Introduction

Circular RNAs (circRNAs) are a family of non-coding RNAs that originate from a non-canonical splicing event, named backsplicing, in which the upstream and downstream splicing sites of an RNA transcript become covalently linked to form a closed loop of RNA as opposed to the typical linear structure of mRNA transcripts1. The lack of 5’ and 3’ termini make circRNAs highly resistant to degradation by exonucleases resulting in their hyperstability and extremely long half-life2. These characteristics make circRNAs excellent candidates for diagnostic biomarkers and therapeutic targets.

Cellular circRNAs play a role in development, cell differentiation, proliferation, response to oxidative stress and starvation and are associated with cancer3,4, neurological disorders5, autoimmune6, and cardiovascular diseases7,8. Although the precise mechanisms by which circRNAs exert physiological or pathological roles in disease remain poorly understood, multiple mechanisms have been proposed for their role in the regulation of gene expression and functions: (i) they act as miRNAs sponges by competitively binding to miRNAs associated with the Ago2 protein, thus inhibiting the miRNA’s functions1, (ii) they interact with RNA binding proteins (RBPs) modulating their availability and localization9,10, (iii) they modulate transcription of their host gene via interactions with RNAPolII and other components of the transcription machinery11,12,13 and lastly (iv) circRNAs containing an internal ribosome entry site (IRES) can be translated into short proteins, which are often found in metabolically active cells such as cancer cells8,14.

circRNAs of viral origin have been identified in DNA viruses from the Herpesviridae family (Kaposi Sarcoma (KSHV) and Epstein Barr (EBV))15,16,17 and in human papillomaviruses (HPV)18,19, although more have been speculated through analysis of available RNA sequencing datasets20,21. circRNAs derived from EBV and KSHV are upregulated in tumors and promote cell proliferation and metastasis by acting as sponges for cellular miRNAs with functions in tumor suppression22. A circRNA coded by HPV is translated into an oncoprotein that promotes cancer growth in vitro and in vivo18,19. Unsurprisingly, since RNA viruses are, with few exceptions, generated and processed within the cytoplasm, which lacks the components of the splicing machinery, only a few circRNAs encoded at extremely low levels by RNA viruses have been characterized20,23. An important exception is the human immunodeficiency type I virus (HIV-1); a retrovirus whose RNA genome is converted into DNA, integrated within the cellular genome, and transcribed by the cellular RNA Polymerase II into a single primary RNA, which is then processed like most cellular transcripts.

An array of alternative splice sites and several cis-regulatory sequences and trans-interacting factors modulate the complex HIV-1 alternative splicing pattern24,25, generating over 50 alternatively spliced mRNA isoforms26 (Fig. 1). The balanced expression of the unspliced and spliced transcripts is critical for viral replication, infectivity, and pathogenesis24,25. Our group and others have previously identified several RBPs which bind in proximity of the exon/intron boundaries of the viral pre-mRNA to regulate splicing and promote the recognition of weaker, alternative splice sites24,25,27. The backsplicing of cellular transcripts is often associated with weak, non-consensus splice sites, the presence of alternatively spliced exons, and RBPs binding to the exon/intron boundaries28,29,30. In HIV-1 the presence of multiple alternatively spliced exons, which are regulated by the binding of RBPs, and the alternative usage of multiple weak splice sites, suggests that multiple circRNAs might be generated by backsplicing of the primary viral transcript. We sought to determine if circRNAs were generated from the HIV-1 transcript. We identified 15 HIV circRNAs and showed that a subset of these circRNAs promote viral replication by functionally interacting with the cellular miRNAs miR-6727-3p and miR-4722-3p.

Fig. 1: Alternative splicing of the HIV-1 pre-mRNA transcript.
figure 1

The main isoforms generated by the splicing of the major acceptor and donor splice sites are indicated. The use of additional minor donor (D) and acceptor (A) splice sites (labeled a, b, c, d) yields over 50 distinct mRNAs. These minor mRNA isoforms represent <1% of the total amount of viral mRNAs. The 15 main circRNAs that are generated by backsplicing of the major donor and acceptor splice sites are indicated. Primer sets utilized to detect total (P-HIV_F/R), unspliced (P-GP_F/R), and multiple spliced mRNAs (P-MS_F/R) are indicated.

Full size image

Results

15 circRNAs are generated from backsplicing of the HIV-1 transcript

Analysis of the complex HIV-1 splicing pattern indicates that the virus has the potential to generate at least 15 distinct circRNAs by backsplicing of 3 major donor and 6 major acceptor splice sites (Fig. 1). Additionally, the usage of several suboptimal donor and acceptor splice sites, which are used with low frequency in the splicing of the linear viral mRNAs, could generate over 200 additional circRNAs. Although, the resulting circRNAs will have minimal sequence diversity given the relative proximity of the suboptimal and major splice sites.

We designed 7 sets of divergent PCR primers to amplify the predicted HIV circRNA isoforms (Fig. 2a). The divergent primers specifically amplify circRNAs but not linear RNAs since they extend in opposite directions on linear transcripts but extend towards each other on circularized transcripts. RNA extracted from the leukocyte-derived RevCEM-D4 cell line infected with HIV-1 (NL4-3 strain) was amplified utilizing the divergent primer sets. All the predicted HIV-derived circRNAs were amplified (Fig. 2b), although several PCR artifacts were also obtained by the amplification of rolling circle concatemers generated by the reverse transcriptase (Fig. 2c). The concatemers provide multiple binding sites within the same template for the primers, thus several PCR amplicons of varying size can be generated from each circRNA. Direct sequencing and cloning of the PCR amplicons confirmed the presence of all the predicted backsplicing junctions (BSJs) and the predicted circRNAs (Fig. 2d).

Fig. 2: HIV circRNAs are amplified by divergent PCR primers.
figure 2

a Schematic representation of the viral genome, and location of the exons that can be alternatively spliced into circRNAs. b 7 divergent primer sets were utilized to amplify the predicted HIV circRNAs in infected RevCEM-D4 lymphocytes. The donor and acceptor sites utilized in each BSJ are indicated. The expected amplicon size for each circRNA is also indicated. The alternatively spliced HPCAL1 gene was utilized as an amplification and alternative splicing control (HPCAL1 exon 3 is alternatively spliced generating two amplicons of 434 nt and 264 nt). c The 7 HIV circRNA PCR reactions were visualized on an agarose gel. (* red) indicates circRNAs amplicons. Several PCR artifacts are observed in most reactions due to reverse transcriptase artifacts that generate rolling circle concatemers which provide multiple primer binding sites generating multiple PCR amplicons and bands of intermediate mobility constituted by the hybridization of reverse complimentary strands from amplicons of diverse sizes that share ample sequence identity. d The results of the sequencing of the PCR amplicons and BSJs are summarized. Each BSJ can be present in multiple circRNAs and each primer pair can amplify multiple circRNA isoforms as indicated. The chromatogram of the sequence of each BSJ is shown (nucleotides −8 to +8).

Full size image

A panel of divergent qPCR primers can detect the individual HIV circRNAs

Since the combination of multiple major and minor splice sites generates 20 major linear RNA and 15 circRNA isoforms, with overlapping sequences and sizes, the identification of single circRNA isoforms with standard techniques such as northern blot and BaseScope is challenging. To overcome this problem, we utilized a qPCR-based strategy to identify each HIV circRNA. 15 primer sets were selected with one or both primers binding to a backsplicing junction and the others to canonical splice junctions allowing for amplification of specific isoforms (Fig. 3a). All the 15 predicted circRNA isoforms were detected in infected RevCEM-D4 cells utilizing the isoform-specific qPCR primers. RNA samples were treated with RNase R, a 3’ exoribonuclease which digests linear RNAs but not circRNAs, to confirm that each qPCR reaction specifically amplifies the target circRNA (Fig. 3b). HIV circRNAs, linear spliced mRNAs (HIV MS), total viral RNA (HIV All), and a cellular mRNA (RPL13A) were analyzed by qPCR. circRNAs were digested less efficiently by RNAse R than the control linear viral and cellular RNAs (Fig. 3a). The viral linear RNAs and cellular control RPL13A RNAs were degraded by over 97% following RNase R digestion while the different circRNAs exhibited a higher resistance to degradation (Fig. 3c), confirming the higher stability of the covalently-closed HIV circRNAs when compared to the linear messengers. Resistance to RNase R digestion varied greatly among the different circRNAs, in agreement with previous observations indicating that RNase R activity is dependent on the substrate RNA structure and size, with short RNAs (<200 nt) being less resistant to RNase R than longer ones31. In addition, divergent primers specific for the linear gag-pol transcript (CirGP) did not yield any amplification (Supplementary Source Data), confirming the specificity of the divergent qPCR primer sets for the amplification of circRNAs.

Fig. 3: Divergent qPCR primers selectively detect HIV circRNAs.
figure 3

a Divergent qPCR primer sets binding across either a BSJ or a canonical splice junction are utilized to amplify each HIV circRNA independently. b Total RNA was extracted from infected RevCEM-D4 cells and digested with RNase R or mock-digested as a control. Samples were amplified with the divergent circRNA qPCR primer sets and divergent primers (CirGP) for the linear Gag-Pol transcript as a negative control. Primers for the cellular gene RPL13A, all HIV-1 multiple spliced messengers (MS All, Fig.1) and all HIV-1 linear transcripts (HIV All, Fig. 1) were also utilized as controls. Data are represented as means ± SEM. P value (paired t-test) *<0.05, **<0.01, ***<0.001. No RT controls are shown in the Supplementary Source Data. c The relative percentage of each circular and linear RNA detected after RNase R digestion relative to the mock-digested sample is indicated.

Full size image

HIV circRNAs are detected in infected CD4+ T cells

Since circRNA expression is often tightly regulated in different tissues, we wanted to determine if the HIV circRNAs are expressed in primary leukocytes similarly to the RevCEM-D4 cell line. We analyzed the expression of HIV circRNAs in infected primary CD4+ T cells isolated from 6 donors (Fig. 4). Since expression of the viral transcripts varies in the infected population, the qPCR cycle threshold (Ct) value for each circRNA was normalized for the total amount of viral RNA present in each sample. Our data shows that the relative amount of each circRNA varied greatly among infected individuals. It is plausible that this variation might contribute to variations in viral replication, immune response and disease progression often observed in the infected population. Additionally, Ct values cannot be utilized to reliably quantify the absolute amount of each circRNA because of inherent differences in the amplification efficiency of different primer pairs, the results obtained suggest that circRNAs containing exon 2 and 3 either alone, or in combination, are highly expressed and in amounts comparable to the linear mRNAs containing the same exons (e.g., Cir23, 27.5 Ct vs linear exons 2-3, 26.4 Ct). It is plausible that the relative abundace of exon 2 and 3 in the viral circRNAs might be connected to their functions since exons 2 and 3 are noncoding and, although conserved in multiple isolates, no clear function has been assigned to their inclusion in the 5’ UTR of the mature linear viral mRNAs26,32,33.

Fig. 4: HIV circRNAs are detected in infected CD4+ T cells.
figure 4

a CD4+ T cells were purified from 6 healthy donors, activated with anti-CD3/CD28 antibodies and IL2 and infected with HIV-1 (NL4-3 clone). HIV circRNAs were detected utilizing the divergent qPCR primer sets. The Ct of each circRNA was normalized utilizing all HIV-1 linear transcripts (HIV All, Fig. 1) as normalizer. b Linear HIV-1 mRNAs were amplified utilizing isoform-specific qPCR primer pairs that detect the mRNAs containing either the spliced Exon 2, Exon 3, Exon 2 + 3, the unspliced Gag/Pol gene, all the multiple spliced transcripts (HIV MS), or all linear viral mRNAs (HIV All)24. No RT controls are shown in the Supplementary Source Data.

Full size image

Cellular miRNAs functionally interact with HIV circRNAs

circRNAs can bind and sequester miRNAs inhibiting their functions. We utilized the miRNA target prediction tool MirTarget to determine possible HIV circRNA–cellular miRNA interactions34. The results were filtered utilizing a target score ≥60 (a score ≥60 represents a high degree of confidence for the predicted miRNA–circRNA interaction) and a minimum free energy (MFE) ≤−10 kcal/mol35. We identified 11 candidate miRNAs predicted to bind the HIV circRNA sequences (Supplementary Information). 7 of the 11 miRNAs have functions in cellular proliferation and/or HIV-1 replication, suggesting that the HIV circRNAs might function as miRNA sponges to regulate the expression of genes required for viral replication and pathogenesis.

Since all miRNAs predicted to bind the HIV circRNAs, except for miR-5583-3p, were easily detectable by RT-qPCR (Ct <30) in both HEK293 and RevCEM-D4 cell lines (Fig. 5a), we utilized a luciferase reporter system to test the activity of the putative miRNA binding sites in HEK293 cells. 45 nt long fragments derived from the HIV circRNAs and centered around the predicted seed miRNA binding sequence for each miRNA were inserted within the 3’ UTR of the luciferase gene in the reporter minigene (pmiR-rep series, Fig. 5a), while a series of vectors containing the mutated seed recognition sequences were generated as controls. Since some of the miRNAs have identical predicted recognition sequences, two of the clones were predicted to bind multiple miRNAs (pmiR-6727-3p/miR-4722-3p and pmiR-18a-5p/miR-18b-5p/mir-4735-3p). Mutation of the seed binding sequence for miR-6727-3p/miR-4722-3p induced an increase in luciferase activity when compared to the wild-type insert indicating that this viral sequence functionally interacts with its cognate cellular miRNAs (Fig. 5c). Next, we sought to determine if miRNA-6727-3p and miRNA-4722-3p were functionally interacting with the target sequence by delivering synthetic miRNA mimics that simulate the endogenous miRNAs (Fig. 5d). While control miRNA mimics did not decrease the expression of the reporter construct, the miR-6727-3p mimic significantly downregulated expression of the luciferase reporter carrying the wild-type target sequence but had no effect on the plasmid carrying a mutation in the miR-6727-3p seed binding sequence. Delivery of the miR-4722-3p mimic caused only a marginal effect on the reporter expression; this might be due to a high ratio of target substrate to mimic and/or a lower affinity of the mimic for the target substrate. The extent of the predicted base pairing between miR-6727-3p /miRNA-4722-3p and the target circRNA sequences (Fig. 5e) is similar to what is observed in other viral and cellular miRNA-circRNA systems where a region with high complementarity to the miRNA seed sequence is flanked by a bulge and a secondary binding region with limited complementarity36,37,38,39. Taken together these data indicate that a subset of the HIV circRNAs (Supplementary Information; Cir-2T Cir-T Cir-23T Cir-3T Cir-R Cir-2R Cir-3R Cir-23R) can functionally interact with and possibly modulate the activity of miR-6727-3p and miR-4772-3p.

Fig. 5: Cellular miRNAs functionally interact with HIV circRNAs.
figure 5

a Predicted miRNA binding sequences within the HIV circRNAs are inserted in the 3’ UTR of the luciferase reporter minigene (pmiR-rep). 45nt inserts are centered around the seed binding sequence for each miRNA. The seed binding sequences are mutated in the control reporter constructs (pmiR-rep-M). b Expression of the miRNAs predicted to bind sequences within the HIV circRNAs in HEK293 and RevCEM-D4 cells. The Ct obtained by the amplification of each miRNA in both cell lines was normalized utilizing the U6 snRNA gene. c The wild type and seed mutated reporter minigenes were transfected in HEK293 cells and luciferase activity was quantified after 48 h. P value (independent t-test) *<0.05, **<0.01, ***<0.001. d miR-6727-3p and miR-4722-3p synthetic mimics were co-transfected in HEK293 cells with the pmiR-6727-3p/4722-3p reporter and the pmiR-6727-3p/4722-3p-Mut minigenes. All assays were carried out with a minimum of three biological replicas. Data are represented as means ± SEM. P value (paired t-test) *<0.05, **<0.01, ***<0.001. No RT controls are shown in Supplementary Source Data. e Base pairing of miR-6727-3p and miR-4722-3p to the target HIV circRNA sequence. The miRNA seed sequence is indicated in red.

Full size image

HIV circRNAs and their interacting miRNAs affect viral replication

miR-6727-3p has been found to be associated with congenital heart defects and is upregulated in gestational diabetes mellitus40,41, while miR-4722-3p has been shown to promote the progression of colorectal cancer by regulating the G-protein coupled receptor P2RY842, which regulates the activity of CREB, AP-1 and c-Myc; key transcription factors with known roles in T cell functions and HIV-1 replication43,44,45. The functional association of the HIV circRNAs with miR-6727-3p and miR-4772-3p implies that these miRNAs might play a role in viral replication in a step that follows integration and expression of the proviral genome.

The proviral clone pNL4-3 and miR-6727-3p, miR-4722-3p, or control miRNAs mimics were transfected in HEK293 cells to determine the role of these miRNAs in HIV-1 replication. In this system, the viral genome is expressed directly from the proviral molecular clone, mimicking the post-integration expression of the virus, thus limiting secondary effects due to different rates of infection and integration that can be observed utilizing infective virions and the primary cells targeted by the virus. Although HEK293 cells are not a leukocyte cell line, they express miR-6727-3p and miR-4722-3p (Fig. 5b), are routinely utilized to produce viral preparations24,46 and can be employed in HIV-1 gene expression studies since they regulate viral RNA biogenesis and replication similarly to leukocytes46,47. HEK293 cells present the advantage of being transfected by lipofection with high efficiency (>90%) and little physiological damage to the cell compared to electroporation; a method often utilized to transfect leukocytes. Quantification of the viral RNA present in the supernatant of the HIV-1 producing cells showed that expression of miR-6727-3p and miR-4722-3p correlates with a decrease in the viral RNA present in the supernatant (Fig. 6a). Accordingly, the infectious viral titer of the supernatant was reduced in cells transfected with the miR-6727-3p and miR-4722-3p mimics (Fig. 6a).

Fig. 6: miR-6727-3p and miR-4722-3p inhibit HIV-1 replication.
figure 6

a HEK293 cells were transfected with the molecular clone pNL4-3 and miRNA mimics. Viral RNA was quantified in the supernatant of the transfected cells 72 h post-transfection. The supernatant viral titer was determined by infecting the reporter cell line TZM-bl. P value (independent t-test) *<0.05. b The sequence of the circRNA CirR and Cir-M, carrying the mutation of the miR-6727-3p/miR-4722-3p seed sequence, were cloned in the RNA circularizing expression vectors pLac2-MCS and pMini-MCS. The ratio of circularized RNAs generated by the plasmids was determined utilizing the qPCR primer sets (p-Lac-RV/FW, p-Mini-RV/FW) that detect all plasmid transcripts (circular and linear) and compared to a primer set that detects only the circularized viral transcripts (p-Rev-RV/FW). The relative amount of the total and circular RNAs detected after RNase R digestion relative to the mock-digested sample is also indicated. Data obtained following expression of the negative control empty plasmids are shown in the Supplementary Source Data. P value (paired t-test) **<0.01. c HEK293 cells were co-transfected with the HIV-1 molecular clone pNL4-3, the indicated circularization vector, or the empty control vectors (pLac2-MCS, pMini-MCS). P value (one-way ANOVA) *<0.05, **<0.01, ***<0.001. d Biotinylated antisense RNA pull-down assay. Fold enrichment of miR-6727-3p and miR-4722-3p pulled down by the biotinylated CirR ASO, the control oligonucleotide, and the streptavidin magnetic beads alone in HEK293 cells transfected with the circularizing RNA plasmid pMini-CirR or pMini-CirR-M. All data are representative of a minimum of three biological replicas and are represented as means ± SEM. P value (independent t-test) *<0.05. No RT controls are shown in Supplementary Source Data.

Full size image

If circRNAs containing the miR-6727-3p/miR-4722-3p target sequence can sponge and inhibit these miRNAs, it is plausible that expression of these circRNAs might modulate viral replication. To test this hypothesis, we cloned the HIV CirR RNA sequence (Fig. 1) into two RNA circularizing vectors; one derived from the human ZSCAN gene and one from the Drosophila laccase2 gene (Fig. 6b). Both vectors contain a mini exon constituted by a multiple cloning site (MCS) flanked by two intronic sequences with high complementarity which base pairing brings the intervening splice sites into close proximity promoting backsplicing and the circularization of the sequences inserted within the MCS48. As expected, the backsplicing vectors generated both linear and circularized transcripts with the circularized messengers representing roughly 45–50% of the total RNAs transcribed from the vectors. A higher resistance to RNase R was observed for the RNA detected with the circRNA specific primers when compared to the primers detecting both circular and linear RNAs, confirming the synthesis of circRNAs from the circularizing vectors and the specificity of the primers utilized. Transfection of both CirR circularizing vectors upregulated virion production, which was monitored by measuring the viral RNA in the supernatant and the relative viral titer (Fig. 6c). Consistently with these results, the expression of empty vector controls and plasmids carrying mutations in the miR-6727-3p/4722-3p seed complementary sequence showed little to no change in virion production and infectivity. Moreover, the pulldown of a biotin-labeled CirR antisense oligonucleotide (ASO) could specifically recruit miR-6727-3p and miR-4722-3p while no enrichment was observed when a CirR circRNA carrying the miRNAs mutated seed binding region was expressed (Fig. 6d). Taken together, these data indicate that CirR upregulates viral replication by sponging miR-6727-3p and miR-4722-3p.

miR-6727-3p and miR-4722-3p are upregulated in infected T cells

Several miRNAs are upregulated in HIV-1 patients and in infection models49,50,51. We investigated changes in the expression of miR-6727-3p and miR-4722-3p in primary CD4+ T cells following HIV-1 infection. CD4+ T cells were isolated from healthy donors, activated by incubation with IL2 and anti-CD3/28 antibodies, infected with HIV-1, and cultured for 5 days to favor the homogeneous infection of the cells in culture. miR-6727-3p and miR-4722-3p expression significantly increased by a factor comparable or higher to what was observed in a series of miRNAs previously shown to be upregulated in response to HIV-1 infection49,50,51, while miR-218-5p, which expression is not altered following HIV-1 infection, was unchanged (Fig. 7a).

Fig. 7: miR-6727-3p and miR-4722-3p are upregulated in infected CD4+ T cells.
figure 7

a CD4+ T cells were isolated by negative selection from three healthy donors, activated and infected with HIV-1 (NL4-3) or mock-infected. Expression of miR-6727-3p, miR-4722-3p, a series of cellular miRNAs known to be upregulated following HIV-1 infection and the negative control miR-218-5p were analyzed by qPCR and normalized utilizing the U6 snRNA. The relative increase in the miRNA expression in the infected cells was calculated relative to their expression in the mock-infected cells. All data are representative of a minimum of three biological replicas and are represented as means ± SEM. The change in expression was compared to the negative control miR-218-5p, P value (independent t-test) *<0.05, **<0.01. No RT controls are shown in Supplementary Source Data. b Proposed model for HIV circRNA function in HIV-1 replication. HIV-1 infection promotes miR-6727-3p and miR-4722-3p expression which restricts viral replication. Processing of the viral transcript leads to the production of a series of circRNAs that can sponge miR-6727-3p and miR-4722-3p, inhibit their functions, and increase viral replication possibly via the P2RY8 dependent activation of CREB, AP-1, and c-Myc.

Full size image

Our data suggest a regulatory mechanism where HIV-1 modulates the mir-4722-3p/P2RY8/AP-1, CREB, c-Myc axis to increase viral replication and possibly modulate the T cell and immune response. miR-6727-3p and miR-4722-3p are upregulated shortly following viral infection and restrict viral replication. The integrated provirus counteracts this antiviral activity by producing a series of circRNAs that sponge and inhibit the miRNAs. Since miR-4722-3p downregulates the activity of the G-coupled receptor P2RY8, which activates several transcription factors, inhibition of miR-4722-3p might promote the activity of CREB, AP-1 and c-Myc42, thus modulating T cell proliferation, immune functions, and HIV-1 replication43,44,45 (Fig. 7b).

Discussion

Deep sequencing studies have identified thousands of circRNAs that are generated from eukaryotic genes and have multiple functions in cell replication, response to injury, inflammation, and disease3,4,6,7,8. However, only a few circRNAs coded from viral genomes have been identified. The best studied examples are circRNAs generated from the large genomes of DNA viruses of the Herpesviridae family15,16,17,18,19. We have now shown that HIV-1 generates several circRNAs by backsplicing of the viral transcript in both a T cell line and primary CD4+ T cells. This is the first experimental characterization of circRNAs being produced by an integrated retroviral genome. The relative abundance of the different HIV circRNAs varies greatly between individual donors, suggesting that their biogenesis is dependent on processing factors which may vary among single individuals and might in turn explain variations in viral spread, and possibly latency, often observed in patients. In this study, we have utilized the viral strain NL4-3, a subtype B chimeric virus, which is widely used in HIV splicing studies. Given the high variability of the viral sequences found in infected individuals, it is plausible that not all the circRNAs we have characterized are generated by viral strains found in patients. Nevertheless, alternative splicing studies carried out in subtype B transmitted/founder viruses have shown that the virus splicing patterns are highly conserved in different isolates26, suggesting that backsplicing events and circRNAs are also conserved. Further studies are warranted to characterize the circRNA species that are generated by viral isolates in the infected population.

Since HIV-1 replication and the cellular response are modulated by several cellular miRNAs52,53, we focused on putative interactions between the HIV circRNAs and cellular miRNAs. A targeted approach that utilized data mining and a series of validation assays indicated that a subset of HIV circRNAs can functionally interact with miR-6727-3p and miR-4722-3p, which are upregulated in response to HIV-1 infection and can restrict virion production in a post-integration model of viral replication. Although expression data from available databases (mirbase.org, mirgenedb.org) indicate that miR-6727 and miR-4722 are expressed at low level in most cell types and tissues, experimental data from multiple sources suggest that miRNAs derived from both miRNAs precursors play key functions in several physiological processes and diseases40,41,42, suggesting that inherent biases in RNA sequencing and data analysis might be limiting the number of reads assigned to these miRNAs in RNAseq assays. Given their high stability, circRNAs can sponge and downregulate the activity of miRNAs even if expressed in minimal amounts1 and their functions are often dependent on the relative amount of the miRNAs they sequester. Measurement of the correct copy numbers of circRNAs and sponged miRNAs is essential to study circRNAs functions. Unfortunately, standard qPCR is not well suited for the absolute quantification of circRNAs. Constraints in the choice of the primers sequence, which need to span at least a backsplicing junction (BSJ) or, in our case, a BSJ and a regular splicing junction, limit the optimization of the reaction. In addition, the reverse transcription step allows the reverse transcriptase to begin a rolling circle amplification that generates single-stranded cDNAs formed by a long chain of repetitive fragments that amplify the signal. The recent development of reverse transcription-droplet digital polymerase chain reaction has shown potential for the quantification of the absolute copy numbers of low abundance circRNAs without the need for a circular RNA quantification standard and might allow the precise quantification of HIV-1 cirRNAs in patients’ samples54.

One of the known miR-4722-3p targets, P2RY8, is a G coupled receptor that regulates the activity of c-Fos (AP-1), CREB and c-Myc; key transcription factors in the regulation of T cell and immune functions. AP-1, CREB and c-Myc have also been shown to modulate HIV-1 transcription in acute and latently infected cells43,55,56. Given their hyperstability, circRNAs are likely to accumulate in long-lived infected cells and might play a role in viral latency by regulating the activity of miRNAs with functions in the maintenance of the viral reservoir57,58. Studies that utilize latency models and possibly patient samples are needed to understand the role of the viral circRNAs in the establishment and/or maintenance of the latent viral reservoir. Our future work will be aimed at better defining the cellular targets of miR-6727-3p and miR-4722-3p and the role of the mir-4722-3p/P2RY8/AP-1, CREB, c-Myc axis in viral replication in relevant cellular models.

12 of the 15 HIV circRNAs we have characterized contain HIV-1 exon 2 or 3 or both. These are noncoding sequences included in the 5’ UTR of several viral messengers (Fig. 1) and although they are conserved in multiple isolates, they don’t have a defined role in HIV-1 replication, infectivity or pathogenesis26,32,33. Circ23, which contains exon 2 and 3, appears to be the most abundant among the viral circRNAs characterized. It is plausible that a key function of these untranslated exons is to form circRNAs with roles in HIV-1 replication/pathogenesis. Future studies will focus on the HIV circRNA species containing exon 2 and 3 or both, to characterize their role in viral replication/pathogenesis and determine the circRNA–miRNA–mRNA networks centered around these RNAs. In addition, we plan to utilize ASOs complementary to each one of the 8BSJs to target the viral circRNAs but not the linear spliced isoforms to define the role of each HIV circRNA in viral replication and cellular functions

Materials and methods

Plasmids

The pLac2-CirR and pMini-CirR circularizing plasmids were obtained by inserting the CirR sequence in the pcDNA3.1(+) Laccase2 MCS Exon Vector (a gift from Jeremy Wilusz, Addgene plasmid #69893; http://n2t.net/addgene:69893; RRID:Addgene_69893) and pcDNA3.1(+) CircRNA Mini MCS Vector (a gift from Jeremy Wilusz, Addgene plasmid #60648; http://n2t.net/addgene:60648; RRID:Addgene_60648). Complementary oligonucleotides were synthesized (Eurofins Genomics) (Supplementary Information), annealed, and cloned utilizing the SacII and AgeI restriction sites. pMIR-Report (Ambion) was utilized to insert the predicted miRNA binding sequences into the 3’ UTR of the luciferase gene utilizing the MluI and HindIII restriction sites. The inserts were obtained by annealing complementary oligonucleotides (Eurofins Genomics) (Supplementary Information).

Cell lines and cellular assays

HEK293 cells were obtained from the ATCC (CRL-1573). The human lymphoblast cell line RevCEM-D4 was obtained from the NIH HIV Reagent Program, Division of AIDS, NIAID, NIH: RevCEM-D4 Cells (ARP-13437, contributed by Dr. Alex Sigal). The TZM-bl cell line was obtained through the NIH HIV Reagent Program, Division of AIDS, NIAID, NIH (ARP-8129, contributed by Dr. John C. Kappes, Dr. Xiaoyun Wu and Tranzyme Inc). Cells were grown and maintained following the repositories’ guidelines. Cell transfections were carried out utilizing Lipofectamine 2000 (Thermo Fisher Scientific) in 24-well plates according to the manufacturer’s protocols utilizing 100 ng of plasmid DNA. miRVana miRNA mimics (Thermo Fisher Scientific) hsa-miR-6727-3p (cat# 4464066), hsa-miR-4722-3p (cat# 4464066), Negative Control #1 (cat# 4464058), and miR-1 Positive Control (cat# 4464062) were transfected at a final concentration of 50 pM.

The viral replication assay was carried out utilizing TZM-bl cells seeded 24 h before infection in 96 well plates at 50% confluence in 200 μL of D-MEM supplemented with 8% fetal calf serum and gentamicin. HEK293 cells were transfected with 50 ng of the proviral construct pNL4-3 and the indicated plasmid or mimic miRNA. Each transfection was carried out in a minimum of three biological replicas. Cells were washed twice with PBS 24 h post transfection to eliminate the transfected plasmids from the media. Supernatant was collected 72 h after the transfection and utilized to infect the TZM-bl cells. At 48 h post-infection, cells were lysed, and luciferase assays were carried out using the Illumination™ Lyophilized Firefly Luciferase Enhanced Assay Kit in technical triplicates according to the manufacturer’s protocol (Gold Biotechnology) and quantified utilizing a BMG PolarStar Omega reader and the MARS data analysis software. Data were represented as the means of three independent biological replica ± SEM. The Student’s t test (2 groups) or ANOVA (3 or more groups) were performed to compare differences between groups. A P < 0.05 was considered statistically significant.

Primary T cells isolation, activation, and infection

Blood apheresis samples from de-identified healthy individuals were obtained from OneBlood (www.oneblood.org) and classified as non-human research by the Florida Atlantic University Health Sciences IRB. PBMCs were isolated utilizing Ficoll®-Paque Plus (MilliporeSigma) and SepMate™-50 (STEMCELL) tubes following the manufacturers’ protocols. The isolation of naïve CD4+ T cells from PBMCs was performed using the negative selection EasySep™ Human CD4+ T Cell Isolation Kit (STEMCELL Technologies) according to the manufacturer’s procedure. T cells were activated for 120 h prior to infection by incubating 2.5 × 106 cells in 1 mL of RPMI 1640 medium supplemented with 10% heat-inactivated fetal bovine serum in 24-well plates coated with anti-CD3 antibody (BioXCell). Anti-CD28 antibody (BD bioscience) at a final concentration of 6 μg/mL and 20 units of interleukin-2 (PeproTech) were added to each well. Infections were carried out in 1 mL of RPMI 1640 supplemented with 10% heat-inactivated FBS, 20 units/mL of interleukin-2 and 6 μg/mL of polybrene with a HIV-1 preparation at a high multiplicity of infection (MOI = 1), which was calculated by dividing the viral titer by the number of cells. HIV-1 preparations were obtained by transfecting HEK293 cells with the molecular clone pNL4-3, washing the cells twice 24 h post-transfection to eliminate residual plasmid DNA, and harvesting the supernatant 72 h post-transfection. The viral titer was determined utilizing the TZM-bl reporter cell line as previously described59. T cells were mock-infected as a control utilizing the supernatant from mock-transfected HEK293 cells. Cells were washed twice with PBS 24 h post-infection to eliminate residual virions and incubated for 5 days following infection.

circRNA isolation and quantification

Total RNA was extracted utilizing a TRIzol (Thermo Fisher Scientific) based protocol we have optimized for the extraction of viral RNA60 and DNase treated with Turbo DNase (Thermo-Fischer Scientific). RNA was reverse transcribed utilizing a random pd(N)6 primer and Superscript II RT (Thermo Fischer Scientific). RT-PCR assays for the amplification of the circRNAs were carried out utilizing the 3 G HotStart Taq DNA Polymerase (BioBasic). We utilized a slow renaturation PCR step to minimize the formation of hybrid PCR products between alternatively spliced circRNA isoforms as previously described24. Quantitative PCR analysis of the linear viral transcripts was carried out as previously described24 utilizing the primers shown in Supplementary Information. qPCR data were normalized for the relative expression of the housekeeping gene RPL13A or the total HIV-1 RNA where indicated61. Total HIV-1 RNA was quantified utilizing primers P-HIV-F and P-HIV-R which amplify a region of the viral 3’UTR that is present in all the viral transcripts. The absolute number of viral RNA copies/mL was obtained by interpolating the Ct obtained by the amplification of the supernatant sample within a pNL4-3 standard qPCR curve. qPCRs were performed utilizing the Agilent AriaMx real time PCR system, the Green-2-Go SYBR green qPCR Kit (BioBasic), and analyzed utilizing the AriaMx software. miRNA RT-qPCR assays were carried out utilizing the Mir-X™ miRNA First-Strand Synthesis and miRNA qRT-PCR TB Green® Kits (Takara Bio) and quantified utilizing the small RNA U6 as a normalizer control. miRNA quantification of supernatant samples was carried out by normalizing the samples utilizing miRNA miR-16-5p, which is present in exosomes secreted from several cell types to account for differences in cell number and RNA extraction efficiency. The reaction conditions for all qPCR assays were 95 °C for 10 min, followed by 40 cycles of 95 °C for 5 s and 60 °C for 20 s. qPCR assays were carried out in technical duplicates and are representative of a minimum of three independent biological replicas. Data are represented as means ± SEM. The Student’s t-test (2 groups) or ANOVA (3 or more groups) were performed to compare differences between groups. A P < 0.05 was considered statistically significant. The minimum difference between a sample and NTC (no template) or NRT (no reverse transcriptase) was kept to 5 Ct to minimize the amplification of contaminant DNA. The qPCR denaturation profiles for all the primer pairs utilized in this study showed single distinct denaturation peaks. DNA sequencing was performed by excising the desired band from the agarose gel and using the Eurofins Genomics pre-mixed tube sequencing service. Short amplicons, which are difficult to direct sequence, were cloned and sequenced. RNAse R digestion was carried out by treating 10 μg of total RNA with 0.5 μl of RNAse R (Lucigen) at 37 °C for 15 min. A mock-digested RNA sample containing the RNase R reaction buffer and incubated at 37 °C at the same time was used as a negative control.

Biotinylated RNA pull-down assay

Pull-down assay were carried out as previously described62. The biotinylated CirR-AS-BYT ASO was designed to bind to the CirR sequence while the oligo CirR-M-AS-BYT was used as control probe. 3.5 × 106 HEK293 cells were transfected 48 h prior to the pull-down assay with the circularizing RNA plasmid pMiniRev31 and pMiniRev31-M as control. Cells were lysed to isolate the cytoplasmic fraction. The lysate was split in three and incubated with 100 pmol of either the biotinylated CirR ASO, control oligo, or a mock control at 4 °C for 90 min. The biotin-coupled RNA complex was then incubated with streptavidin magnetic beads (Life Technologies) for 2 h at 4 °C. The beads were washed three times and the miRNAs bound to the circRNA-ASO complex were isolated using Trizol and analyzed by qRT-PCR. CirR circRNAs were amplified utilizing the primer set LacRev31-FWa/qPCR-LacRev31-RVa to confirm the specific pull down of the circularized cirR RNA (Supplementary Source Data) and the miRNA qRT-PCR TB Green® Kits (Takara Bio) with the miRNAs specific primers Pr_miR-6727-3p and Pr_miR-4722-3p.