicate SEM (N=2 for stage one; N=3 for stage two).

The first step of the genome editing process was to replace the existing PGAL1 with a short (30bp) region containing a novel CRISPR cut site (Fig. 1A). To do this, gRNAs targeting the nearest available cut sites immediately before and after the GAL1 promoter were added to a pRS315 plasmid containing a yeast centromeric element and a LEU2 marker. In addition, a donor oligonucleotide was created containing the novel cut site flanked by 50bp of homology to each of the regions immediately adjacent to the cut sites described above. The plasmid and donor oligonucleotide were co-transformed into a yeast strain containing CAS9 driven by a constitutive promoter as well as the PGAL1-YFP construct described above. Because of the relative proximity of the cut sites, the cellular repair machinery treats the two as a single break, which is repaired through the use of the added donor oligonucleotide while the intervening region is lost. The end result of this step is the replacement of the original PGAL1 and small adjacent regions including the original PAM sites with a 30bp novel DNA sequence containing a new CRISPR cut site.

The second step of the editing process begins upon creation of the new strain described above. This involves inserting a gRNA targeting the introduced PAM site into a pRS315 plasmid and co-transforming it into the new yeast strain along with a new piece of donor DNA which consists of the original promoter containing any desired edits and 50bp of homology to each end of the new break (Fig. 1B). The repair will replace the adjacent regions which were lost in the previous step and add the newly edited PGAL1 region. Ultimately, the only changes to the original strain are those desired. The targeting and PAM sites targeted in the first step were reintroduced such that no trace of the editing process remains in the final genome (Fig. 1B).

Characterization of editing efficiency

We measured the editing efficiency for both steps of our technique (Fig. 1C), as efficiency is a key factor for any genome editing technique. The initial step had an efficiency of 63%. We note that the first step must only be completed once for any editing to be conducted at the relevant locus. The efficiency of the second editing step is much more important and this step was found to have an efficiency of 77% (Fig. 1C). This editing efficiency is comparable with the efficiencies reported in literature (Horwitz et al. 2015, Mans et al. 2015).

Editing GAL1 promoter architecture and measuring promoter activity in single cells

To functionally validate our technique, we decided to explore the phenotypic consequences of introducing rational edits to a promoter. For this, we investigated the effects of removing Gal4 binding sites from the canonical GAL1 promoter which is a faithful reporter of the yeast galactose (GAL) network (Fig. S1) activity (Acar et al. 2005). The constitutively expressed Gal4 protein serves as the main transcriptional activator of the GAL network (Mizutani and Tanaka 2003). It activates transcription from all network promoters, including that of the GAL1 promoter, by binding to a 17bp site (CGG-N(11)-CCG) on the promoters (Johnston 1987). On the other hand, Gal80 repressor proteins negatively regulate the network activity by binding to Gal4 on promoter sites (Mizutani and Tanaka 2003) (Fig. 2A). Galactose relieves Gal80 repression via Gal3 proteins. Activated Gal3 binds to Gal80, leading to the dissociation of Gal80 from Gal4 proteins (Peng and Hopper 2002). There is no known crosstalk between the GAL network and any other gene except for an inhibitor which is dependent on the presence of glucose, making it ideal for study in conditions without the presence of glucose.

Figure 2. Components of the GAL network and characterization of GAL1 promoter activity.

Figure 2

Open in a new tab

(A) Relevant components of the yeast GAL network. GAL4 is constitutively expressed and activates PGAL80 and PGAL1. Gal80 inhibits the activity of Gal4, thus inhibiting both its own and PGAL1 expression. (B) Architecture of the GAL1 promoter. The first line represents the wild type GAL1 promoter. The red boxes are Gal4 binding sites and the yellow-filled triangle is the TATA box. The second line represents a GAL1 promoter in which the first Gal4 binding site has been recoded (empty black box). The third line represents a GAL1 promoter in which the first Gal4 binding site has been removed entirely (black line), shortening the length of the promoter by 17bp. (C) Phenotypic characterization of wild type PGAL1-YFP activity. The bar at the top indicates that the data refers to the wild type architecture of the GAL1 promoter. The seven histograms show the flow cytometry data at seven concentrations of galactose as indicated from 0% to 0.5%. The top right panel shows the fraction of ON cells as galactose concentration increases, and the bottom right panel shows the mean expression level of the ON state for concentrations with at least 5% ON cells. Error bars indicate SEM (N=3). (D) Phenotypic characterization of an edited GAL1 promoter used in PGAL1-YFP. The bar at the top indicates that the data refers to an edited GAL1 promoter with the first and second Gal4 binding sites recoded. The top and bottom right panels show the fraction of ON cells and mean ON expression levels, respectively, as galactose concentration increases (blue). The data from the wild type GAL1 promoter (black) is included for comparison purposes. The mean ON expression level plots only include galactose concentrations at which both strains have at least 5% ON cells. Error bars indicate SEM (N=3). See also Figure S1. See also Table S1.

The GAL network is induced by galactose (John and Davis 1981) in a bimodal fashion (Acar et al. 2005). To use as a reference, we measured the activity of the unedited wild type GAL1 promoter driving YFP. For this, we grew yeast cells for 22 hours in the presence of various galactose concentrations (0%, 0.025%, 0.05%, 0.1%, 0.2%, 0.4%, 0.5%) in addition to 0.1% mannose. Mannose was used as a non-inducing sugar that is metabolized by both OFF and ON cells of the bimodal distribution. At the end of the 22-hour period, single cell YFP expression levels of ~3,000 cells were measured using a flow cytometer (FACS), and two important phenotypes associated with a bimodal distribution were quantified for each condition: fraction of ON cells and mean expression level of the ON cells (Fig. 2B).

Altering the architecture of the GAL1 promoter can potentially change the fraction of ON cells, the mean expression level of the ON state, or both. The native GAL1 promoter (Johnston and Davis 1984, Kellis et al. 2003) contains three Gal4 binding sites immediately adjacent to each other followed by a fourth binding site with a different sequence 45 base pairs downstream of the previous three (Fig. 2B). 184 base pairs downstream of the fourth site is a TATA box and the ATG start codon is 145 base pairs downstream of the TATA box (Fig. 2B). We altered the four Gal4 binding sites in two ways: by simply removing the 17bp binding site from the promoter outright and by recoding the conserved CGG and CCG sequences of the site into randomized A/T sequences of the same length, abolishing the binding of Gal4 on that position while keeping the number of base pairs intact.

To investigate the phenotypic consequences of these binding site alterations on the activity of the GAL1 promoter, we performed galactose induction experiments and measured the resulting promoter activity profiles in single cells. Figure 2C shows results from the wild type strain, while Figure 2D shows, as an example, results from one of the promoter-edited strains. In the edited strain, the first and second Gal4 binding sites were recoded to neutral sites on the GAL1 promoter. Compared to the wild type promoter activity, the recoding of these two binding sites resulted in a reduction of the ON-state mean expression level, but a nearly identical fraction of ON cells. These results verify that edits made using this method can produce phenotypic changes in vivo.

Editing the GAL80 promoter sequence

To move the phenotypic characterization of our editing approach to the gene network level and to test the approach on the editing of a promoter functioning in its natural location, we next introduced rational edits on the endogenous GAL80 promoter which contains a single Gal4 binding site (Fig. 3A, red box). For this purpose, we constructed four strains carrying different edits on the GAL80 promoter and measured the resulting activity of the GAL network using YFP driven by the wild type GAL1 promoter.

Figure 3. Editing the GAL80 promoter and phenotypic characterizations.

Figure 3

Open in a new tab

(A) Cartoons illustrating promoter elements and edits on the GAL80 promoter. The first depicts the wild type GAL80 promoter with a single Gal4 binding site (red box) and a TATA box (yellow triangle). The second shows a second Gal4 binding site (red arrow) inserted before the original site. The third has a recoded Gal4 binding site (empty red box) before the original site. (B–E) Each horizontal bar indicates one strain and illustrates the specific edits introduced in the GAL80 promoter. The filled red boxes are Gal4 binding sites while recoding an existing sequence to a Gal4 binding site is depicted by an empty red box. A pointing red arrow indicates the second Gal4 binding site inserted. The wild type data is plotted in black as the reference while the data of each bar-illustrated strain are shown in blue. Each panel depicts either the fraction of ON cells in percent or the mean expression level of the ON cells in arbitrary units (a.u.), reported by the unedited PGAL1-YFP. The mean ON expression level plots only include galactose concentrations at which both strains plotted have at least 5% ON cells. Error bars indicate SEM (N=3). See also Table S1.

In the first strain, we inserted a new Gal4 binding site immediately before the existing one (Fig. 3B) on the GAL80 promoter, while in the second strain we recoded the existing base pairs before the existing Gal4 binding site into a second Gal4 binding site (Fig. 3C), a change doubling the number of binding sites but not increasing the length of the promoter. Both of these strains showed a massive increase in the fraction of ON cells – in the recoded strain, virtually all cells were in the ON state even in the absence of galactose. Since Gal80 is a repressor protein, these findings indicate that the alterations we made on the GAL80 promoter greatly reduced or eliminated GAL80 expression, thus allowing GAL1 promoter to be activated in a much larger fraction of cells.

In the third strain, we added a new Gal4 binding site immediately downstream of the existing one, while in the fourth strain we recoded the existing base pairs downstream of the existing Gal4 binding site into a second Gal4 binding one. Surprisingly, compared to the unedited strain, the strain with one Gal4 binding site added downstream from the original site showed a reduction in the fraction of ON cells while keeping the mean YFP expression levels of the ON cells roughly the same (Fig. 3D). This is in contrast to the results obtained from the recoded strain (Fig. 3E), which resembled the initial two strains (Fig. 3B–C) discussed above in that the expression of GAL80 seemed to be completely eliminated at all galactose concentrations used. This result from the promoter containing an additional downstream binding site (Fig. 3D) indicates higher GAL80 expression levels compared to wild type. These results suggest that while binding site number is clearly important for setting promoter expression level or strength, it is not the only factor. The location of binding sites within the promoter is also of utmost importance. To give a specific example, when two Gal4 binding sites are adjacent to each other (Fig. 3B–E), their relative positioning to the other promoter elements determines if transcription will proceed (Fig. 3D) or be blocked (Fig. 3B–C,E). Future studies utilizing single molecule imaging techniques may directly elucidate how such transcriptional inactivation may occur.

Introducing simultaneous edits on the GAL1 and GAL80 promoters

To further investigate our findings and expand the applicability of our method, we set out to simultaneously introduce edits on two different genomic locations: GAL80 promoter at its endogenous location and GAL1 promoter driving YFP at the ho locus. To do this, we first constructed a strain containing novel targeting and PAM sites replacing both GAL1 and GAL80 promoters. We then transformed this strain with a plasmid containing two gRNAs, one for each promoter site, as well as donors carrying the edited promoters (Fig. 4A). To find out the combined efficiency of simultaneous editing at both genome locations, we examined editing efficiency for both locations individually and together. Using targeted sequencing, we found that 76% of the colonies were successfully edited at the GAL1 promoter, 83% were successfully edited at the GAL80 promoter, and 69% were successfully edited at both locations (Fig. 4B). These efficiency levels are in line with efficiencies reported in yeast by recent CRISPR papers (Horwitz et al. 2015).

Figure 4. Introducing simultaneous edits into the GAL1 and GAL80 promoters.

Figure 4

Open in a new tab

(A) Cartoon diagram showing the two-step CRISPR editing technique. The second step of the technique introduces two edited sequences (GAL1 and GAL80 promoters) simultaneously. (B) Editing efficiency for the second step of the technique. 76% of the colonies had the correct GAL1 promoter edits, 83% had the correct GAL80 promoter edits, and 69% were correct for both GAL1 and GAL80 promoters. Error bars indicate SEM (N=3). (C) Phenotypic characterization of the strain carrying the dual promoter edits. Each horizontal bar indicates the GAL1 (green) or GAL80 (sky blue) promoter and illustrates the specific edits introduced in them. The filled red boxes are Gal4 binding sites while recoding an existing sequence to a Gal4 binding site is depicted by an empty red box. An empty black box shows recoding a previously existing Gal4 binding site to a null sequence with the same length. A yellow triangle indicates the TATA box. The wild type data is plotted in black as the reference while the data from the dual-bar-illustrated edited strain are shown in blue. The left and right panel depicts, respectively, the fraction of ON cells in percent or the mean expression level of the ON cells in arbitrary units (a.u.), reported by the edited PGAL1-YFP. Green (orange) data points show the phenotypic levels measured from a strain carrying the same edits but only on the GAL1 (GAL80) promoter, not both. The mean ON expression level plot only includes galactose concentrations at which all strains plotted have at least 5% ON cells. Error bars indicate SEM (N=3). See also Table S1.

To measure the phenotypic consequences of the dual edits, we constructed a strain by combining two specific sets of edits previously introduced on GAL1 (Fig. 2D) and GAL80 (Fig. 3C) promoters. When introduced into the wild type background, the GAL1 promoter edits (the first and second binding site locations recoded) gave rise to ON cell fractions similar to the wild type strain, however, the expression level of the ON state was drastically reduced (Fig. 2D, blue; Fig. 4C, green). On the other hand, when introduced into the wild type background, the GAL80 promoter edit (recoding a second Gal4 binding site upstream of the original one), led to increases in the fraction of ON cells and ON-state expression levels (Fig. 3C, blue; Fig. 4C, orange). Analyzing the fraction of ON cells in the strain carrying the GAL1 and GAL80 dual edits (Fig. 4C, left panel, blue) showed us that the GAL80 edit, which lowered GAL80 protein expression, dictated the phenotype of the dual edit strain. This was expected, as the phenotypic effect of the GAL1 edits was neutral (Fig. 4C, left panel, green). However, when analyzing the phenotype of mean ON-state expression levels, it was clear that the GAL1 edit, which lowered the GAL1 expression mean (Fig. 2D, right panel, blue), dictated the phenotype of the dual edit strain. This was also expected as the phenotypic effect of the GAL80 edits was considerably less intense than that of the GAL1 edit (Fig. 4C, right panel, orange).

Mechanistic insights into the differential activities of the edited GAL1 promoters

In an effort to further understand how the changes in promoter activity are caused by the addition or removal of Gal4 binding sites, we combinatorially removed or recoded each of the three main Gal4 binding sites on the PGAL1-YFP construct and measured the resulting promoter activity profiles compared to the unedited promoter’s activity. We first examined YFP profiles from the two strains in which all three of the immediately adjacent Gal4 binding sites were removed or recoded and found that these strains lacked any YFP expression (Fig. S2A). In addition, when we removed or recoded the fourth binding site, we found that these changes did not have any effect on GAL1 activity (Fig. S2B–C). These results indicate that the fourth binding site is neither necessary nor sufficient for promoter activity. We thus concluded that this site was not relevant to the GAL1 promoter’s activity and treated it as such for the remainder of this study. The difference of this site from the canonical motif (Supplemental Information) further supports our phenotypically validated conclusion.

Focusing our attention on the first three canonical Gal4 binding sites, we next examined the promoter-edited strains to determine if removing or recoding the first Gal4 binding site (by itself or together with the second site) had an effect on transcriptional activity from the GAL1 promoter. Similar to the results observed by editing the GAL80 promoter (Fig 3), removing or recoding the first site prevented any activity from the promoter despite the presence of the remaining two binding sites (Fig. 5A). More surprising was the observation that recoding both the first and third Gal4 binding sites restored the wild type activity in the promoter in terms of the fraction of ON cells but not in terms of the expression level of the ON state (Fig. 5B). Removing the first and third binding site locations had the same effect (Fig. S3A). Similar phenotypes were also seen for other GAL1 promoter edits (Fig. S3B–C). Removing or recoding the third binding site prevented any activity from the promoter despite the presence of the other two binding sites (Fig. S3B–C). While the hypothesis that the number of transcription factor binding sites on a promoter is the key determinant of promoter activity was initially appealing, results from three promoter architectures contradicted this idea. Two of these represent promoter designs in which the second and third binding sites had both been either removed or recoded, leaving only the first Gal4 binding site intact (Fig. 5C). According to the binding site number hypothesis, the activity from these two promoters should have been similar to the activity from those promoters carrying only the second Gal4 binding site intact (Fig. 5B, Fig. S3A). Instead, having only the first Gal4 binding site intact resulted in the abolishment of all activity from the promoter, while having only the second Gal4 binding site resulted in a promoter activity profile matching the wild type in terms of the fraction of ON cells. The other promoter architecture which generated results contradictory to the binding site number hypothesis was the one containing intact first and third Gal4 binding sites with the second site recoded (Fig. 5D). Despite having two functional binding sites, the phenotypic characteristics of this promoter resembled those promoters (Fig. 2D, Fig. 5B, Fig. S3A) with only one functional binding site. These findings, together with the ones observed from the edited GAL80 promoters (Fig. 3), made us reject the hypothesis that the number of transcription factor binding sites was the sole key factor in setting promoter activity levels.

Figure 5. Systematically editing the canonical Gal4 binding sites on the GAL1 promoter.

Figure 5

Open in a new tab

Each horizontal bar indicates one strain and illustrates the specific edits introduced in the GAL1 promoter. The filled red boxes are Gal4 binding sites while removing a Gal4 binding site by recoding is depicted by an empty black box. A yellow triangle indicates the TATA box and the black line indicates the removal of a Gal4 binding site. (A–D) Wild type data is plotted in black as the reference while the data of each bar-illustrated strain are shown in blue. Each panel depicts either the fraction of ON cells in percent or the mean expression level of the ON cell in arbitrary units (a.u.). The mean ON expression level plots only include galactose concentrations at which both strains plotted have at least 5% ON cells. Error bars indicate SEM (N=3). See also Figure S3. See also Table S1.

As an alternative hypothesis, we propose that the relative position of the Gal4 binding site(s) and other promoter elements, such as the TATA box, is an important parameter influencing the promoter activity levels. The fact (Dion and Coulombe 2003) that Gal4 binding site(s) are involved in the looping of DNA for a stable transcriptional initiation provides a mechanistic basis for this hypothesis. Accepting distance between important promoter elements as a key factor does not fully discard potential contributions from the number of binding sites on promoter activity. Indeed, removing a binding site automatically alters the relative distances between promoter elements, meaning that the parameters of binding site number and distance cannot be fully orthogonal. Also, it is more than plausible to think that, when crucial spacing needs between promoter elements are not met, the effects of transcription factor binding site number could be muted due to disruption of the spacing necessary for proper folding of DNA and transcriptional initiation.

Going back to our experimental observations, the hypothesis based on the relative positioning of promoter elements would explain why the two strains with only the first Gal4 binding site intact (Fig. 5C) failed to be inducible. Having only one binding site was not the culprit as having only the second binding site had produced a wild-type-like fraction of ON cells (Fig. 5B, Fig. S3A). This indicates that, compared to the phenotypic effect of the second binding site, having only the first site increased the distance between the site and downstream promoter elements, leading to the deactivation of the promoter. The relative positioning hypothesis also explains the re-activation of the GAL1 promoter when two Gal4 binding sites are separated by 20bp (Fig. 5D), as the promoter was inactive when two binding sites were separated by 2bp (Fig. 5A).

Systematic alterations of distance between specific promoter elements

To further confirm the hypothesis that the spacing of key promoter elements can have a drastic impact on promoter activity, we constructed three additional strains in which the distance between the third Gal4 binding site and the TATA box was shortened by 5, 10, or 17 base pairs. This was done by moving the 5, 10, or 17 bp-long GAL1 promoter region immediately downstream of the third Gal4 binding site to a location immediately upstream of the first Gal4 binding site (Supplemental Information). These strains were built off of the strain seen in Figure 2D, as we reasoned that having only one canonical binding site would make results easier to interpret. The resulting three strains therefore had the single intact Gal4 binding site moved 5, 10, or 17 bp closer to other promoter elements, such as the transcription start site (TSS) and the TATA box, and moved the same distance away from the 5′ end of the promoter. The parent strain (Fig. 2D and Fig. 6A) displayed a fraction of ON cells virtually identical to that of the wild type. In contrast, the strain in which the Gal4 binding site was moved 5 bp closer to the TSS and TATA box (Fig. 6B) displayed an almost complete lack of promoter activity. However, the strain in which the Gal4 binding site was moved 10 bp closer to the TSS and TATA (Fig. 6C) closely resembled its parent (Fig. 6A). Even more interestingly, the strain in which the Gal4 binding site was moved 17 bp closer to the TSS and TATA box (Fig. 6D) had a similar fraction of ON cells to both its parent and to the wild type strain, but had a much higher mean expression level than its parent, even higher than the wild type.

Figure 6. Phenotypic characterization of systematic spacing changes in the GAL1 promoter.

Figure 6

Open in a new tab

Each horizontal bar indicates one strain and illustrates the specific edits introduced in the GAL1 promoter. The filled red boxes are Gal4 binding sites while removing a Gal4 binding site by recoding is depicted by an empty black box. A yellow triangle indicates the TATA box and the black line indicates the removal of a Gal4 binding site. Arrows above the promoter indicate the number and position of base pairs moved from one location within the promoter to another. Wild type data is plotted in black as the reference while the data of each bar-illustrated strain are shown in blue. Histograms show the flow cytometry data at seven concentrations of galactose as indicated from 0% to 0.5%. Each panel depicts either the fraction of ON cells in percent or the mean expression level of the ON cells in arbitrary units (a.u.), reported by the edited PGAL1-YFP. The mean ON expression level plots only include galactose concentrations at which both strains plotted have at least 5% ON cells. (A) Strain with first and second Gal4 binding sites recoded. This is the parent of the three strains (carrying specific spacing changes) whose phenotypes displayed in the remaining panels of this figure. (B) Strain with the 5 bp sequence immediately downstream of the third Gal4 binding site moved immediately upstream of the first Gal4 binding site. (C) Strain with the 10 bp sequence immediately downstream of the third Gal4 binding site moved immediately upstream of the first Gal4 binding site. (D) Strain with the 17 bp sequence immediately downstream of the third Gal4 binding site moved immediately upstream of the first Gal4 binding site. Error bars indicate SEM (N=3). See also Table S1.

These results definitively show that the spacing between various promoter elements is crucial for promoter activity. The stark difference among strains with their Gal4 binding sites located only a few base pairs closer to or farther from the TSS and TATA box illustrates that a change in the relative position of various promoter elements is capable of almost completely eliminating activity from a promoter (Fig. 6B), keeping activity almost exactly the same (Fig. 6C), or massively increasing expression compared to the parent strain (Fig. 6D). The changes made to the number of Gal4 binding sites also resulted in small spacing changes which we originally ignored, such as those in Figure 5C (right panel), but are likely to be the cause of the loss of expression from the promoter. We hypothesize that these phenotypic changes can be explained by the positioning of the binding sites on the DNA helix itself. A helical turn of the DNA is known to be ~10.5 bp, so a change of 5 bp should place the Gal4 binding sites on the opposite side of the DNA compared to the original, potentially resulting in a loss of interaction between the DNA-bound Gal4 and the transcriptional machinery. However, a change of 10 bp would bring the Gal4-bound site back to its original position in three-dimensional space, potentially explaining why expression is restored. The increased expression level resulting from the 17 bp shortening is harder to fully explain as the 17 bp change is expected to lead to both a partial helical turn and a change in position along the DNA strand. However, this result still reinforces the main hypothesis that the spacing of key promoter elements is critical for setting specific promoter activity levels.

In summary, while there may still be a number of unknown factors influencing the overall promoter activity from each different configuration of the GAL1 promoter, we have elucidated an initial model (Fig. 7) based on strains in which the third Gal4 binding site is systematically moved towards the TSS and TATA box. The wild type promoter architecture carries three canonical binding sites on which Gal4 proteins bind. The presence of transcriptional mediators induces the bending of the Gal4-bound DNA in order to facilitate stable assembly of the transcriptional machinery including RNA polymerase (Fig. 7A). On a promoter with a single active binding site, Gal4 binds to that site and interacts with the transcriptional machinery, causing promoter activation but with a lower expression level (Fig. 7B). When the single Gal4 binding site is moved 5 bp closer to the TATA box, the site-bound Gal4 protein is rotated almost a full half turn around the DNA helix, causing all expression to be lost, likely due to the loss of interactions between Gal4 and the transcriptional machinery on the promoter (Fig. 7C). However, when the single binding site is moved 10 bp closer to the TATA box, the site-bound Gal4 protein is rotated almost full circle around the DNA helix, leading to promoter activity levels very similar to the levels seen without the 10 bp change. This result indicates that the 10 bp move restores the interaction of the Gal4 protein with the transcriptional machinery (Fig. 7D). This model explains why small changes in spacing caused loss of activity from the GAL1 promoter in many of the edited promoters examined in this study.

Figure 7. Model incorporating distance-based factors influencing GAL1 promoter activity.

Figure 7

Open in a new tab

DNA is shown in green. The filled red boxes are Gal4 binding sites while removing a Gal4 binding site by recoding is depicted by an empty black box. A bright yellow box indicates the TATA box. Gal4 proteins are depicted by blue pacman symbols. TATA binding protein is shown in cherry. DNA polymerase is shown in dark yellow. Mediator proteins are shown in purple. Black dotted lines indicate interactions between Gal4 proteins and the mediator complex. Red arrows indicate the turn of the DNA helix (A) Schematic of the wild type GAL1 promoter. (B) Schematic of the promoter architecture present in the parent strain from which the three strains carrying three spacing changes were constructed. The first and second Gal4 binding sites have been recoded. (C) Schematic of the promoter having the third Gal4 binding site move 5 bp towards the TSS and TATA box (rotated nearly half circle around the DNA helix). (D) Schematic of the promoter having the third Gal4 binding site move 10 bp towards the TSS and TATA box (rotated almost one full circle around the DNA helix).

DISCUSSION

In this study, we have applied the CRISPR/Cas9 genome editing technology to introduce a technique that allows truly scarless editing of large genomic regions in live yeast cells by preserving endogenous PAM sites. We have shown that, by removing the region of interest and then replacing it with an edited version, any number of edits can be made to a genomic region. We recognize that the necessity to first construct an intermediate strain by introducing a novel PAM site may be less than ideal; however it is indispensable when completely scarless editing is needed. This is due to the inability to introduce donor oligonucleotides retaining the original PAM sites which are present outside of the region to edit. If we were to use donors with the original PAM sites, the CRISPR machinery would not be able to differentiate between the editing donor and the to-be-edited genome and would cut both, a problem recognized in several previous papers (Horwitz et al. 2015, Mans et al. 2015). The use of this method to overcome this significant obstacle is a major advantage over any other in vivo editing method for sensitive genomic regions.

The method presented here can be extended to many other organisms and we anticipate that it will be widely used in a broad range of fields and applications including directed evolution (Arnold 1993, Kuchner and Arnold 1997, Shao and Arnold 1996) performed in vivo. There is currently no easy method to create synthetically modified organisms containing libraries of mutations at specific regions in their genome. Such an ability would greatly enrich the technique of directed evolution, as edited genes could be expressed in their native loci and without copy number variations, both of which have been inevitable disadvantages of the traditional plasmid-based directed evolution technique.

As a potential limitation of our method, the necessity to have more than one CRISPR step may make the method relatively difficult to perform in animals. Also, we anticipate that essential genes would not be amenable to editing with our method as organisms could not survive the lack of such genes during the intermediate step of editing. Temporarily introducing essential genes during the intermediate step via plasmids should alleviate this limitation.

Using this method, we have elucidated previously unknown characteristics of the canonical GAL1 and GAL80 promoters that are commonly used not only in yeast studies but also in many other eukaryotic organisms. We have constructed yeast strains with edited promoters and examined the effect of transcription factor binding site number and location in vivo. We have found several surprising results providing important insights into a more comprehensive understanding of genotype-phenotype relationships in eukaryotes. For example, having two Gal4 binding sites immediately adjacent to each other did not result in promoter activation even in high concentrations of galactose, while promoters with either one or three Gal4 binding sites immediately adjacent to each other had normal promoter activation. Intriguingly, when the two Gal4 binding sites were separated by only 20bp, the promoter was still functional in an activity pattern similar to the wild type.

Even more surprisingly, we have found that small changes to the spacing between transcription factor binding sites and other promoter elements, such as the TATA box, can have dramatic effects on promoter activity. We acknowledge the existence of previous work (Sharon et al. 2012) in which Gal4 binding sites were added to synthetic promoters including a modified GAL1/10 promoter that had its native regulatory elements inactivated by mutation, and which reported that increasing the number of binding sites (up to 5 at predefined locations) generally increased the level of transcription (with one notable exception for two binding sites placed 1bp apart), in seeming contradiction to our observations. However, the positions of the added Gal4 binding sites in that study were artificially determined and not related to their native position. As we have shown here, the distance between the binding site and other promoter elements (such as the TSS and TATA box) is critically important to the binding site’s ability to activate transcription, so the fact that we reach different conclusions with respect to the number of promoter binding sites should be unsurprising. Our results, obtained using a marker-free precise genome editing technique on native promoters, reinforce the idea that the relative positioning of promoter elements is of critical importance to promoter activity in live single cells, and demonstrates that simply adding transcription factor binding sites to a promoter can have nonlinear consequences for transcriptional activation.

EXPERIMENTAL PROCEDURES

Construction of plasmids

The base plasmid used for all plasmid construction was pRS315, a yeast centromeric plasmid containing a LEU2 marker, which was a gift of Mark Hochstrasser. In order to create plasmids containing a single gRNA cassette, we ordered (Integrated DNA Technologies) a 388bp gRNA cassette sequence modified from the template created by DiCarlo et al. (DiCarlo et al. 2013) (Supplemental Information). The cassette consists of a promoter, terminator, and gRNA sequence which function well in yeast. These sequences were constructed with a BamHI cut site on the 3′ end and a HindIII cut site on the 5′ end of the cassette. Upon receipt, these sequences and pRS315 were digested with the appropriate enzymes and ligated to form plasmids containing a gRNA cassette. In order to create plasmids containing two gRNA cassettes, the two cassettes were individually ordered with the first containing a NotI site and a BamHI site and the second containing an XmaI site and a HindIII site. The two were then ligated into pRS315 in two separate transformations to create a plasmid containing two gRNAs separated by 12 base pairs.

Introducing the CAS9 gene into yeast

All strains were created from WP35, a W303 strain with one copy of the PGAL1-YFP construct inserted into the ho locus. The rest of the yeast genome was unaltered. To insert CAS9 into this strain, we obtained plasmid #43802 from Addgene. This plasmid, p414-TEF1p-Cas9-CYC1t, described by DiCarlo et al. (DiCarlo et al. 2013), contains the CAS9 gene under a constitutive TEF1 promoter along with a TRP1 marker. To insert it into WP35, we removed the centromeric region from the plasmid and then linearized the plasmid in the TRP1 gene using MfeI enzyme and transformed the linear product into WP35 using the standard lithium acetate (LiOAc) transformation technique. This resulted in the strain GE1.

Introducing rational edits on the GAL1 and GAL80 promoters

The first step of editing at the GAL1 promoter (PGAL1) consisted of transforming GE1 with a plasmid containing gRNAs targeting two sites flanking PGAL1, with the goal of cutting out the entirety of PGAL1. A donor oligonucleotide carrying a novel 30bp CRISPR cut and PAM site in between 50bp regions homologous to each genomic region flanking the GAL1 promoter was co-transformed (LiOAc technique) with the plasmid containing two gRNAs so that cells could use the donor as a repair template. The transformed cells were then grown on –LEU plates for two days after which colonies were PCR tested and sequenced to verify that editing had taken place. After this, replica plating between rich media (YPD) and –LEU media was used to locate colonies which had lost the centromeric plasmid used in the first step. This was necessary because the plasmid used in step 1 would have been able to cut the desired product after the second round of editing if it remained in the cells.

The second step of the editing process consisted of transforming (using the LiOAc technique) the intermediate strain carrying the 30bp novel PAM site with a plasmid containing a gRNA cassette targeting the new CRISPR cut site in the presence of a new donor oligonucleotide. The donor consisted of an edited version of the PGAL1 in between 50bp sequences homologous to the genomic regions flanking the cut site. The transformed cells were then grown on –LEU plates for two days after which colonies were PCR tested and sequenced to verify that editing had taken place.

Editing PGAL80 involved the same two-step process, but used gRNAs and donors that were relevant to PGAL80. Finally, to construct the multiplexed strains, the transformations for the first step of the editing process were performed successively in order to create an intermediate yeast strain with both PGAL1 and PGAL80 replaced by two novel PAM sites. The second step of the editing process was completed simultaneously for the PGAL1 and PGAL80 loci by using a single plasmid containing two gRNAs (each targeting one locus) and two separate donor oligonucleotides carrying the edited PGAL1 and PGAL80 sequences.

Quantifying the efficiency of the CRISPR editing

The editing efficiency was calculated by dividing the number of correctly edited colonies (as identified by PCR confirmation and sequencing) by the total number of colonies examined. The use of a selectable marker on the gRNA-containing plasmid ensured that only colonies which received a plasmid could grow and be used for analysis. To quantify the efficiency of the first step of the editing process, we used the results obtained from targeting the GAL80 promoter site. Two independent transformations were conducted, 8 and 24 colonies were collected from the two transformations, and the editing efficiency values were calculated separately. As a result, 6 out of 8 colonies and 14 out of 24 colonies were counted as correct. The results were combined to provide the overall efficiency in terms of mean and standard error of the mean (S.E.M., N=2) as presented in the manuscript. In quantifying the efficiency of the second step of the editing process, three independent transformations were conducted by simultaneously editing the GAL80 and GAL1 promoters, 24 colonies were collected from each transformation, and the editing efficiency values were calculated separately. The results were combined to provide the overall efficiency in terms of mean and standard error of the mean (S.E.M., N=3) as presented in the manuscript.

Growth conditions, media, and flow cytometry data analysis

Cells were grown in the appropriate synthetic amino acid dropout media. All growths were conducted in duplicate at 30°C in a shaking incubat or in 5 mL of media. Cells were first grown overnight for 22 hours in minimal media containing 0.1% mannose as the carbon source, reaching to an optical density (OD600) between 0.075 and 0.15. They were subsequently diluted into the induction media containing 0.1% mannose and the appropriate concentration of galactose, and grown for another 22 hours, reaching to a cell density between 0.075 and 0.15. After the induction period, single cell fluorescence values were analyzed using flow cytometry (Stratedigm-8 with HTAS). Each FACS sample had on average 3,000 cells after gating. Log-amplified fluorescence measurements for the gated cells were converted to linear scale for analysis. A threshold for ON state (75.7 a.u.) was selected based on fluorescence measurements from uninduced, unedited cells and uniformly applied to all samples. The fraction of ON cells was then quantified for each sample. For each galactose concentration that resulted in at least 5% ON cell, the mean expression level of such cells was also quantified.

Supplementary Material

1
2

Acknowledgments

The authors thank Acar Lab members for useful discussions and W. Peng for providing the base strain WP35. MA acknowledges funding through NIH Director’s New Innovator Award (1DP2AG050461-01) and a Junior Faculty Fellowship from Yale University. GLE was in part supported by the National Institute of Health T32 GM007499. p414-TEF1p-Cas9-CYC1t was a gift from George Church (Addgene plasmid # 43802).

Footnotes

AUTHOR CONTRIBUTIONS

GLE and MA designed the experiments and analyses. GLE constructed the strains, performed the experiments, and analyzed the data quantifying editing efficiency. RS analyzed the flow cytometry data. GLE and RS prepared the figures. MA guided the study. GLE and MA wrote the manuscript. All authors approved the manuscript.

COMPETING FINANCIAL INTERESTS

MA and GLE have filed a provisional application with the US Patent and Trademark Office on this work.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Acar M, Becskei A, Van Oudenaarden A. Enhancement of cellular memory by reducing stochastic transitions. Nature. 2005;435(7039):228–232. doi: 10.1038/nature03524. [DOI] [PubMed] [Google Scholar]
  2. Acar M, Pando BF, Arnold FH, Elowitz MB, van Oudenaarden A. A general mechanism for network-dosage compensation in gene circuits. Science. 2010;329(5999):1656–60. doi: 10.1126/science.1190544. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Arnold FH. Engineering proteins for nonnatural environments. The FASEB Journal. 1993;7(9):744–749. doi: 10.1096/fasebj.7.9.8330682. [DOI] [PubMed] [Google Scholar]
  4. Bao Z, Xiao H, Liang J, Zhang L, Xiong X, Sun N, Si T, Zhao H. Homology-Integrated CRISPR-Cas (HI-CRISPR) System for One-Step Multigene Disruption in Saccharomyces cerevisiae. ACS Synth Biol. 2015;4(5):585–594. doi: 10.1021/sb500255k. [DOI] [PubMed] [Google Scholar]
  5. Boeke JD, Trueheart J, Natsoulis G, Fink GR. 5-Fluoroorotic acid as a selective agent in yeast molecular genetics. Methods in Enzymology. 1987;154:164–175. doi: 10.1016/0076-6879(87)54076-9. [DOI] [PubMed] [Google Scholar]
  6. Cong L, Ran FA, Cox D, Lin S, Barretto R, Habib N, Hsu PD, Wu X, Jiang W, Marraffini LA, Zhang F. Multiplex genome engineering using CRISPR/Cas systems. Science. 2013;339(6121):819–823. doi: 10.1126/science.1231143. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. DiCarlo JE, Norville JE, Mali P, Rios X, Aach J, Church GM. Genome engineering in Saccharomyces cerevisiae using CRISPR-Cas systems. Nucleic Acids Res. 2013;41(7):4336–4343. doi: 10.1093/nar/gkt135. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Dion V, Coulombe B. Interactions of a DNA-bound transcriptional activator with the TBP-TFIIA-TFIIB-promoter quaternary complex. J Biol Chem. 2003;278(13):11495–501. doi: 10.1074/jbc.M211938200. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Horwitz AA, Walter JM, Schubert MG, Kung SH, Hawkins K, Platt DM, Hernday AD, Mahatdejkul-Meadows T, Szeto W, Chandran SS, Newman JD. Efficient Multiplexed Integration of Synergistic Alleles and Metabolic Pathways in Yeasts via CRISPR-Cas. Cell Syst. 2015;1(1):88–96. doi: 10.1016/j.cels.2015.02.001. [DOI] [PubMed] [Google Scholar]
  10. Jiang W, Bikard D, Cox D, Zhang F, Marraffini LA. RNA-guided editing of bacterial genomes using CRISPR-Cas systems. Nat Biotechnol. 2013;31(3):233–239. doi: 10.1038/nbt.2508. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. John TPS, Davis RW. The organization and transcription of the galactose gene cluster of Saccharomyces. J Mol Biol. 1981;152(2):285–314. doi: 10.1016/0022-2836(81)90244-8. [DOI] [PubMed] [Google Scholar]
  12. Johnston M. A Model Fungal Gene Regulatory Mechanism: the GAL Genes of Saccharomyces Cerevisiae. MicroBiological Reviews. 1987;51(4):458–476. doi: 10.1128/mr.51.4.458-476.1987. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Johnston M, Davis RW. Sequences That Regulate the Divergent GALJ-GALIO Promoter in Saccharomyces cerevisiae. Mol Cell Biol. 1984;4(8):1440–1448. doi: 10.1128/mcb.4.8.1440. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Kellis M, Patterson N, Endrizzi M, Birren B, Lander ES. Sequencing and comparison of yeast species to identify genes and regulatory elements. Nature. 2003;423:241–254. doi: 10.1038/nature01644. [DOI] [PubMed] [Google Scholar]
  15. Kuchner O, Arnold FH. Directed evolution of enzyme catalysts. Trends in Biotechnology. 1997;15(12):523–530. doi: 10.1016/S0167-7799(97)01138-4. [DOI] [PubMed] [Google Scholar]
  16. Li-En Jao SRW, Chen Wenbiao. Efficient multiplex biallelic zebrafish genome editing using a CRISPR nuclease system. Proc Natl Acad Sci U S A. 2013;110(34):13904–13908. doi: 10.1073/pnas.1308335110. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Mali P, Yang L, Esvelt KM, Aach J, Guell M, DiCarlo JE, Norville JE, Church GM. RNA-guided human genome engineering via Cas9. Science. 2013;339(6121):823–826. doi: 10.1126/science.1232033. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Mans R, van Rossum HM, Wijsman M, Backx A, Kuijpers NG, van den Broek M, Daran-Lapujade P, Pronk JT, van Maris AJ, Daran JM. CRISPR/Cas9: a molecular Swiss army knife for simultaneous introduction of multiple genetic modifications in Saccharomyces cerevisiae. FEMS Yeast Res. 2015;15(2):1–15. doi: 10.1093/femsyr/fov004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Mizutani A, Tanaka M. Regions of GAL4 critical for binding to a promoter in vivo revealed by a visual DNA-binding analysis. The EMBO Journal. 2003;22(9):2178–2187. doi: 10.1093/emboj/cdg220. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Peng G, Hopper JE. Gene activation by interaction of an inhibitor with a cytoplasmic signaling protein. Proc Natl Acad Sci U S A. 2002;99(13):8548–53. doi: 10.1073/pnas.142100099. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Ryan OW, Poddar S, Cate JH. CRISPR-Cas9 Genome Engineering in Saccharomyces cerevisiae Cells. Cold Spring Harb Protoc. 2016;2016(6):525–533. doi: 10.1101/pdb.prot086827. [DOI] [PubMed] [Google Scholar]
  22. Ryan OW, Skerker JM, Maurer MJ, Li X, Tsai JC, Poddar S, Lee ME, DeLoache W, Dueber JE, Arkin AP, Cate JH. Selection of chromosomal DNA libraries using a multiplex CRISPR system. Elife. 2014;3:e03703. doi: 10.7554/eLife.03703. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Shao Z, Arnold FH. Engineering new functions and altering existing functions. Current Opinion in Structural Biology. 1996;6(4):513–518. doi: 10.1016/s0959-440x(96)80117-5. [DOI] [PubMed] [Google Scholar]
  24. Sharon E, Kalma Y, Sharp A, Raveh-Sadka T, Levo M, Zeevi D, Keren L, Yakhini Z, Weinberger A, Segal E. Inferring gene regulatory logic from high-throughput measurements of thousands of systematically designed promoters. Nat Biotech. 2012;30(6):521–530. doi: 10.1038/nbt.2205. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Wang T, Wei JJ, Sabatini DM, Lander ES. Genetic screens in human cells using the CRISPR-Cas9 system. Science. 2014;343(6166):80–84. doi: 10.1126/science.1246981. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS835858-supplement-1.pdf (722.8KB, pdf)
2
NIHMS835858-supplement-2.pdf (27.1KB, pdf)

RESOURCES