Multiplexed Simian Immunodeficiency Virus-Specific Paired RNA-Guided Cas9 Nickases Inactivate Proviral DNA

ABSTRACT

Human and simian immunodeficiency virus (HIV and SIV) infections establish lifelong reservoirs of cells harboring an integrated proviral genome. Genome editing CRISPR-associated Cas9 nucleases, combined with SIV-specific guiding RNA (gRNA) molecules, inactivate integrated provirus DNA in vitro and in animal models. We generated RNA-guided Cas9 nucleases (RGNu) and nickases (RGNi) targeting conserved SIV regions with no homology in the human or rhesus macaque genome. Assays in cells cotransfected with SIV provirus and plasmids coding for RGNus identified SIV long terminal repeat (LTR), trans-activation response (TAR) element, and ribosome slip site (RSS) regions as the most effective at virus suppression; RGNi targeting these regions inhibited virus production significantly. Multiplex plasmids that coexpressed these three RGNu (Nu3), or six (three pairs) RGNi (Ni6), were more efficient at virus suppression than any combination of individual RGNu and RGNi plasmids. Both Nu3 and Ni6 plasmids were tested in lymphoid cells chronically infected with SIV_mac239, and whole-genome sequencing was used to determine on- and off-target mutations. Treatment with these all-in-one plasmids resulted in similar levels of mutations of viral sequences from the cellular genome; Nu3 induced indels at the 3 SIV-specific sites, whereas for Ni6 indels were present at the LTR and TAR sites. Levels of off-target effects detected by two different algorithms were indistinguishable from background mutations. In summary, we demonstrate that Cas9 nickase in association with gRNA pairs can specifically eliminate parts of the integrated provirus DNA; also, we show that careful design of an all-in-one plasmid coding for 3 gRNAs and Cas9 nuclease inhibits SIV production with undetectable off-target mutations, making these tools a desirable prospect for moving into animal studies.

IMPORTANCE Our approach to HIV cure, utilizing the translatable SIV/rhesus macaque model system, aims at provirus inactivation and its removal with the least possible off-target side effects. We developed single molecules that delivered either three truncated SIV-specific gRNAs along with Cas9 nuclease or three pairs of SIV-specific gRNAs (six individual gRNAs) along with Cas9 nickase to enhance efficacy of on-target mutagenesis. Whole-genome sequencing demonstrated effective SIV sequence mutation and inactivation and the absence of demonstrable off-target mutations. These results open the possibility to employ Cas9 variants that introduce single-strand DNA breaks to eliminate integrated proviral DNA.

INTRODUCTION

AIDS, caused by human immunodeficiency virus (HIV), continues to be a worldwide health problem, with the World Health Organization estimating that 38 million people were infected worldwide by HIV by the end of 2019. The best approach to reduce AIDS-related deaths is early diagnosis and lifelong combination antiretroviral therapy (cART) (1–3). While cART reduces virus replication, it does not eliminate virus-infected cells completely; consequently, there is a continued need for the development of new HIV therapeutics that target all virus-infected cells. As with HIV infection in humans, infection with simian immunodeficiency virus (SIV) in rhesus macaques results in a persistent, lifelong infection that progresses to AIDS (4), making it one of the best animal models to study HIV pathogenesis and therapy.

Therapies targeting HIV-1 proviral DNA to eliminate all virus-infected cells, including latently infected cells constituting viral reservoirs, is one cure strategy that has advanced rapidly in recent years. The first approaches to disrupt integrated HIV DNA involved the development of evolved Cre recombinases (5, 6), zinc finger nucleases (ZFNs), and transcription activator-like effector nucleases (TALENs) (reviewed in reference 7). More recently, the clustered, regularly interspaced, short palindromic repeats (CRISPRs) and CRISPR-associated nuclease (Cas) (CRISPR/Cas) system has been adapted for genome editing with revolutionary success to modify genes located in the human genome (8, 9). Cas9 is a nuclease produced by Streptococcus pyogenes (SpCas9) that binds to protospacer-adjacent motifs (PAMs) of 5′-NGG-3′ nucleotide sequence (10) directed by guiding RNAs (gRNAs) to induce DNA double-strand breaks (DSBs) (11, 12). For the Cas9 nuclease (Cas9Nu) to function, Cas9 must first undergo a conformational change that occurs when it associates with gRNA comprised of CRISPR RNA (crRNA) and transactivating crRNA (tracrRNA), two strands of RNA that anneal to each other (13, 14). Once Cas9Nu recognizes a PAM, the helicase domain of Cas9 unwinds the DNA and a 20-nucleotide (nt) region of the crRNA, called the protospacer region, binds to the DNA strand opposite of the PAM via Watson and Crick base pairing (9). If there is absolute complementarity between the protospacer and genomic DNA, then the Cas9Nu cleaves both strands of DNA 3 bp upstream of the PAM sequence with two enzymatic domains: the RuvC domain (noncomplementary to the gRNA) and the HNH domain (complementary to the gRNA) (9, 12, 15). These DSBs are generally repaired imperfectly by the host nonhomologous end-joining (NHEJ) DNA repair enzymes resulting in insertions and deletions (indels), a process that frequently renders a gene nonfunctional (16). There has been success in preventing HIV-1 replication in infected cell lines and even in primary CD4⁺ T cells (17, 18). Removal of integrated provirus DNA from infected cells can occur when DSBs occur at the long terminal repeats (LTRs) located at the 5′ and 3′ ends of the proviral genome (18–25). One caveat to using a single gRNA to target HIV provirus is the appearance of mutations that allow for viral escape (26–29), similar to what is observed for antiretroviral drugs. Consequently, it was found that targeting highly conserved HIV sequences delayed the appearance of escape mutants (27), and the combination of two strongly suppressive gRNAs inhibited HIV more efficiently (30). More recently, an important proof-of-concept study in SIV-infected rhesus macaques inoculated with adeno-associated virus 9 (AAV9)-CRISPR/Cas9 demonstrated that these molecular tools can lead to the in vivo removal of fragments on integrated proviral DNA from infected cells (31).

While RNA-guided Cas9 nucleases (RGNu) show great promise for precise genome editing, there is the risk of Cas9Nu-induced indel formation at sites in the host genome that are incompletely complementary to the protospacer region of the gRNA, often referred to as off-target cleavages. Multiple studies have shown that approximately 3 to 5 mismatch bases of the gRNA in regions distal to the PAM site are permissive to Cas9 cleavage events (32–34). Various improvements to the design of the gRNA have been made in order to increase the precision of Cas9Nu and reduce off-target cleavage while maintaining the robustness of on-target mutagenesis; among these modifications, there is (i) truncation of the protospacer region to 17 nt (35), (ii) using a chimeric, single gRNA that fuses the crRNA and tracrRNA into one RNA strand that assembles into a structure identical to crRNA and tracrRNA duplex (9), and (iii) increasing the length of the tracrRNA 3′ tail (32). An additional approach has been the use of a nickase mutant Cas9 (Cas9ni), which induces a single-strand DNA break and requires the concerted action of two different adjacent gRNAs to mimic DSBs, minimizing the risk of off-target effects (34, 36–38). Use of paired Cas9ni can be as effective in inducing indels at target site as single nucleases while reducing the frequency of off-target indel formation to below the limit of detection using next-generation sequencing (NGS) (34, 36, 38). With RNA-guided Cas9ni (RGNi), one needs to generate two different gRNAs per provirus target and test their efficiency of indel formation compared to the single gRNA associated with Cas9.

Here, we describe the development of very effective SIV-specific gRNAs and show that multiplexed RGNu and RGNi plasmids decrease the level of virus production in a chronically SIV-infected T-cell line, leading to the recovery of CD4 surface expression. We also provide evidence that these changes are associated with the induction of specific mutations in integrated proviral DNA at the expected Cas9 cleavage sites, with little effect outside the target sequences.

RESULTS

SIV-specific RGNu inhibit virus production in a cotransfection model of HEK293T cells.

We developed CRISPR/Cas9 molecules to induce indels in conserved, essential regions of the SIV provirus with the purpose of disrupting SIV replication. Seven conserved regions of SIV proviral DNA were identified and targeted (Fig. 1). Conserved regions of the SIV proviral DNA were identified by aligning SIV_mac239 and SIV_mac251 lineage viruses (39) in Geneious R7 and were mined for potential protospacer sequences and adjacent PAM sites. We designed our gRNA to be 17 nt in length, as truncated gRNAs increase specificity while not losing any efficacy in inducing indels (35). An area of the LTR U3 region was targeted by LTR1, LTR2, LTR3, LTR4, and LTR5 gRNA constructs. The trans-activation response (TAR) element in the LTR repeat (R) region was targeted with TAR1, TAR2, TAR3, TAR4, and TAR5 constructs. These LTR and TAR sequences were present at both the 5′ and 3′ LTRs. The tRNA binding site was targeted by tRNA1 and tRNA2 constructs, of which tRNA1 is in the LTR region. The matrix protein was targeted by Gag1 and Gag2 constructs. There were four gRNA constructs that targeted the ribosome slip site (RSS): RSS1, RSS2, RSS3, and RSS4. The reverse transcriptase (RT) and integrase (IN) enzymes were targeted by RT1 and RT2 and by INT1 and INT2 constructs, respectively.

FIG 1.

Schematic representation of gRNA targets in the SIVmac239 proviral genome. Seven conserved regions of the SIVmac proviral genome were targeted with 22 gRNAs. Yellow boxes symbolize the location of the PAM. Green arrows represent the location and orientation of the gRNAs. Red arrows indicate the target location for Cas9.

Each one of the 22 individual SIV-specific gRNAs (see Table S1 in the supplemental material) was screened for its ability to reduce virus production in a cotransfection model of SIV infection using HEK293T cells (40). A plasmid encoding green fluorescent protein (GFP), Cas9Nu, and a SIV-specific gRNA was cotransfected into HEK293T cells along with a plasmid containing the whole SIV provirus (pMA239) to detect proviral inactivation correlating with indel formation at the target sites. Transfection of HEK293T cells with pMA239 results in production of infectious SIV, which can be measured by quantifying SIV Gag p27 protein from the supernatant posttransfection. However, new rounds of infection do not occur since HEK293T cells do not express CD4/CCR5 and are not susceptible to infection with SIV.

When plasmids encoding RGNu LTR1, LTR2, LTR3, TAR1, or RSS2 were cotransfected with pMA239 into HEK293T cells, there was a dramatic inhibition of virus production as detected by a reduction of SIV Gag p27 capsid protein in the supernatant (Fig. 2A). On the other hand, TAR2, tRNA1, RSS1, RT1, and INT1 RGNu significantly inhibited virus production compared to empty guide RGNu control (Fig. 2A) but not as impressively as LTR1, LTR2, LTR3, TAR1, TAR3, RSS2, and RSS4 as seen in the percent inhibition of SIV production over empty Cas9 control conditions (Fig. 2B). Pairing LTR1 and LTR2 RGNu or LTR1 and LTR3 RGNu suppressed virus production as significantly as single LTR1, LTR2, or LTR3 RGNu. The same trend occurred when pairing TAR1 and TAR2 RGNu, TAR1 and TAR3 RGNu, and RSS2 and RSS4 RGNu (Fig. 2A). Interestingly, for some of our pairs where one gRNA inhibited virus production and one could not, such as tRNA-specific gRNAs, we observed an additive effect in the inhibition of virus production, suggesting that one RGNu did not negatively affect the other but instead that they worked in concert to inhibit SIV production. Lastly, neither Gag1, Gag2, nor paired Gag1 and Gag2 RGNu suppressed the production of p27 in the supernatant (Fig. 2A), suggesting that this genomic area was not an optimal site for CRISPR/Cas9 targeting RGNu. We also demonstrated that SIV-specific gRNAs that were not efficient at suppressing production of SIV p27, such as RT-, IN-, or Gag-specific gRNAs, generated virus able to induce infection in the permissive cell line CEM-x-174 (Fig. 2C), indicating that production of SIV p27 correlated directly with levels of infectious virus, even when using gRNAs that did not target the capsid protein.

FIG 2.

Screening of RGNu for inhibition of SIV production. (A) HEK293T cells were transfected with pMA293 and RGNu-encoding plasmids, and after 24 h the supernatant was harvested to measure SIV Gag p27 capsid protein. (B) The percent inhibition of SIV Gag p27 capsid protein was calculated against RGNu empty control. Assays were performed three times in triplicate each time. (C) Infectivity of SIV_mac after treatment with RNA-guided Cas9 nucleases. Supernatant of transfected HEK cells was used to infect the SIV-permissive cell line CEM-x-174. SIV p27 concentration was determined in the supernatant of infected CEM-x-174 cells with a Luminex assay. Statistical analysis was performed using Student’s t test with Welch’s correction. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001. ns, not significant. Error bars represent the standard errors of the means (SEM). The level of detection (LOD) of SIV Gag p27 is 1.98 pg/ml.

The ability of some of the LTR-, TAR-, and RSS-specific gRNAs to cleave SIV DNA was corroborated with reconstituted ribonucleoproteins (RNP) made of purified Cas9Nu and synthetic gRNA. After in vitro incubation of pMA239 DNA with synthetic RNP containing LTR3, TAR1, or RSS2 gRNAs, gel electrophoresis analysis showed cleavage at the two LTR sites for LTR3 or TAR1 and a single cut for RSS2 (Fig. 3A). Next-generation sequencing analyses of DNA from transfected HEK cells also demonstrated the accumulation of indels in the SIV areas targeted by the LTR-, TAR-, and RSS-specific gRNAs (Fig. 3B). In summary, gRNAs targeting Cas9Nu to the LTR, TAR, and RSS regions of SIV were efficient at inducing indels and inhibiting virus production.

FIG 3.

Analysis of SIV editing by RGNu on plasmid pMA239. (A) In vitro cleavage of pMA239 DNA by SIV-specific gRNAs and Cas9Nu. (Left) Diagram showing the location of LTR3, TAR1, and RSS2 target sequences; (right) agarose gel electrophoresis of pMA239 DNA incubated with reconstituted ribonucleoprotein made of purified Cas9Nu and different SIV-specific gRNAs. (B) Determination of frequency of SIV-specific insertions and deletions by deep sequencing (MiSeq). DNA was purified from HEK cells transfected with pMA239 and plasmids expressing either LTR-specific (top graph), TAR-specific (middle graph), or RSS-specific (bottom graph) gRNAs. Curves represent the percentage of mutations (insertion and deletions) identified at each SIV nucleotide position.

SIV-specific paired RGNi and multiplexed RGNu and RGNi inhibit virus production in a cotransfection model of HEK293T cells.

In the previous section, we showed that LTR1, LTR2, LTR3, TAR1, TAR3, RSS2, and RSS4 gRNAs paired with Cas9 nuclease were able to dramatically inhibit virus production in a HEK293T cotransfection model. These gRNAs were next tested with Cas9 nickase as paired RGNi in the same HEK293T cotransfection model to measure virus production inhibition. We observed some SIV inhibition when Cas9 nickase was coexpressed with individual LTR1, LTR2, and LTR3 gRNAs but not at the dramatic level seen for RGNu LTR1, LTR2, and LTR3 (Fig. 2A). Virus production was reduced 100-fold when pairing LTR1 and LTR2 RGNi and pairing LTR1 and LTR3 RGNi. Interestingly, RGNi TAR1 inhibited virus production as well as RGNu TAR1, while paired TAR1 and TAR3 RGNi had an SIV inhibitory effect intermediate between RGNi1 and RGNi3. Finally, paired RGNi RSS2 and RSS4 inhibited virus production similarly to RGNi RSS4 (Fig. 4A). These data suggest that it is possible to inhibit virus replication with paired RGNi, but we observed that the degree of suppression was not as effective as with single or paired RGNu.

FIG 4.

Screening of RGNi and multiplex plasmids for inhibition of SIV production. (A) HEK293T cells were transfected with pMA293 and RGNi or multiplex plasmids, and after 24 h, supernatant was harvested for detection of SIV Gag p27 capsid protein. Assays were performed in triplicate and repeated three times. Statistical analysis was performed using Student’s t test with Welch’s correction, comparing all Nis to empty control. *, P ≤ 0.05; **, P ≤ 0.01; ****, P ≤ 0.0001. Error bars represent the SEM. (B to F) At 24 h posttransfection, cells were harvested, intracellularly stained with an anti-SIV Gag p27 antibody, and analyzed by flow cytometry. HEK293T cells were either not transfected (B), cotransfected with RGNu with empty gRNA (C), cotransfected with RGNi with empty gRNA (D), cotransfected with Nu3 (E), or cotransfected with Ni6 (F).

To further enhance the inhibition of virus production in our HEK293T cotransfection model, we multiplexed the three best gRNAs targeting the LTR, TAR, and RSS region by creating a single plasmid that encoded Cas9 nuclease and cotranscribed the LTR3, TAR1, and RSS2 gRNAs (designated Nu3). We also combined the three best pairs of gRNAs in a single plasmid that encoded Cas9 nickase and cotranscribed LTR1, LTR3, TAR1, TAR3, RSS2, and RSS4 gRNAs (designated Ni6). The Nu3 multiplex plasmid inhibited virus production from HEK293T cells by 5 orders of magnitude, almost to the limit of detection of our SIV p27 assay. Although not as effective as Nu3, the Ni6 multiplex plasmid inhibited virus production from HEK293T cells by 3 orders of magnitude, in an additive manner compared to the paired RGNi targeting LTR, TAR, and RSS regions separately (Fig. 4A).

In addition to measuring virus production in the supernatant of Nu3- and Ni6-transfected HEK293T cells, the percentage of cells producing virus was quantified by intracellular staining of cells with an anti-p27 antibody. Less than one percent of HEK293T cells were positive for intracellular SIV p27 with either Nu3 (Fig. 4E) or Ni6 (Fig. 4F) multiplex plasmids, a dramatic inhibition of p27 compared to matched RGNu (Fig. 4C) or RGNi (Fig. 4D) empty control plasmids. The lack of p27-positive cells corroborates the Luminex assay p27 data (Fig. 4A) showing almost complete inhibition of virus production.

In summary, we were able to produce plasmid DNA molecules that expressed multiple gRNAs and were effective at inhibiting SIV production in the context of both Cas9 nuclease and nickase.

Targeting integrated proviral DNA with RGNu and RGNi to inhibit SIV replication in a T-lymphoblast cell model.

Our RGNu, RGNi, and multiplexed constructs inhibited SIV production in an HEK293T cotransfection model when the proviral DNA target was contained in a plasmid vector. In order for RGNu and RGNi to be successful in inhibiting SIV replication in vivo, they first must be able to target provirus integrated into host cell chromosomes of lymphoid cells and induce indels at gRNA-targeting cleavage sites. SIV, like HIV, infects primarily CD4⁺ cells via the CCR5 coreceptor. Therefore, we used CEM.NK^R-CCR5 cells, a T-lymphoblast line that expresses CD4, CXCR4, and CCR5 and can be infected by SIV_mac strains, as a cell model to study proviral targeting by our RGNu and RGNi (40, 41). CEM.NK^R-CCR5 cells were infected with SIV_mac239 and CD4-negative cells were sorted and serially diluted to isolate chronically infected clones. Clone 2C11 is an isolate that is predominantly CD4 negative and produces SIV at a high rate. We used the 2C11 cells as a biologically relevant cell line to test whether our RGNu, paired RGNi, Nu3, and Ni6 constructs can target integrated proviral DNA to suppress virus production. We nucleofected 2C11 cells with a cocktail of three plasmids encoding RGNu (LTR3, TAR1, and RSS2 gRNAs), a cocktail of six RGNi-coding plasmids (LTR 1/3, TAR 1/3, and RSS 2/4), multiplex RGNu Nu3 plasmid, or multiplex RGNi Ni6 plasmid, sorted cells that were GFP positive, and then cultured the sorted cells for 15 to 17 days.

Flow cytometry analysis for surface expression of CD4 and intracellular detection of SIV p27 for untransfected 2C11 cells showed less than 9% expression of CD4 (Fig. 5A); however, more than 92% of 2C11 cells became CD4⁺ SIV p27⁻ after treatment with Nu3 (Fig. 5B). Treatment with Ni6 resulted in the CD4⁺ SIV p27⁻ cells increasing to 80% (Fig. 5C). Western blot analysis of 2C11 cells treated with Nu3 and Ni6 showed that SIV Gag expression was undetectable (for Nu3) or barely detectable (for Ni6) compared to that in untreated 2C11 cells, further confirming that virus production was inhibited in Nu3- and Ni6-treated 2C11 cells (Fig. 5D).

FIG 5.

RGNu and RGNi restore CD4 expression and decrease SIV p27 production in 2C11 cells. 2C11 cells were nucleofected with control plasmid (Nu Empty), a cocktail of three RGNu plasmids (NuLTR/TAR/RSS), a cocktail of six RGNi plasmids (NiLTR/TAR/RSS), the multiplex Nu3 plasmid, or the multiplex Ni6 plasmid; cells were sorted for GFP expression and cultured for 15 to 17 days. (A to E) At day 15 to 17 of culture, cells were stained for flow cytometry to assess CD4 and p27 expression. (A) 2C11 cells not nucleofected. (B) 2C11 cells nucleofected with Nu3 multiplex plasmid. (C) 2C11 cells nucleofected with Ni6 multiplex plasmid. (D) Western blot of transfected 2C11 cells probed with serum from an SIV-infected rhesus macaque (top) or anti-GAPDH antibody (bottom). Lane 1, Ni6; lane 2, Nu3; lane 3, 2C11 cells; lane 4, uninfected CEM cells. (E) Percentage of CD4⁺ p27⁻ 2C11 after nucleofection, sorting, and 2-week incubation. (F) Inhibition of SIV p27 in the supernatant of electroporated 2C11 cells. Twice a week, the entire supernatant was collected and analyzed for SIV p27 by Luminex assay. Viral inhibition is calculated in reference to SIV p27 content of 2C11 cells treated with an empty RGNu plasmid.

The importance of coexpression of multiple gRNAs along with Cas9 nuclease or nickase was also identified in these assays. While transfection of 2C11 cells with a cocktail of the RGNu plasmids LTR1, TAR1, and RSS2 resulted in about 75% of 2C11 cells becoming CD4⁺ SIV p27⁻, this percentage was still below the 92% level achieved by Nu3. However, multiplexing contribution was even more relevant for the RGNi plasmids, which showed that transfection with a cocktail including the six RGNi plasmids LTR1, LTR3, TAR1, TAR 3, RSS2, and RSS4 resulted in less than 15% of the 2C11 cells being CD4⁺ SIV p27⁻, while transfection with Ni6 increased that number to 80%; this percentage was even higher than the one produced by the cocktail of 3 RGNu (Fig. 5E). These data confirm that multiplexing gRNAs onto one plasmid molecule is more efficient in suppressing SIV production in 2C11 cells, ensuring that all gRNAs are expressed simultaneously in the same cell.

As expected, the number of CD4⁺ SIV p27⁻ 2C11 cells at the end of the 2-week culture directly correlated with viral inhibition measured as SIV p27 released in the supernatant of transfected cells. Supernatant from the 15- to 17-day cultures postsorting was collected twice a week, and SIV p27 levels were measured to quantify virus production using a Luminex assay. With the 2C11 cells treated with the Nu3 or Ni6 multiplex plasmids, approximately 2 logs less virus was detected in the supernatant compared to RGNu empty control, which corresponded to almost 95% inhibition of virus production (Fig. 5F).

Multiplexed RGNu and RGNi induce mutations in integrated proviral DNA.

The multiplexed RGNu and RGNi plasmids suppressed virus production in 2C11 cells, which correlated with a reduction in the percentage of virus-producing cells and inversely correlated with the cellular expression of the CD4 surface marker, providing indirect evidence that proviral DNA is targeted by our CRISPR reagents. To investigate whether the suppression of virus production was directly caused by Cas9-induced DSBs/indel formation, mutations at all three SIV target sites of Nu3 and Ni6 were quantified via MiSeq NGS. Analysis of 2C11 cells treated with Nu3 showed many unique and large proviral deletions and a large number of small identical deletions and insertions at the LTR3 (Fig. 6A) and TAR1 (Fig. 6C) cleavage sites. Ni6-treated 2C11 cells showed both large and small proviral deletions and small insertions that spanned both LTR1 and LTR3 (Fig. 6B) and TAR1 and TAR3 (Fig. 6D) cleavage sites, indicating that the paired RGNi targeting the LTR and TAR regions were able to more effectively edit the integrated proviral DNA. Cells treated with Nu3 showed that the RSS region contained many unique and large deletions and a large number of small, identical insertions at the RSS2 cleavage site (Fig. 6E), but Ni6 did not produce a similar large pattern of indels generated between the RSS2 and RSS4 cleavage sites, suggesting that this pair of gRNAs targeting the RSS region did little to edit SIV proviral DNA (Fig. 6F).

FIG 6.

Incidence of indels in 2C11 proviral DNA with multiplex RGNu and RGNi. Shown are 2-dimensional (2D) indel plots of insertion and deletion frequencies at every nucleotide position generated using MiSeq amplicon data. DNA was isolated from 2C11 cells cultured 15 days after sorting. Vertical and horizontal lines within each plot indicate the Cas9 cleavage site location targeted by each gRNA: LTR regions of Nu3 (A) and Ni6 (B) where the first line represents LTR1 and the second line represents LTR3, TAR regions of Nu3 (C) and Ni6 (D) where the first line represents TAR1 and the second line represents TAR3, and RSS regions of Nu3 (E) and Ni6 (F) where the first line represents the site for RSS2 and the second line represents RSS4. Each dot represents a unique indel, with the size of the dot proportional to the number of supporting reads and the color of the dot indicating the relative density of unique indels in that portion of the plot. All dots within the upper left triangle of the plot represent insertions; those in the lower right triangle represent deletions.

In order to identify the direct effect of mutations on the SIV genome, we quantified the percentage of MiSeq reads covering the targeted regions that were identical (perfect match) to the original proviral DNA sequence (i.e., no mutations had been introduced). Across the three targeted regions, we observed the same relative editing efficiencies in both the Nu3- and Ni6-nucleofected cells, with the highest rate of mutated proviruses observed for LTR, followed by TAR and RSS (Fig. 7B). However, across all three target sites, we observed a higher percentage of nonmutated (i.e., perfect match) amplicons in Ni6-nucleofected cells (1.4 to 1.8 times higher than in Nu3-nucleofected cells), indicating that Nu3 resulted in more efficient proviral editing at all three sites (Fig. 6). Considering all three regions together, and assuming that mutations in each region were independently distributed, the estimated percentages of viable provirus were 6.3% for Nu3-nucleofected 2C11 cells and 25% for Ni6-nucleofected 2C11 cells (Fig. 7B). These data show that Nu3 and Ni6 are both effective at suppressing virus production via Cas9-induced DSBs/indel formation, with Nu3 having a more robust suppression of virus production due to greater indel incidence.

FIG 7.

NGS analysis of genomic data from 2C11 cells and cells treated with Nu3 and Ni6. (A) Depth of coverage by WGS of the genomic area of SIV integration. Samples were normalized to the average depth of coverage within the first 4,000 bp of the cellular genome upstream of the proviral insertion. Red dashed vertical lines represent the start and end of SIV proviral DNA on chromosome 9. (B) Percentage of MiSeq-generated amplicons identical to SIV_mac239. Shown are the percentages of identical (nonmutated) amplicons generated from Nu3- and Ni6-nucleofected 2C11 cells using MiSeq NGS analysis of the LTR, TAR, and RSS regions; “ALL” represents the estimated fraction of functional provirus in Nu3- and Ni6-treated 2C11 cells assuming that mutations in each region are independently distributed. (C) Number of single-nucleotide polymorphisms identified by WGS. (D) Number of indels identified by WGS. The number in the maroon circle represents mutations called by GATK, the number in the green circle represents mutations called by CRISPResso2, and the number in the overlapping circles represents mutations called by both GATK and CRISPResso2. All numbers were called within 10 bp upstream of PAM for all sgRNAs used and normalized by genome coverage.

WGS reveals single provirus integration in 2C11 cells and absence of significant off-target mutations for Nu3- and Ni6-treated cells.

The purpose of designing paired RGNi to target SIV provirus for CRISPR genome editing was to limit the risk of off-target indels to host chromosomes relative to RGNu. To determine the indel profile of multiplex-treated 2C11 cells that inhibit virus production, we employed whole-genome sequencing (WGS) analysis on 2C11 cells and 2C11 cells treated with Nu3 or Ni6 that were then sorted and cultured for 15 days. WGS was performed using NovaSeq with genome coverage rates of 39.3×, 38.8×, and 59.4× for untreated 2C11 cells, Nu3 2C11 cells, and Ni6 2C11 cells, respectively. WGS-generated sequences from untreated 2C11 cells and Nu3- and Ni6-treated 2C11 cells were aligned to the human genome GRCh38 (RefSeq accession no. GCF_000001405.39) assembly combined with the proviral sequence of SIV_mac239 (GenBank accession no. M33262.1). DELLY v. 0.8.1 (42) was used to concurrently identify structural variants in all three samples. DELLY identified a shared “breakend” structural variation between human chromosome 9 and SIV_mac239 in all the samples, suggesting that chromosome 9 is the site of the viral insertion. The SIV_mac239 provirus is integrated into the intron of the excision repair cross-complementing 6-like 2 (ERCC6L2) gene located on human chromosome 9 (Fig. S1). This single integrated provirus in all three cell cultures verified the clonal nature of our 2C11 cells. Normalized sequence coverage of SIV provirus in Nu3- or Ni6-treated 2C11 cells showed a substantial reduction compared to the SIV coverage seen for untreated 2C11 cells (Fig. 7A), suggesting that there was proviral excision in some of the treated cells.

Analyzing WGS data with both GATK (v. 4.1.0.0 [43]) and CRISPResso2 (v. 2.0.23 [44]) to call unique off-target single-nucleotide polymorphisms (SNPs) and indels not detected in untreated 2C11 cells, we performed genome-wide analyses and also concentrated on genomic regions within 10 bp upstream of PAM sites identified by Cas-OFFinder (45) for both Nu3- and Ni6-treated 2C11s. The direct analysis of the WGS data available for the parental 2C11 cell line and for each antiviral treatment, compared against the human reference genome, produced higher numbers of GATK and CRISPResso2 SNPs and indel calls for Ni6 cells than for 2C11 and Nu3 cells, due to the much higher sequencing coverage for Ni6- versus Nu3-treated cells (59.4× for Ni6 versus 38.8× for Nu3). When normalized by genome coverage depth, the number of variant calls for Ni6 was lower than what was generated for the parental 2C11 and Nu3 cell lines. When compared to the reference human genome, the normalized GATK variant calls were found to be similar to each other, and even the numbers of singletons (variants only present in a particular sample and not shared with the other samples) found for Nu3 cells (1,129 SNPs and 276 indels) and Ni6 cells (988 SNPs and 262 indels) were similar to or smaller than the ones found for 2C11 cells (1,140 SNPs and 268 indels). Genome-wide analyses with CRISPResso comparison, for variants present in Nu3- and Ni6-treated cells not present in 2C11 cells, also produced normalized lower numbers for Ni6 (129 SNPs and 103 indels) than for Nu3 (182 SNPs and 112 indels). When restricting the analyses to the 10-bp regions upstream of PAM sites, the total numbers of normalized SNPs (Fig. 7C) and indels (Fig. 7D) continued to be smaller for Ni6-treated cells than for Nu3-treated cells, based both on calls by each analysis program and on the calls shared by GATK and CRISPResso2. Interestingly, the number of normalized indel calls recognized by both programs was smaller than 1 (0.5 indel for Nu3 and 0.2 indel for Ni6 per genome coverage [Fig. 7D]), indicating that practically there were no off-target indels. These data suggest that the paired RGNi multiplex strategy was effective at inhibiting virus production while limiting the potential of off-target effects; however, neither Nu3 nor Ni6 plasmids generated detectable off-target indels identifiable by more than one analysis program.

DISCUSSION

CRISPR/Cas9 RGNu are genome editing reagents that can robustly target DNA with the purpose of eliminating gene expression through the introduction of indels. Thus, these reagents have been repurposed with the goal of suppressing HIV production by targeting viral genes as a gene editing therapeutic. Here, we demonstrate a novel strategy of genome editing with enhanced specificity using truncated gRNA, with paired nickases targeting three regions to successfully inactivate SIV. Here, we show that SIV-specific RGNu and paired RGNi can disrupt virus production via mutations caused by NHEJ repairs of DNA DSB both in a HEK293T cotransfection model of SIV infection and in an SIV-infected T-lymphoblast line. The SIV-infected T lymphoblasts that we developed were infected with SIV_mac239, which is a highly pathogenic clonal isolate of SIV_mac251 (46); however, virus-infected cells in vivo are not clonal and therefore pose a challenge for single-target RGNu therapies that may not recognize a mutated proviral target. Multiplexing three gRNAs or three gRNA pairs with RGNu or RGNi, respectively, may overcome this obstacle. In addition to possibly targeting a quasispecies more effectively, multiplexed RGNu and multiplexed paired RGNi may prevent viral escape seen by single RGNu (26–29). Multiplexing two gRNAs has been shown to reduce viral escape (26), and if those two individual gRNAs are strongly suppressive, the inactivation of HIV is enhanced (30). Therefore, by generating multiplex plasmids that target three different regions of the SIV provirus, we increase our chances of reducing viral escape and enhancing virus inactivation. In addition, all the LTR, TAR, and RSS gRNAs in our multiplexed plasmids target well-conserved sequences of SIV, which may likewise contribute to reducing viral escape. More importantly, and as demonstrated in these studies, multiplexing RGNu or RGNi into single DNA molecules resulted in a higher inhibitory effect than cocktails of individuals RGNu and RGNi.

With the HEK293T cotransfection model, it was observed that multiplex Nu3 and Ni6 plasmids could almost completely suppress virus production as determined by assessing supernatant SIV p27 levels; however, for 2C11 cells, the suppression was about 2 logs less than the RGNu empty control condition but was not complete. This discrepancy in the effectiveness of multiplex RGNu and RGNi in 2C11 cells compared to transfected HEK cells may be due, in part, to the integrated proviral DNA in 2C11 cells being wrapped around histones to form nucleosomes, since it is known that Cas9 has been shown to target nucleosome DNA less effectively in vitro and in vivo than pure DNA (47–50). While a 2-log decrease in virus detected in the culture supernatant may not appear to be a substantial and biologically relevant inhibition of virus replication, it has been shown that for HIV-infected individuals, having even a half-log decrease in viral loads correlates with a delay to AIDS death by two additional years and lower transmission rates (51). Further work is needed to demonstrate the effectiveness of our RGNu and RGNi tools in cells latently infected with SIV, in which the proviral DNA may not be easily accessible.

The utilization of WGS for analysis of treated 2C11 cells allowed us to determine that a single copy of SIV_mac239 proviral DNA was integrated into chromosome 9, which confirmed the clonal nature of our 2C11 cells. Additionally, WGS was used to investigate whether the multiplex paired RGNi construct would reduce off-target effects compared to multiplex RGNu. Potential off-target sites were identified using Cas-OFFinder, assigning parameters designed to detect a large sample of unbiased, possible off-target sites; these potential sites were examined for actual off-target indels using GATK and CRISPResso2. With the dozens of variant-calling software to detect single nucleotide variants (SNVs) and indels within WGS data, it is unclear which are best suited for detecting true Cas9-mediated indels. It is becoming more common to utilize two or three different algorithms to enhance the probability that variants called are true variants and not artifacts of the software’s unique algorithm (52, 53). The variant-calling software currently available vary a great deal, which is indicated by the small number of findings in common for both GATK and CRISPResso2 approaches for Nu3 and Ni6 despite analyzing similar sites identified by Cas-OFFinder. In any case, due to the low levels of detected SNPs and indels for RGNu and RGNi, it is unclear whether the rate of Cas9-mediated off-target indels is greater than the de novo rate of mutations occurring with cell culture passage.

Currently, the majority of WGS analysis for detecting off-targets has been carried out on a small number of cell clones (54, 55) because of the high costs of WGS methodology, thus limiting the use of WGS for finding low-frequency indels (56). Additionally, there is debate as to how much read coverage is adequate to detect these low-frequency indels, because higher coverage increases costs, which limits researchers in the number of samples they can investigate. Typically, 30 to 60× coverage WGS has been used to analyze low-frequency indels of clonal cell populations (56). We employed over 38× coverage WGS to detect off-target indels in 2C11 cells, which were not a clonal cell population after RGNu or RGNi exposure and were expanded for 15 days postnucleofection. Consequently, low-frequency indels may also have been expanded during culture. It is likely that we are unable to detect all low-frequency indels in these nonclonal cell populations with the 30 to 60× coverage WGS; nevertheless, we were able to identify a small difference in the numbers of off-target indels between the RGNu and paired RGNi constructs. Both Nu3 and Ni6 dramatically inhibited virus production via Cas9-mediated indel formation, but it is important to note that Ni6 was slightly less effective than Nu3 in suppressing virus production and inducing genome editing of proviral DNA. For Ni6, we observed that only LTR and TAR gRNA pairs were able to induce significant numbers of indels, while the RSS pair resulted in low levels (Fig. 6); reasons for this low RSS activity are not known, since the RSS2 and RSS4 pairs were efficient when tested in HEK cells (Fig. 3).

Genome editing for reaching an HIV cure offers a great potential. However, for this work to be translated into a realistic strategy for an HIV cure therapeutic, it must also be demonstrated that these effective, multiplexed RGNu and paired RGNi are highly specific and therefore confer limited off-target genome editing. Recently, Mancuso et al. (31) showed as proof of concept that CRISPR genome editing reagents packaged into an adeno-associated virus (AAV) can target SIV_mac239 in an in vivo rhesus macaque/SIV AIDS model. With these experiments, they relied on Staphylococcus aureus Cas9 (SaCas9) targeting two regions of SIV provirus using 3 gRNAs delivered via an AAV vector with wide tropism. AAV can accommodate only the smaller SaCas9 and therefore limits any SIV and HIV proviral targets to its unique PAM. Our strategy uses SpCas9, which allows for more potential HIV targets (57) and is therefore a more useful molecule to employ when translating this strategy to HIV-1 proviral targets. Also, in contrast, with our three gRNAs and six paired gRNAs, we are able to target three regions of SIV provirus: LTR, TAR, and RSS. Furthermore, our gRNAs are truncated, and we designed and implemented paired RGNi, both approaches intended to enhance specificity by reducing off-target events. Lastly, we were able to characterize the on- and off-target events of our CRISPR reagents using highly accurate and innovative NGS methods, which represent important tools for understanding the mechanism of our reagents before utilizing them for in vivo experiments. Importantly, we show that strategies that work to reduce off-target indels may sacrifice genome editing efficiency and are important considerations for designing Cas9 reagents for human therapeutics. However, we believe that inclusion of a third pair of gRNAs with capacity to induce indel formation will produce a Cas9 nickase-based system as efficient as the one that utilizes the Cas9 nuclease.

MATERIALS AND METHODS

Designing SIV-specific protospacer sequences.

Conserved regions of the SIV proviral DNA were found by aligning in Geneious R7 the following SIV_mac239 and SIV_mac251 lineage viruses (39) to find potential protospacer sequences and PAM sites: SIVMM251 (GenBank accession no. M19499), SIVMM239 (accession no. M33262), SIV1A11AA (accession no. M76764), SIVMM32H (accession no. D01065), and MAC239-87801 (accession no. AY587015). Finding paired, conserved PAM sites that had a tail-to-tail protospacer orientation on opposite strands of DNA that fit the formula 5′-CCN_(32–72)GG-3′ (37) was used to first assess areas of the SIV proviral sequence. Truncated protospacer regions of 17 nucleotides (35) were identified that were adjacent to Streptococcus pyogenes Cas9 PAM sites. Using this strategy, two potential protospacers were selected: a 5′ protospacer binding the sense strand of proviral DNA and a 3′ protospacer binding the antisense strand. Once candidate SIV-specific protospacer targets were found to meet the paired Cas9 nickase parameters, the sequences were searched in both the human and rhesus macaque genomes (called a BLAST search) to look for similar cleavage sites that might produce off-target indels. BLAST parameters in the GenBank NCBI NIH nucleotide BLAST “blastn suite” consisted of the full 17-nucleotide protospacer sequences plus the four potential PAM sites that would allow for RGNu cleavage (AGG, CGG, GGG, and TGG). For searching for potential off-target cleavage events in the human genome, sequences were BLAST searched against “Human genomic plus transcript (Human G+T)” in the database “GPIPE/9606/current/all_top_level (Homo sapiens GRCh38.p12 [GCF_000001405.38]) chromosomes plus unplaced and unlocalized scaffolds (reference assembly in Annotation Release 109)” and “GPIPE/9606/current/rna (Homo sapiens Annotation Release 109 RNAs),” which are the human genome and human transcriptome, respectively. Because of the short 20 nucleotide sequence imputed, the search parameters were adjusted to search for a short input sequence automatically by “blastn suite.” The BLAST conditions were the same for rhesus macaques as for humans but using the “rhesus macaque (taxid:9544)” database. If any of the protospacer with PAM sequences hit a 100% match in either the human or rhesus genome, that candidate protospacer target was eliminated from the potential SIV proviral target. If there were 19 or 18 nucleotides that matched the 20-nucleotide sequence, only those with PAM matches were excluded. The oligonucleotide sequences that were ordered from Life Technologies to generate gRNAs are listed in Table S1.

Generating plasmid-encoded Cas9 variants and gRNA.

Top and bottom protospacer oligonucleotides were resuspended to 100 μM in water, and pairs were phosphorylated with T4 polynucleotide kinase and annealed together (36). Protospacers were ligated into BbsI-digested and Antarctic phosphatase-treated pSpCas9(BB)-2A-GFP (pX458; Addgene; plasmid 48138) for Cas9 nuclease and pSpCas9n(BB)-2A-GFP (pX461; Addgene; plasmid 48140) for Cas9 nickase; both plasmids were gifts from Feng Zhang (36). Plasmids that contained protospacer sequences were verified by Sanger sequencing.

Golden Gate assembly of multiplexed plasmids.

To express multiple gRNAs on a single plasmid encoding Cas9-2A-GFP or Cas9n-2A-GFP, we utilized a Golden Gate assembly method with a CRISPR kit used for constructing multiplex CRISPR/Cas9 vectors that was a gift from Takashi Yamamoto (Addgene; kit no. 1000000054) (58). Briefly, the TAR1 protospacer oligonucleotide was cloned into the pX330A_1×3 and pX330A_D10A_1×6 plasmids that encode SpCas9-2A-GFP and SpCas9n-2A-GFP, respectively. The LTR3 protospacer oligonucleotide was cloned into pX330S_2, RSS2 into pX330S_3, LTR1 into pX330S_4, TAR3 into pX330S_5, and RSS4 into pX330S_6. Then using the Golden Gate assembly method of a single reaction, we digested all plasmids with BsaI and ligated them into assembly using T4 DNA ligase, and the resulting plasmids that were transformed into Zymo 5α competent Escherichia coli were pX330-3 (LTR3, TAR1, and RSS2 gRNA and Cas9) and pX330-6 (LTR1, LTR3, TAR1, TAR3, RSS2, and RSS4 gRNA and Cas9n). We then cloned the 2A-GFP sequence from pX458 into the pX330 Nu3 and pX330 Ni6 plasmids by EcoRV and NotI sites to generate pX330-GFP-3 (Nu3) and pX330-GFP-6 (Ni6). Plasmids were determined to have the correct gRNA by PCR amplification, restriction endonuclease digestions, and Sanger sequencing.

Cell lines.

Human embryonic kidney cell line 293T/17 (HEK293T, ATCC CRL-11268) was maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 5% fetal bovine serum (FBS), 2 mM GlutaMAX (Life Technologies), 25 mM HEPES, 1 mM sodium pyruvate, and 1× nonessential amino acids (NEAA) at 37°C with 5% CO₂ incubation. CEM.NK^R-CCR5 cells were maintained in RPMI 1640 medium (RPMI) supplemented with 10% FBS, 2 mM GlutaMAX (Life Technologies), 100 U/ml of penicillin (pen), 100 μg/ml of streptomycin (strep), 25 mM HEPES, and 1× NEAA (RPMI-10) at 37°C with 5% CO₂ incubation and were obtained through the NIH AIDS Reagent Program, Division of AIDS, NIAID (catalog no. 4376) deposited by Alexandra Trkola (59). CEM.NK^R-CCR5 cells were infected with SIV_mac239 at a multiplicity of infection (MOI) of 0.1 using ViroMag R/L magnetofection (OZ Biosciences). Infected CEM.NK^R-CCR5 cells were isolated from uninfected cells by sorting CD4-negative and CCR5-negative cells. The double negative CD4 and CCR5 cells were serially diluted to isolate single cell clones with a limiting dilution assay. One clone, 2C11, was found to produce large amounts of SIV Gag p27 as determined by Luminex assays and does not express CD4 or CCR5.

Transfections.

For plasmid transfection into HEK293T cells, cells were seeded onto 6-well plates (Corning) at a density of 7 × 10⁵/well 16 to 24 h prior to transfection. Cells were transfected with plasmids at 70 to 80% confluency using Transit-LT1 (Mirus) following the manufacturer’s recommended protocol. A total of 500 ng of a plasmid that carries the entire SIV_mac239 proviral genome (pMA239 [60]) and 2 μg of pX458 or pX461 plasmid were transfected per well. The supernatant was collected 24 h posttransfection and frozen at −30°C, and cells were harvested for flow cytometry analysis. 2C11 cells were nucleofected at a concentration of one million cells per 100 μl of 1SM buffer (61) (5 mM KCl, 15 mM MgCl₂, 120 mM NaH₂HPO₄/Na₂HPO₄ [pH 7.2], 25 mM mannitol, and 25 mM sodium succinate) containing 5.4 μg of pX458 or pX461 vector in an electroporation cuvette using program X-001 with the Nucleofector 2b device (Lonza). Cells were then cultured at 1 × 10⁶/ml, and twice a week for 15 to 17 days, cells were counted and all the supernatant was collected and frozen at −30°C.

Infection of CEM-x-174 cells using magnetofection.

CEM-x-174 cells were infected with supernatant collected from HEK293T cell cultures 24 h posttransfection using ViroMag R/L magnetofection (OZ Biosciences). Fifty microliters of supernatant and 3 μl of ViroMag particles were mixed and incubated on ice for 15 min. Per infection condition, 2 × 10⁵ CX-1 cells were pelleted via centrifugation at 300 × g for 5 min, and as much liquid as possible was removed by vacuum as to not resuspend the cell pellet. The virus preparation was gently pipetted on top of the CEM-x-174 cell pellet. Tubes were placed on a rack with an OZ Biosciences magnet underneath and were placed in a 37°C incubator for 2 h. Cells were resuspended in up to 1 ml of warmed RPMI-10 (complete with pen/strep) and transferred to a 24-well plate. The plate was incubated at 37°C, and twice a week for 3 weeks, half the culture was removed and replaced with RPMI-10 (complete with pen/strep). The removed culture was centrifuged at 10,000 × g to pellet the cells; the supernatant was transferred to a sterile tube and frozen at −30°C.

In vitro DNA cleavage assay.

For Cas9/single guiding RNA (sgRNA) ribonucleoprotein (RNP) formation, purified Cas9 nuclease fused with a nuclear localization signal (Cas9-NLS; catalog no. Cas9-NLS-50) was purchased from Eupheria Biotech; synthetic sgRNA with 2′-O-methyl 3′ phosphorothioate modifications of the first and last three nucleotides were purchased from Synthego and resuspended to 3 μM per the manufacturer’s recommended protocol (CRISPRevolution sgRNA EZ kit). The Cas9-NLS was complexed with the sgRNA for 10 min at room temperature (RT) in Opti-MEM I reduced serum medium (Gibco). A total of 500 ng of pMA239, 6 nM sgRNA, and 1 nM Cas9-NLS were incubated at 37°C for 1 h, and DNA fragments were resolved by agarose gel electrophoresis.

Cell sorting.

HEK293T cells were harvested 24 h after transfection of pMA239 and gRNA/Cas9-encoding plasmids followed by a wash with DMEM complete medium with 1% FBS. Cells were resuspended in 5 to 10 million cells per ml of DMEM complete medium with 1% FBS, and GFP-positive cells were sorted on a four-laser FACSAria III flow cytometer (BD Biosciences). GFP-positive HEK29T cells were then collected and DNA was extracted and purified using the Gentra Puregene cell kit (Qiagen). 2C11 cells were harvested 24 h after nucleofection of gRNA/Cas9-encoding plasmid(s), followed by a wash with RPMI complete medium with 1% FBS. Cells were resuspended in 5 to 10 million cells per ml of RPMI complete medium with 1% FBS, and GFP-positive cells were sorted on a four-laser FACSAria III flow cytometer (BD Biosciences). GFP-positive 2C11 cells were then washed in RPMI-10 and cultured for 15 to 17 days.

Flow cytometry.

HEK293T cells were harvested 24 h posttransfection via trypsinization followed by a wash with DMEM complete medium. Cells were then fixed in 1× BD FACS lysing solution (BD Biosciences), washed with 1× phosphate-buffered saline (PBS) plus 1% bovine serum albumin (BSA) (washing buffer), and permeabilized with 1× BD FACS permeabilizing solution 2 (BD Biosciences) as per the manufacturer’s recommended protocol. Cells were intracellularly stained by incubation with anti-SIV Gag p27 antibody (clone 55-2F12; NIH AIDS Reagent Program) conjugated with Dylight633 (Thermo Scientific; catalog no. 46414) at room temperature for 1 h. Cells were then washed twice with washing buffer and then fixed with 1× PBS containing 1.6% methanol-free formaldehyde (Polysciences) and analyzed on a three-laser CyAn Advanced Digital Processing (ADP) (Beckman-Coulter), and data were evaluated on FlowJo version 10 software. Nucleofected 2C11 cells were harvested day 15 or 17 of culture postsorting for surface and intracellular staining. Cells were washed with cold 1× PBS and resuspended in 100 μl of cold RPMI plus 10% FBS. Cells were surface stained with anti-CD4 peridinin chlorophyll protein (PerCP)-Cy5.5 (clone L200; BD Biosciences) antibody for 30 to 60 min at 4°C in the dark. Cells were then washed with washing buffer and intracellularly stained with anti-SIV Gag p27 Dylight633 (clone 55-2F12; NIH AIDS Reagent Program).

Luminex assays.

Virus production was examined by quantifying SIV Gag p27 (p27) from cell culture supernatants. Supernatants from HEK293T and 2C11 cell cultures were centrifuged at 10,000 × g for 5 min at room temperature and then lysed with a buffer containing 0.2% Tween 20 for 1 h at room temperature to disassociate SIV capsid protein from virus particles. Lysed supernatants were added to in-house-prepared polystyrene microspheres conjugated with anti-SIV Gag p27 antibody (clone 55-2F12; NIH AIDS Reagent Program) and incubated overnight at 4°C with shaking at 225 rpm. The next day, beads were washed and incubated with SIV-positive rhesus macaque serum diluted in buffer for 1 h at room temperature with shaking at 225 rpm, followed by two washes of beads and incubation with goat anti-human IgG-phycoerythrin (PE) antibody (Santa Cruz Biotechnology) for 1 h at room temperature with shaking at 225 rpm before reading. The levels of p27 from the supernatants were calculated by generating a standard curve from an SIV_mac239 viral stock of known concentration (62).

Amplicon sequencing and indel characterization.

In order to characterize indel creation at RGNu or RGNi cleavage sites, the target sites were PCR amplified using SIV-specific primers with Illumina overhang adapter sequences. Briefly, DNA was isolated from sorted HEK293T cells using the Gentra Puregene cell kit (Qiagen). An approximately 300-bp region consisting of the cleavage site(s) for each gRNA target was amplified using the primers listed in Table S1. PCR products were gel purified through a 0.8% agarose gel, and DNA was extracted using a gel extraction kit (Qiagen). Purified, extracted PCR products were given to the Greehey Children’s Cancer Research Institute’s Genome Sequencing Facility, where the core facility generated the second-round PCR products using the Nextera XT index kit set for barcoding. These barcoded PCR produces were pooled and run on an Illumina MiSeq (300 bp, pair end).

The raw paired-end sequence reads were first processed using BBDUK to remove adapter sequences and to discard low-quality sequences (ktrim=r tpe tbo k=23 mink=8 hdist=1 hdist2=1 ftm=5; qtrim=rl trimq=15 maq=20 k=31 hdist=1 minlen=50). The individual reads were then combined using FLASH v. 1.2.11 (63) with minimum and maximum overlaps of 40 and 300, respectively. The combined reads were then mapped on to the SIV_mac239 reference genome (GenBank accession no. M33262.1) using BWA MEM v. 0.7.17-r1188 (64), and custom Python scripts were used to identify unedited (i.e., perfect match) reads and to characterize introduced indels relative to the reference.

Western blot sample preparation.

CEM.NK^R-CCR5 cells, 2C11 cells, and 2C11 cells nucleofected with Nu3 or Ni6, sorted, and then cultured for 15 days were lysed in radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris-Cl [pH 7.4], 150 mM NaCl, 1% Triton X-100, 0.25% sodium deoxycholate, 1 mM phenylmethylsulfonyl fluoride [PMSF], 1× HALT protease inhibitor cocktail) at 3 × 10⁵ cells/100 μl for 1 h at room temperature. Cell lysates were diluted in 4× NuPAGE lithium dodecyl sulfate (LDS) sample buffer with additional RIPA buffer and boiled at 90°C for 10 min. Samples were added to a 4% to 12% Bis-Tris gel along with a SeeBlue Plus2 protein ladder and run with morpholinepropanesulfonic acid (MOPS) buffer at 170 V for 70 min. Proteins were transferred to a polyvinylidene difluoride (PVDF) membrane in transfer buffer loaded into for 1 h at 30 V. Membranes were blocked in 5% nonfat dry milk in PBS for 1 h at room temperature while rocking. For detecting SIV proteins, plasma from an SIV-infected rhesus macaque was diluted in antibody dilution buffer (2.5% nonfat dry milk in PBS). Membranes were incubated overnight with rocking at 4°C. The following day, the membranes were washed 3 times with PBS-Tween buffer (PBS-T); the washings were done for 5, 10, and then 15 min with rocking at room temperature. Goat anti-human IgG-horseradish peroxidase (HRP) antibody (Millipore; AP504P) was diluted 1:2,000 in PBS-T and incubated at RT for 1 h while rocking. For membranes detecting glyceraldehyde-3-phosphate dehydrogenase (GAPDH), anti-GAPDH antibody (Invitrogen; AM4300) was diluted 1:10,000 in antibody dilution buffer, and the membranes were incubated overnight with rocking at 4°C. The following day, the membranes were washed 3 times with PBS-T. Goat anti-mouse HRP antibody (Promega; W402B; 1 mg/ml) was diluted 1:10,000 in PBS-T and added to the membrane and incubated at RT for 1 h while rocking. The membranes were washed 3 times. Both anti-SIV and anti-GAPDH membranes were developed with SuperSignal West Pico CL (catalog no. 34080; ThermoFisher Scientific).

WGS.

In order to determine the frequency of off-target generation of indels by RGNu and RGNi, we used whole-genome sequencing (WGS) analysis of 2C11 cells nucleofected with our Nu3 and Ni6 plasmids to detect indels. Briefly, DNA was isolated from sorted and cultured cells using the Gentra Puregene cell kit (Qiagen). Genomic DNA quality and quantity were assessed by PicoGreen assay, and genomic DNA was fragmented by sonication. Bar-coded, paired-end, PCR-free libraries were prepared with the Nextera XT DNA library prep kit (Illumina). One hundred fifty-nucleotide paired-end WGS was performed up to 60-fold coverage with Illumina NovaSeq 6000. Sequences were aligned to human genome assembly GRCh38 (RefSeq accession no. GCF_000001405.39) and SIV_mac239 proviral DNA (GenBank accession no. M33262.1) using BWA mem (64). The proviral integration site was identified using DELLY v. 0.8.1 (42) as a “breakend” structural variation between SIVmac239 and human chromosome 19. Cas-OFFinder (45) was used to identify putative off-target sites using the parameters of 5 potential mismatches, 2 DNA bulges, and 5 RNA bulges between the protospacer and genomic DNA. GATK v. 4.1.0.0 was used to call indels following the best-practices guidelines (43). The CRISPResso2 v. 2.0.23 (44) tools CRISPRessoWGS and CRISPRessoCompare were used to compare the treated and untreated reads by applying a Fisher exact test cutoff of a P value of <0.05 to indels within 10 nucleotides of an NGG or NAG PAM site. To estimate the proportion of cells containing an integrated provirus, we generated a reference sequence containing the provirus plus 5,000 bp of flanking cellular DNA on both sides of the insertion site. Reads from each condition were mapped to this integration site reference sequence using the global read aligner Bowtie2 (65).

Statistical analysis.

Statistical analysis and figure generation were performed with GraphPad Prism 8 software. The statistical tests used for individual experiments are described in the figure legends; significance is defined as a P value of <0.05.