Combinatorial CRISPR-Cas9 and RNA Interference Attack on HIV-1 DNA and RNA Can Lead to Cross-Resistance

ABSTRACT

Many potent antiviral drugs have been developed against HIV-1, and their combined action is usually successful in achieving durable virus suppression in infected individuals. This success is based on two effects: additive or even synergistic virus inhibition and an increase in the genetic threshold for development of drug resistance. More recently, several genetic approaches have been developed to attack the HIV-1 genome in a gene therapy setting. We set out to test the combinatorial possibilities for a therapy based on the CRISPR-Cas9 and RNA interference (RNAi) mechanisms that attack the viral DNA and RNA, respectively. When two different sites in the HIV-1 genome were targeted, either with dual CRISPR-Cas9 antivirals or with a combination of CRISPR-Cas9 and RNAi antivirals, we observed additive inhibition, much like what was reported for antiviral drugs. However, when the same or overlapping viral sequence was attacked by the antivirals, rapid escape from a CRISPR-Cas9 antiviral, assisted by the error-prone nonhomologous end joining (NHEJ) DNA repair machinery, accelerated the development of cross-resistance to the other CRISPR-Cas9 or RNAi antiviral. Thus, genetic antiviral approaches can be combined, but overlap should be avoided.

INTRODUCTION

Over the last 25 years more than 30 potent anti-HIV drugs have been developed, including inhibitors of the viral enzymes reverse transcriptase (RT), protease, and integrase. Early clinical trials with individual drugs indicated potent virus suppression but also documented viral escape by the selection of one or a few point mutations in the viral gene encoding the targeted viral protein, e.g., zidovudine (AZT) resistance mutations in the RT enzyme (1). These mutations are introduced during low-level ongoing virus replication by the error-prone viral RT polymerase. Soon thereafter, clinicians observed the benefits of a combinatorial use of these drugs (2), which led to the formulation of highly effective therapeutic drug regimens consisting of three antivirals. Such drug combinations trigger more powerful virus suppression through additive inhibition, and sometimes even synergistic drug effects are observed (3). But perhaps more important is the impact of combination therapy on the ability of the virus to gain drug resistance. Resistance to drug combinations requires the acquisition of several mutations at multiple positions in the HIV-1 genome. Virus variants with a single mutation can easily be generated under therapy, but they may also be present within the viral quasispecies of an infected individual at the start of therapy (4, 5). However, virus variants with multiple specific drug resistance mutations are much more difficult to generate, which forms the basis for the clinical success of combination antiretroviral therapy.

Despite this success in long-term virus suppression, drug treatment does not lead to a cure, and lifelong therapy is required. That is why research has also focused on the development of gene therapy applications, which could potentially lead to durable virus suppression upon a single treatment. Many different genetic attacks on HIV-1 have been proposed, including transgenes that express transdominant negative viral proteins, antisense RNA, RNA aptamers, or RNA decoys (6–12). Two promising approaches that trigger a sequence-specific attack on the viral genome use the mechanisms of RNA interference (RNAi) and CRISPR-Cas9 (13). RNAi is a cellular mechanism for the regulation of gene expression at the posttranscriptional level via small, noncoding microRNAs (miRNAs). The complete RNAi machinery (e.g., Drosha, Dicer, and RNA-induced silencing complex [RISC]) is therefore present in cells to process these miRNAs and to execute the subsequent mRNA-silencing step. The cell can be equipped with a transgene encoding a short hairpin RNA (shRNA), which is processed into a small interfering RNA (siRNA) that functions as an miRNA mimic against the HIV-1 RNA genome. Potent HIV-1 suppression was reported but also rapid viral escape due to the selection of a point mutation within the short target sequence (14, 15). This result underscores the exquisite sequence specificity of RNAi action but also the ease of viral escape. Combinatorial RNAi regimens with at least three shRNAs could durably block viral escape (16, 17) and produced promising results in preclinical studies (18).

CRISPR-Cas9 is a more recent tool that was developed to cleave DNA genomes in a sequence-specific manner (19). Unlike RNAi, CRISPR-Cas9 systems are of bacterial origin and novel to human cells. This requires the transgene-mediated expression of the complete machinery consisting of the Cas9 endonuclease and the guide RNA (gRNA) that mediates the sequence specificity. Several groups of investigators have documented its potential for virus inhibition by cleavage of the integrated HIV-1 DNA genome (20–26). We along with others reported a novel viral escape mechanism that is actually facilitated by the cellular DNA repair machinery, more specifically the nucleotide insertions, deletions (indels), or substitutions introduced at the cleavage site by the nonhomologous end joining (NHEJ) mechanism (27–30). Cleavage-induced NHEJ repair triggers extremely rapid viral escape but only when genome regions are attacked that can absorb such changes without loss of virus replication capacity. The combination of two anti-HIV gRNAs boosted the antiviral effect, but viral escape (via NHEJ-introduced mutations) still occurred for most combinations (31, 32). Intriguingly, we identified two double-gRNA combinations that durably suppressed virus replication (31). These combinations did not result in viral escape and effectively sterilized the infected culture. Inspection of the leftover HIV-1 genomes revealed inactivating indels at both targets, thereby preventing virus replication. It has also been proposed that a double CRISPR-Cas9 attack on HIV-1 DNA can lead to genome excision (23–25). Although we also observed this excision route, our results indicate that the efficiency of dual cleavage and fragment deletion may be rather low.

We here set out to study the impact of combining CRISPR-Cas9 and RNAi antivirals. In particular, we investigated virus inhibition and escape when dual gRNAs or gRNA/shRNA combinations are used that target separate or overlapping viral sequences.

RESULTS AND DISCUSSION

Design of overlapping and nonoverlapping gRNA and shRNA reagents.

Figure 1A summarizes the mode of action of RNAi and CRISPR-Cas9 attack on HIV-1 RNA and DNA, respectively. RNAi attack calls for virus escape through regular RT-mediated point mutations (Fig. 1A, small diamond). CRISPR-Cas9-mediated DNA cleavage triggers fast NHEJ-assisted DNA repair that facilitates escape via introduced indels or nucleotide substitutions (Fig. 1A, large diamond). Repeated cleavage and repair, which will occur as long as the sequence is still recognized by the Cas9 nuclease, may lead to more severe mutations that will permanently inactivate the virus (Fig. 1A, multiple large diamonds).

FIG 1.

Targeting HIV-1 RNA and DNA with RNAi and CRISPR-Cas9. (A) In the RNAi pathway, the siRNA guides the RISC to the target sequence in the viral RNA. Binding of a 100% complementary sequence leads to cleavage and degradation of the RNA, thus inhibiting virus replication. Point mutations (small diamond) introduced in the viral target sequence during the error-prone reverse transcription process can prevent this RNAi-induced degradation and result in virus escape. In the CRISPR-Cas9 pathway, the gRNA directs the Cas9 nuclease to the target sequence in the viral DNA. Subsequent cleavage triggers NHEJ DNA repair, which frequently causes mutations (indels or nucleotide substitutions; large diamond) in the target DNA. These mutations may be compatible with virus replication (causing virus escape) or may permanently inactivate the virus (no escape). Repeated cleavage and repair may occur as long as the target sequence is recognized by the gRNA/Cas9 and may cause major mutations at the target site (multiple large diamonds), which can permanently inactivate the virus. (B) The target sites of the shNef and gRNAs in the HIV-1 RNA and DNA, respectively, are shown. The U3, R, and U5 domains of the LTR and the gag and nef genes are indicated in the viral DNA. Transcription starts at the first nucleotide of the 5′ LTR, and the transcripts are polyadenylated at the last nucleotide of the 3′ LTR. The gRNAs targeting the sense and antisense strand are indicated above and below the HIV-1 DNA, respectively. In the lower panel, the shNef target sequence in the HIV-1 RNA and the gRNA target sequences that overlap in the HIV-1 DNA are shown, with the PAM (underlined) and Cas9 cleavage sites (arrowhead) indicated.

We used various anti-HIV gRNA and shRNA reagents in this study (Table 1 and Fig. 1B). The gRNAs look at the viral DNA in the CRISPR-Cas9 mechanism. The shRNA targeting Nef (shNef) looks at the viral RNA in the RNAi mechanism. A dual CRISPR-Cas9 attack was analyzed without and with target sequence overlap (gRNA targeting Gag3 [gGag3] plus gGag1 and gGag3 plus gGag5, respectively). In the latter situation, the gRNAs recognize opposite strands of the double-stranded DNA (dsDNA) (Fig. 1B, lower panel). The shNef was combined with guide RNAs that target other HIV-1 loci (gGag1 and the gRNA targeting long terminal repeat 6 [gLTR6]; the latter actually cleaves HIV-1 DNA in both the 5′ and 3′ LTR) or an overlapping HIV-1 DNA locus (gNefL and gNefR, targeting opposing HIV-1 DNA strands).

TABLE 1.

Selected gRNA and shNef targets in HIV-1

Name	Position(s) in HIV-1 LAI DNA	Target sequence plus PAMa	Orientation	On-target activity (%)b	Conservationc
gLTR6	343–365, 9475–9497	GCTACAAGGGACTTTCCGCTGGG	Sense	88	0.52
gGag1	1389–1411	GTTAAAAGAGACCATCAATGAGG	Sense	64	0.15
gGag3	1480–1502	CCAAGGGGAAGTGACATAGCAGG	Antisense	67	0.07
gGag5	1480–1502	CCAAGGGGAAGTGACATAGCAGG	Sense	67	0.07
gNefL	8992–9014	CTACCAATGCTGCTTGTGCCTGG	Sense	74	0.42
gNefR	9010–9032	CCTGGCTAGAAGCACAAGAGGAG	Antisense	65	0.28
shNef	9007–9025	GTGCCTGGCTAGAAGCACAd			0.25

^aPAM sequence for gRNAs is indicated in bold.

^bThe on-target activity was predicted with the CRISPR design web tool (http://crispr.mit.edu/).

^cThe Shannon entropy measures the variation in the gRNA target sequence conservation among HIV-1 isolates (HIV 2014 sequence database [https://www.hiv.lanl.gov/]; only group M isolates with complete sequences were included). The entropy can vary from 0 to 1.5, with an invariant sequence having a score of 0.

^dThe actual target is RNA.

Inhibition of HIV-1 gene expression by gRNA and gRNA/shRNA combinations.

The impact of different gRNA/shRNA combinations versus the individual antivirals on HIV-1 gene expression was measured in a transient-transfection experiment. 293T cells were transfected with a plasmid encoding the HIV-1 LAI strain (pLAI) and plasmids expressing CRISPR-Cas9, gRNA, and/or shNef. Virus production was quantified by measuring the viral CA-p24 level in the culture supernatant at 2 days after transfection (Fig. 2). A high CA-p24 level was produced in the control experiments, i.e., upon pLAI transfection without any other plasmid (labeled 293T) and upon addition of the Cas9-expressing plasmid but without any gRNA construct. All tested gRNAs profoundly reduced HIV-1 gene expression when tested as individual inhibitors, and the tested combinations (nonoverlapping gGag3 plus gGag1 and overlapping gGag3 plus gGag5) yielded antiviral effects at least as strong. The shNef also potently reduced viral gene expression and a similar or further reduced virus level was apparent in combinations with a gRNA, either nonoverlapping (gLTR6 and gGag1) or overlapping (gNefL and gNefR). Additive effects are present for different combinations, but a more quantitative analysis is hampered by the potent antiviral effects of each individual inhibitor.

FIG 2.

RNAi and CRISPR-Cas9 inhibition of HIV-1 gene expression. The efficiency of shNef and the gRNAs to silence HIV-1 DNA was tested in 293T cells transfected with plasmids expressing HIV-1 LAI (pLAI), Cas9, single or dual gRNAs, and/or shNef. To quantify viral gene expression, the CA-p24 level was measured in the culture supernatant at 2 days after transfection. As control experiments, cells were transfected with only pLAI (293T bar; a control for shNef) or with pLAI and the Cas9 plasmid (Cas9; a control for all gRNAs). The CA-p24 levels were compared to the control values (set at 100%). Average values (± standard errors of the means) of four experiments are shown. The gGag1/gGag3 data were previously presented (31) and are included for comparison.

Inhibition of HIV-1 replication in a stably transduced T cell line and viral escape.

The SupT1 T cell line was stably transduced with different combinations of lentiviral and retroviral vectors expressing the Cas9 endonuclease, single or dual gRNAs, and/or shRNA-Nef. The transduced cells were infected with the CXCR4-using LAI isolate (1 ng of CA-p24), and virus replication was monitored by measuring the CA-p24 level in the culture supernatant. Efficient spreading of the virus in control Cas9-expressing SupT1 cells resulted in a rapid increase in the CA-p24 level (Fig. 3) and the formation of massive syncytia in the first week of infection. Coexpression of single or dual inhibitors blocked virus replication for a variable period. These cultures were maintained for up to 60 days to select for escape virus variants. As virus evolution is a chance process, we performed these experiments in quadruplicates. To quantify this virus escape moment, we plotted the average day at which HIV-1 breakthrough replication became apparent in the cultures (Fig. 4).

FIG 3.

HIV-1 replication in CRISPR-Cas9- and RNAi-protected T cells. Stably transduced SupT1 T cells expressing Cas9, single or dual gRNAs, and/or shNef were infected with HIV-1 LAI virus (corresponding to 1 ng of CA-p24) and cultured for 60 days. Virus replication was monitored by measuring the CA-p24 level in the culture supernatant. Each combination was tested four times, and all replication curves are shown. HIV-1 replicates with similar efficiency on untransduced control SupT1 and Cas9-only cells (data not shown).

FIG 4.

Time to HIV-1 escape from CRISPR-Cas9 and RNAi inhibition. Upon infection of the stably transduced SupT1 T cells expressing Cas9, single or dual gRNAs, and/or shNef with HIV-1 LAI virus (as shown in Fig. 3), the day at which massive virus-induced syncytia were observed (reflecting breakthrough virus replication) was scored. When no virus replication was detected, we scored 60 days. The average values for the four cell cultures (± standard errors of the means) are shown. SupT1, control nontransduced cells. The gGag1/gGag3 data were previously presented (31) and are included for comparison.

As previously shown (27, 31), gGag3 and gGag1 when applied individually inhibited HIV-1 replication for approximately 20 and 28 days, respectively. Sequencing of the breakthrough virus revealed the presence of nucleotide substitutions around the Cas9 cleavage site, which are typical for the Cas9 cleavage/NHEJ repair-mediated escape pathway (Fig. 5 and 6C) (27). The combination of these nonoverlapping gRNAs completely blocked replication for the time of the experiment (60 days). The gRNA combination thus significantly delayed viral escape, probably by increasing the genetic barrier for escape, similar to what has been described for drug combinations. The situation was quite different when gGag3 and gGag5, which target an overlapping sequence in HIV-1 DNA, were tested. gGag5 inhibited virus replication for a period similar to that with gGag3 in two cultures and completely blocked replication in the other two cultures (an average breakthrough replication day of 42). Combining this gRNA with gGag3 did not increase but decreased the period of inhibition as breakthrough replication was detected in all four cultures at approximately 20 days, which is similar to the period of inhibition obtained with gGag3 only. Sequencing of the breakthrough virus in the gGag5-only cell cultures identified mutations around the gGag5/Cas9 cleavage site (Fig. 5). Sequencing of the breakthrough virus in gGag3 plus gGag5 cultures revealed mutations around the gGag3/Cas9 cleavage site and not around the gGag5/Cas9 cleavage site. Thus, virus escape in these dual-gRNA-protected cell cultures resembles escape in cell cultures protected with gGag3-only treatment, both with respect to timing and to the position of the escape mutations, which indicates that the gGag3-induced mutations allow escape from both gGag3 and gGag5 inhibitors.

FIG 5.

Target site mutations in HIV-1 upon escape from single- and dual-gRNA CRISPR-Cas9 inhibition. The sequence of the gGag3/gGag5 target region was analyzed for the breakthrough viruses obtained in two to four independent HIV-1 cultures on SupT1-Cas9-gRNA cells expressing either only gGag3 or gGag5 or both gGag3 and gGag5 (indicated in yellow). At top, the wild-type HIV-1 nucleotide sequence (Gag gene; with the lines indicating the overlapping gRNA binding sequences and the arrowheads indicating the Cas9 cleavage sites) and translated amino acid sequence (Gag protein) are shown. Nucleotide and amino acid substitutions are indicated for every culture.

FIG 6.

Target site mutations in HIV-1 upon escape from RNAi and CRISPR-Cas9 inhibition. (A) The Nef region targeted by shNef, gNefL, and gNefR was analyzed for the breakthrough viruses obtained in independent HIV-1 cultures on SupT1 cells expressing either only shNef, gNefL, or gNefR or the combination of shNef with gLTR6, gGag1, gNefL, or gNefR. The latter two gRNAs target a sequence that overlaps the shNef target (indicated in yellow). (B) The gLTR6 target region was analyzed for the breakthrough viruses obtained in independent HIV-1 cultures on SupT1 cells expressing only gLTR6 or gLTR6 combined with shNef. (C) The gGag1 target region was analyzed for the breakthrough viruses obtained in independent HIV-1 cultures on SupT1 cells expressing only gGag1 or gGag1 combined with shNef. For panels A to C, the wild-type HIV-1 nucleotide sequence (with the lines indicating the gRNA binding sequences, the arrowheads indicating the Cas9 cleavage sites, and, for panel A only, the shNef target site boxed in gray) and translated amino acid sequence are shown at the top of the panels. Nucleotide and amino acid substitutions, insertions, and deletions (Δ) are indicated for every culture (*, translation stop codon).

Several possible explanations can be envisaged to explain the dominant character of gGag3- over gGag5-related mutations. It could be that gGag3 is a more efficient gRNA than gGag5 or that DNA repair at the former site occurs more swiftly, but such direct explanations seem unlikely. As we previously pointed out, restrictions in the HIV-1 escape options may indirectly but profoundly influence the outcome of virus evolution. Inspection of the genetic changes induced at the gGag3 and gGag5 cleavage sites reveal an interesting difference in the translated Gag protein sequence (Fig. 5, right-hand column). gGag3-induced mutagenesis yields 1- to 4-nucleotide (nt) changes that cause an amino acid substitution in this Gag domain, whereas only silent 1-nt codon changes are selected by gGag5 treatment. The latter observation suggests that amino acid substitutions at the Gag5 site are not compatible with virus replication and thus not selected. Both target sequences are highly conserved (Table 1), but more genetic variation seems to be allowed in the gGag3 target. We recently distinguished two types of HIV-1 target sequences: those that can absorb mutations and those that lose function upon acquisition of mutations. A CRISPR-Cas9 combination therapy with two gRNAs that exclusively focuses on the latter class was demonstrated to trigger mutational inactivation of the viral genome, thus resulting in an effective cure of the infected culture (31).

Basically, the same overall pattern was scored for the combined RNAi and CRISPR-Cas9 attack on HIV-1 RNA and DNA (Fig. 3 and 4). In the absence of genetic overlap (shNef plus gLTR6 and shNef plus gGag1), we observed regular additive inhibition for the combinations, resulting in a further delay or even prevention of the selection of escape viruses. Breakthrough virus replication was observed around days 10, 12, and 20 for the individual inhibitors shNef, gLTR6, and gGag1, respectively, but this drops to days 35 and 44 for the shNef plus gLTR6 and shNef plus gGag1 combinations, respectively.

All benefits of the combinatorial approach are lost in shRNA and gRNA pairs that overlap with respect to their genetic targets (shNef plus gNefL and shNef plus gNefR). These combinations did not delay breakthrough virus replication compared with replication in the shNef, gNefL, and gNefR cultures as breakthrough replication was detected in all cultures around day 10 (Fig. 3 and 4). The genotype analyses revealed a switch in the evolutionary path for the shNef target (Fig. 6). shNef monotherapy and combinations with nonoverlapping gRNAs (gLTR6 and gGag1) resulted in standard RT-introduced point mutations scattered across the target sequence in the nef gene. But in combination therapy with overlapping gRNAs (gNefL and gNefR), typical NHEJ-induced indels appeared around the gRNA/Cas9 cleavage site that were similar to the escape mutations observed in the gNefL-only and gNefR-only cultures. This result demonstrates that the NHEJ-assisted escape at the gNefL and gNefR target sequences in HIV-1 DNA also mediates shNef escape at the RNA level. Sequencing of the gLTR6 and gGag1 target sites of escape viruses in gLTR6-only and gGag1-only cultures, respectively, revealed indels and nucleotide substitutions around the Cas9 cleavage site, as previously described (27). A similar mutation pattern was observed when these gRNAs were combined with shNef.

To summarize, we demonstrate that CRISPR-Cas9 and RNAi antiviral approaches are similar in their abilities to reduce viral gene expression, with obvious variation in the efficiency of individual gRNAs and shRNAs. The two methods can also be combined for a concerted assault on both forms of HIV genetic information, DNA and RNA. However, one should be careful when overlapping HIV-1 sequences are targeted as cross-resistance can readily occur because CRISPR-Cas9 attack is coupled to a unique mutagenic response. CRISPR-Cas9-induced DNA cleavage triggers immediate NHEJ DNA repair, which generates considerable genetic variation at the site of cleavage (mostly indels but also nucleotide substitutions) that forms the raw material for the selection of escape virus variants. We here demonstrate that this evolutionary pathway can trigger rapid virus escape not only from CRISPR-Cas9 inhibition but also from other gRNA/CRISPR-Cas9 and shRNA/RNAi attacks targeting the same or overlapping sequences in the viral DNA and RNA, respectively.

Whereas we focused in this study on antiviral approaches that directly target the HIV-1 genome, it should be mentioned that one could also silence, using RNAi, CRISPR-Cas9, or other strategies, critical host factors. The major advantage of such indirect antivirals could be the difficulty in selecting resistant viruses (33). The major disadvantage is potential toxicity when a critical cellular protein is knocked down (RNAi) or even knocked out (CRISPR-Cas9). The CCR5 receptor is a logical candidate, and clinical trials that focus on this target are currently in preparation or ongoing (34, 35).

We describe the impact of therapeutic attack, different mutagenic pathways, and resistance development on genetically overlapping HIV-1 DNA and RNA target sequences. Such effects can even radiate onto the encoded viral proteins when the codons are altered. One can even set up nucleic acid selection scenarios that influence the evolution of certain protein mutants. For instance, we used RNAi strategies to skew the evolution toward certain drug-resistant HIV-1 variants that are less fit (36). Other strategies forced the selection of intersubtype recombinants under imposed RNAi pressure (37). With the massive use of CRISPR-Cas9 approaches against HIV-1 DNA, there will be options to design triple-molecule (DNA-RNA-protein) selection scenarios, with the eventual goal to produce a durable cure.

MATERIALS AND METHODS

The lentiviral vector LentiCas9-BLAST (52962; Addgene) containing the human codon-optimized Streptococcus pyogenes Cas9 expression cassette and LentiGuide-Puro (52963; Addgene) used for gGag3 expression were gifts from Feng Zhang (38). The lentiviral vector pLenti-pBsmBI-sgRNA-Hygro (62205; Addgene) was a gift from Rene Maehr (39) and was used for the expression of all other gRNAs. The pRETROSUPER-Puro retroviral vector containing the shNef expression cassette was described previously (14). The pLAI plasmid encoding the HIV-1 isolate LAI (GenBank accession K02013) was a gift from Keith Peden and obtained through the NIH AIDS Reagent Program, Division of AIDS, NIAID, NIH (courtesy of the Medical Research Council AIDS Directed Program) (40). All other reagents and procedures were described previously (14, 27, 31).