Introduction on the RNAi Mechanism

Noncoding RNA plays an important role in the regulation of cellular gene expression, most notably via the RNA interference (RNAi) mechanism, which is evolutionary conserved among eukaryotes. RNAi triggers the sequence-specific inactivation of one or multiple complementary mRNAs. At least three small RNA classes can be distinguished that participate in RNAi mechanisms: microRNAs (miRNAs) [1, 2], endogenous small interfering RNAs (endo-siRNAs) [3], and PIWI-associated RNAs (piRNAs) [4]. The latter two classes are mainly implicated in the suppression of transposons. The miRNAs are involved in the regulated expression of many cellular genes at the posttranscriptional level [5, 6]. miRNA-mediated gene regulation plays an important role in cell metabolism, cellular developmental and differentiation processes in mammals.

More than 1,000 human miRNAs have been identified, which are involved in the regulated expression of at least 30 % of human genes [7]. Many efforts have been made to elucidate the biological function of these miRNAs by identification of their target mRNAs. This miRNA–mRNA matching remains a formidable task in the absence of good algorithms for target site prediction and is complicated by the fact that many mRNAs may be influenced by a single miRNA. The design of man-made miRNA mimics like short hairpin RNAs (shRNAs) provided new tools for the control of introduced transgenes in therapeutic applications. The development of vector-mediated RNAi allowed the establishment of durable gene silencing approaches, in particular for retroviral and lentiviral vector systems that are stably inherited by integration into the host cell genome [8–10]. We describe our research line on the development of an RNAi-based gene therapy against HIV-1, which was initiated around 2003. The scientific discoveries and practical hurdles that we came across along this translational track towards the clinic are described.

RNAi: From Natural Mechanism to Therapeutic Approach

The endogenous pathway for miRNA biosynthesis is widely conserved among vertebrates and invertebrates. First, a primary miRNA transcript (pri-miRNA) is made, of which a hairpin-like RNA structure is processed by the “Microprocessor” complex, which consists of the Drosha nuclease and its dsRNA-binding partner DGCR8 (Fig. 1, endogenous RNAi pathway). Microprocessor recognizes pairing in the stem and multiple primary sequence elements in the single-stranded flanks and loop region [11, 12]. The resulting pre-miRNA is cleaved near the terminal loop by the Dicer nuclease in collaboration with the trans-activation response RNA-binding protein (TRBP) and protein activator of PKR (PACT) cofactors [13]. This miRNA pathway yields the mature RNA duplex, of which one strand of approximately 22-nucleotides is preferentially loaded into the Argonaute (Ago) enzyme as part of the RNA-induced silencing complex (RISC). The miRNA-loaded RISC complex targets mRNA transcripts, usually with multiple partially complementary targets, for translational repression. In exception cases there is full complementary of the miRNA with the mRNA causing degradation of the latter.

Fig. 1
figure 1

The RNAi pathway. The endogenous RNAi pathway of mammalian cells is plotted on the left hand side. Host pri-miRNAs are transcribed in the nucleus, processed by Drosha and DGCR8 into a pre-miRNA and transported by Exportin-5 to the cytoplasm. The pre-miRNA is processed by the Dicer/TRBP/PACT endonuclease complex to yield the miRNA duplex. One strand (the mature miRNA) is loaded into the Ago2 enzyme, thus forming the RNA-induced silencing complex (RISC) that induces gene silencing by translational repression or cleavage of the mRNA target. The exogenous RNAi cascade is shown on the right hand side. Man-made shRNA gene cassettes are expressed in the nucleus and enter the RNAi pathway at the level of Exportin-5. Synthetic siRNA is transfected into the cell and will directly instruct RISC for mRNA silencing. The man-made shRNAs and siRNAs direct mRNA cleavage by perfect base pairing complementarity

Full size image

Short hairpin RNAs (shRNAs) are perfectly base paired minimal miRNA mimics that are synthesized from man-made gene cassettes. These hairpins enter halfway the RNAi pathway as Drosha cleavage is not needed to remove flanking sequences (Fig. 1, exogenous RNAi inducer). The loop of the hairpin is cleaved off by Dicer to produce the small interfering RNA (siRNA) with an active guide strand and an inactive passenger strand. Synthetic siRNAs and shRNA genes are usually designed with full base pairing complementarity with the intended target mRNA. Only a single target sequence is needed to trigger sequence-specific cleavage and inactivation of the mRNA. This specificity is the greatest asset for turning the RNAi mechanism into a therapeutic mode, but it also creates problems when dealing with genetically flexible microorganisms as escape mutations may frustrate the therapy. In general, the idea to develop an RNAi-based antiviral therapy is supported by the recent finding that RNAi represents an antiviral defense mechanism in mammalian cells [14, 15].

HIV-1 as RNAi Target

We previously reviewed the ins and outs of using the RNAi machinery for a specific and durable attack on HIV-1 [16, 17]. We list the most significant points to consider. A relevant question is whether one can target the “incoming” viral RNA genome that is introduced into the cell by virus infection (Fig. 2a, route a). We demonstrated that an early attack on the incoming RNA is impaired because the RNA genome is protected by the viral nucleocapsid protein as part of the virion core structure [18]. Alternatively, transcriptional gene silencing may be induced at the viral LTR promoter much later in the replication cycle after the integrated DNA provirus is established [19, 20] (Fig. 2a, route b). Most likely, the newly produced viral transcripts will be targeted (Fig. 2a, route c).

Fig. 2
figure 2

Where and how to attack HIV-1 with RNAi-based inhibitors? (a) The HIV-1 replication cycle and three RNAi attack possibilities. Upon virus entry into the cell, the viral core with the RNA genome is released into the cytoplasm. This incoming RNA genome represents an attractive target (attack a). HIV-1 RNA is reverse-transcribed and the DNA is integrated into the host cell chromosome. New viral mRNAs are subsequently transcribed from the integrated provirus. Viral mRNA production in the nucleus can be blocked by transcriptional silencing (attack b). RISC-mediated silencing can occur in the cytoplasm (attack c). Attack route c seems the major mechanism of HIV-1 inhibition; see the text for further details. (b) The HIV-1 DNA genome map with some shRNA targets. The shRNAs Gag5, Pol1, Pol47, and R/T5 target conserved and essential HIV-1 sequences. The Gag5 shRNA was subsequently removed from the combinatorial RNAi cocktail because of toxicity (see Figs. 8 and 9). The shRNA Nef was used for viral escape studies (see Fig. 3). (c) Alternative splicing of the HIV-1 primary transcript is illustrated. The shaded terminal boxes mark genome segments (mostly noncoding) that are present in all mRNA forms and thus form ideal shRNA targets

Full size image

It is important to screen many shRNA candidates for their potential to knock down HIV-1 gene expression, but some general rules can be applied. First, it is important to target conserved HIV-1 sequences such that many different isolates and subtypes are sensitive to the therapy (Fig. 2b). Second, but related to the first rule, one should target sequences that are essential for HIV-1, most notably sequences that encode critical domains of the essential viral proteins. These sequences are not only well conserved (rule 1), but will also avoid mutations that are needed to escape from specific RNAi pressure [21]. Third, it is important to consider the complex HIV-1 splicing pattern that generates many different mRNA targets [22]. The best target sequences may be located in the extreme 5′ and 3′ ends of HIV-1 RNA, which are present on each mRNA splice variant (Fig. 2c, shaded areas). The first three rules are not mutually exclusive as some of the 5′ and 3′ end sequences, particularly those overlapping with the regulatory long-terminal repeat (LTR) and critical RNA signals, are fairly well conserved, although not coding for protein [23]. Fourth, it seems important to target sequences that are accessible in the HIV-1 RNA structure for the shRNA-programmed RISC complex [24, 25]. The importance of this rule was underlined by the description of an RNAi-escape variant with an altered local RNA conformation [26]. Fifth, one could argue that it is favorable to interrupt the replication cycle by targeting the early viral functions like the Tat, Rev and Nef proteins that are translated from the early wave of multiply spliced transcripts (Fig. 2c). Mathematical modelling predicted that HIV-1 decay dynamics depend on the stage of the viral replication cycle that is attacked, much more so than the actual drug efficacy [27]. Tat may be a particularly good target because this protein has been proposed to exhibit RNAi suppressor activity [28, 29].

When talking about early and late protein function, we should realize that the actual target for RNAi attack is the RNA molecule. For instance, sequences involving the 3′ end of the Nef gene are present on all HIV-1 transcripts, thus forming an ideal target according to the third rule (Fig. 2c). However, the first selection rule remains of overwhelming importance as some of the Nef sequences are not well conserved among HIV-1 isolates and different subtypes. In addition, the Nef gene is not an essential part of the HIV-1 genome and can even be inactivated by deletions under specific RNAi pressure (see “HIV-1 Escape”).

Despite all these rules for the selection of candidate target sites, it remains important to construct and test the activity of a significant number of distinct shRNAs as only a few will have potent activity [30]. Perhaps we will be able to raise the overall success rate in the future by improved shRNA design algorithms. Anyhow, a large screen will allow one to identify potent inhibitors. Simple co-transfection experiments can be performed to score the silencing efficiency on matching luciferase-HIV reporter constructs. As this may require the comparison between different reporters, each with a different piece of the HIV-1 genome attached, the alternative test system becomes more attractive. Co-transfection of the shRNA construct with a full-length HIV-1 molecular clone can be performed, with silencing scored as a reduction in virus production, which can be measured by different means (e.g., CA-p24 Elisa or Reverse Transcriptase activity). The antiviral potency should subsequently be tested in stably lentivirus-transduced T cells against a spreading HIV-1 infection, preferentially using diverse virus isolates. Subsequent tests will include preclinical studies in vitro and in vivo, e.g., in the humanized mouse model, to confirm the antiviral efficacy and to test for adverse effects (see “Preclinical Efficacy Tests,” “Preclinical Safety Tests,” “The Humanized Mouse Model”).

HIV-1 Escape

Prolonged in vitro HIV-1 replication studies revealed how this virus can escape from the pressure of an antiviral shRNA. We reported an detailed escape study for the shNef inhibitor that targets sequences encoding the Nef gene [31]. As said, this may be a good position for shRNA attack because these sequences are present in all HIV-1 RNAs, full-length or spliced. On the other hand, Nef is not an essential viral function and thus may facilitate “easy” escape. Although robust and reproducible virus suppression was obtained in a T cell line that was transfected with the lentiviral vector encoding shNef, viral escape was eventually observed. Multiple parallel HIV-1 escape experiments with a single shRNA inhibitor are shown in Fig. 3a. To confirm the escape phenotype, the virus was collected at the peak of infection and used to infect a fresh batch of shNef-expressing cells, causing a rapid spreading infection (Fig. 3b).

Fig. 3
figure 3

HIV-1 escape options. (a) Long-term shRNA inhibition study with replication-competent HIV-1. T cells expressing an individual shRNA were infected with HIV-1 and virus replication monitored by measuring CA-p24 production. Control cells with the empty lentiviral vector JS1 show rapid virus spread as measured by CA-p24 ELISA in the culture supernatant. Robust HIV-1 inhibition is observed in all shRNA cultures, but breakthrough replication is apparent at variable times. (b) The virus-containing supernatant from an escape culture was transferred to fresh T cells, demonstrating immediate spread on shRNA and control JS1 cells, demonstrating phenotypic viral escape. (c) An example of HIV-1 escape mutations under pressure of the shNef inhibitor. The shNef target sequence is highlighted in green. Escape is possible by a single or multiple point mutations in the target sequence, partial or complete deletion of the target or a mutation outside the target. (d) Rearrangement of the local RNA structure due to a point mutation causes RNAi-resistance by occlusion of the 3′ end of the target sequence, which forms the annealing site for the shRNA inhibitor. See the text for further details. Panel C was modified from [26]

Full size image

The next step is to determine the sequence changes that underlie the escape phenotype. For shNef, we sequenced the target site and flanking sequences of these escape viruses and documented a wide range of escape options (Fig. 3c). Point mutations were selected in the target site of some cultures (C1, F, I), demonstrating the exquisite sequence specificity of RNAi action. Over time the C1 culture evolved a second point mutation (C2), indicating that the single change does not provide full RNAi-resistance. In other cultures, HIV-1 acquired a deletion in the target site, either removing it partially (B, D, E) or completely (A, G). This result confirms the absence of a replicative function for the Nef protein in this in vitro culture system. As predicted, such deletion-mediated escape was not observed when more critical domains of the HIV-1 RNA genome were targeted [30]. In fact, when important protein domains were targeted, e.g., in the Protease or Integrase enzymes, we observed a preference for mutations that represent “silent” codon changes [32]. These results indicate that HIV-1 prefers not to change the encoded protein function, and RNAi-escape variants do in fact mimic the sequences, present in natural HIV-1 variants [21].

Culture H is unique and remained unexplained as there was no mutation inside the target site. Follow-up studies demonstrated that the point mutation that is located seven nucleotides upstream of the target site does cause RNAi-resistance by a different mechanism. This mutation destabilizes a local hairpin structure in the HIV-1 RNA genome such that an alternative folding is induced (Fig. 3d). Annealing of the shRNA inhibitor is initiated at the 3′ end of the target site, which is accessible in the original structure, but occluded in the induced structure. Several other studies highlighted the importance of target RNA structure on the efficiency of RNAi attack [25, 26, 33–36]. Knowledge on the actual structure of the HIV-1 RNA genome can thus be used for the selection of improved shRNA reagents [24, 37].

Targeting Cellular Cofactors

If targeting of the virus causes so much escape problems, targeting of a host cell cofactor may represent a better antiviral option. The number of candidate cellular cofactors has increased considerably based on RNAi knockdown screens [38–40]. As these candidate cofactors were obtained in transient assays with reporter genes in non-T cells, which is remote from the physiological setting, they first need to be confirmed in regular HIV-1 infection experiments. Some candidates were subsequently knocked down to test for the antiviral activity in T cells [41]. Targeting of cellular cofactors imposes specific advantages and shortcomings. It may obviously cause cytotoxicity, but may have a dual advantage concerning viral escape. Inhibition of an important cofactor will be effective against all viral variants in an infected individual and likely all HIV-1 strains and subtypes that circulate worldwide. Additionally, viral escape would seem possible only through adaptation to an alternative cellular cofactor. Thus, it is important to target components of cellular pathways that lack redundancy [41].

As discussed, RNAi does not allow an early attack on the RNA genome of the infecting virus particle [18]. However, one could target cellular entry factors that facilitate virus–cell contact and entry. The chemokine receptor 5 (CCR5) is the most important HIV-1 receptor and represents a promising target because this protein is not important for human physiology as demonstrated by individuals with a homozygous gene deletion that interrupts CCR5 protein expression [42]. A proof of concept for this concept was obtained by bone marrow transplantation from such a CCR5-minus donor in the “Berlin” HIV-1 patient who subsequently did not need antiviral drugs to maintain an undetectable viral load [43, 44]. This functional cure has spurred a search for alternative cofactors that are vital for HIV-1 replication, yet without an important role in human physiology. It has even been proposed to remove the CCR5 gene with a tailored nuclease that excises the CCR5 gene [45, 46]. When silencing or removing the CCR5 function, a possibility is that HIV-1 escapes by switching to CXCR4 as alternative receptor and such CXCR4-using HIV-1 variants may be more pathogenic [47].

Combinatorial RNAi Approaches

Promising anti-escape approaches include targeting of highly conserved and evolutionary restrained regions of the viral RNA genome, but HIV-1 is still likely to escape from a potent shRNA by selecting a mutation in the target sequence [48]. In case only a few HIV-1 escape routes are observed, it can be proposed to develop modified shRNAs that specifically target these escape variants. These modified shRNAs should be combined with the original inhibitor to prevent viral escape. We used this approach to skew the evolution of resistance, but it became apparent that virus evolution could not be blocked completely as HIV-1 started using new escape routes.

It makes sense that combinatorial RNAi approaches should mimic the action of combinatorial drug regimens that are very successful in the durable control of HIV-1 in patients [30, 49]. In other words, one should simultaneously express multiple shRNA inhibitors that target different parts of the HIV-1 genome or important cofactors [30, 50]. Besides additive inhibition, one will raise the genetic threshold for the development of resistant viruses as multiple mutational hits will be required in multiple target sites.

Combinatorial RNAi can be achieved with multiple shRNA cassettes introduced into the same lentiviral vector [50], but alternative scenarios have been tried with variable success [51, 52]. The different methods are listed in Fig. 4 with some of the major advantages and disadvantages. Multiple inhibitors can be generated from polycistronic miRNA transcripts [53, 54]. Stacking of two shRNAs on top of each other leads to the so called extended shRNA (e-shRNA) design, but most silencing activity is lost upon further extension as in long hairpin RNAs (lhRNAs) [51, 55–57]. The RNAi inhibitors can be combined with other RNA-based inhibitors [49]. One could even create hybrid molecules that combine siRNA and other antiviral activities, e.g., an RNA aptamer that binds to and neutralizes the viral Envelope protein [58, 59].

Fig. 4
figure 4

Combinatorial RNAi strategies. Four inhibitory scenarios are plotted. Expression of multiple shRNAs from independent cassettes with RNA polymerase III, a single transcript composed of multiple miRNAs under control of an RNA polymerase II promoter, processing of multiple shRNAs from e-shRNA transcripts made by RNA polymerase III, or the expression of many siRNAs from long hairpin RNAs (lhRNAs)

Full size image

To further explore the power of combinatorial approaches, we tested the influence of RNAi-mediated knockdown on the activity of conventional antiretroviral drugs (fusion, RT, Integrase and Protease inhibitors). We compared the fold-change in IC50 (FCIC50) of these drugs in cell lines stably expressing anti-HIV and anti-host shRNAs and measured increased values for some combinations [60]. Additive or synergistic anti-HIV effects were observed with combinations of shRNAs and small-molecule drugs. The multiplication of inhibitors that target a single replication step yielded some prominent inhibitory effects. Leonard et al. reported that a combination of RNAi attack with antiretroviral drug did enhance the antiviral activity [61]. We recently demonstrated that second-generation shRNAs can be combined with Protease inhibitors to skew virus evolution, imposing an evolution block or triggering the selection of less fit virus variants [62]. The combination of two siRNAs against the viral Gag mRNA and the cellular CCR5 mRNA provided additive inhibition [63]. These combined results confirm that a high degree of anti-HIV cooperativity between shRNAs, targeting the virus or cellular cofactors, and drugs can be achieved. As previously discussed [64–66], this result supports the therapeutic interest in shRNA-drug combinatorial approaches.

Improved shRNAs and the Novel AgoshRNA Design

Several attempts to improve the shRNA hairpin design have been reported, with a particular focus on the loop segment [8, 67–73]. It is important to note that design algorithms for siRNAs cannot be applied to shRNA design [74]. Some confusion was created by the presentation of the original pSuper system as hairpins with a 9-nucleotide loop [8], which likely do allow the formation of two additional base pairs and consequently a 5-nucleotide loop [71]. The loop sequence may have an effect on shRNA processing, e.g., Dicer recognition and/or cleavage [67, 75]. miRNA-derived loops may interact with specific cell proteins to facilitate processing by Drosha and/or Dicer [76–78]. Recent evidence indicates that one could create shRNAs with a grossly different design due to Dicer-independent processing [79].

The first evidence for Dicer-independent shRNA processing came from studies on synthetic shRNAs [10, 75, 80]. A subclass of short shRNAs (sshRNAs) was described with a short stem of only 16–19 base pairs [81]. Whereas regular shRNAs are Dicer substrates, sshRNAs cannot be cleaved by Dicer in vitro [81, 82]. Yet these sshRNAs are active via RNAi-mediated target RNA cleavage [81], and a cellular endonuclease of unknown origin was suggested to execute the processing [10]. Ago2 involvement was suggested based on modification of the putative Ago2 cleavage site in the middle of the base paired stem [83]. A peculiar feature of sshRNAs that also needed to be explained is their “handedness.” The active guide strand switches to the other side of the hairpin when compared to regular shRNA molecules [81, 83].

Similar results were described for shRNA molecules that are synthesized inside the cell from gene expression cassettes. Early studies described the effect of the shRNA loop sequence [71, 84–86] and stem length [57, 82] as important determinants for regular Dicer processing. We recently identified a specific shRNA design with a short stem length and small loop that triggers an alternative processing route [87]. As described above, we observed a strand switch such that the passenger strand is effectively converted into guide strand. Sequencing indicated that cleavage occurred half-way the 3′ side of the duplex, suggesting a role for Ago2 that is predicted to cleave between base pair 10 and 11 (Fig. 5). Production of the typical approximately 30-nucleotides RNA fragments was abolished with a catalytically defective Ago2 mutant. This new design was termed AgoshRNA because the short shRNAs of 17–19 base pairs are too small to be recognized by Dicer and consequently end up in Ago2 for alternative processing. Ago2 has a dual role in AgoshRNA processing and subsequent target RNA cleavage. It is likely that these two processes are functionally coupled and executed by the same Ago2 molecule. If processed AgoshRNAs leave Ago2 prematurely, they are less likely to be bound again because of the disrupted RNA structure. Furthermore, “pre-sliced” sshRNAs molecules are inactive, which is consistent with a coupled two-step mechanism [83].

Fig. 5
figure 5

Characteristics of the traditional shRNA and novel AgoshRNA design. The regular shRNA is processed by Dicer. The shorter AgoshRNA duplex is not recognized by Dicer and is processed by Ago2, triggering a switch in guide strand from the 3′ to the 5′ arm of the duplex. See the text for further details

Full size image

We discussed the advantages of the AgoshRNA design over regular shRNAs [87] and these differences are listed in Fig. 5. AgoshRNAs produce only a single RNAi-active guide strand, which is an important feature to restrict RNAi-induced off target effects due to the passenger strands. AgoshRNAs will be the silencing method of choice for cells that lack a significant amount of Dicer, including monocytes [88]. AgoshRNAs may be safer than regular shRNAs for several reasons: saturation of Dicer as critical component of the cellular RNAi pathway is less likely and innate immunity mechanisms will be triggered less likely by these short RNA duplexes [89]. Ago2-mediated processing of shRNAs may yield more precise ends compared to Dicer processing, which is notoriously inaccurate [90]. Finally, AgoshRNAs may mimic the Dicer-independent cellular miR-451 that is loaded exclusively into Ago2, thus avoiding off target effects via Ago1, 3, and 4 [91]. Nevertheless, it is too early to tell whether the regular shRNA or special AgoshRNA design yields more RNAi-active molecules as this requires the testing of many more molecules in similar experimental settings.

Vector Issues

Although promising effects have been reported for a transient siRNA treatment in an HIV mouse model [92], long-term control of the virus would require a gene therapy approach to modify the target cells such that they can resist virus infection [93]. The idea is to develop a gene therapy with a durable effect that lasts the life span of the infected individual. To achieve such a durable effect, the optimal delivery system is based on the lentiviral vector that stably integrates in one of the chromosomes. This HIV-based vector system has recently been demonstrated to be safe in vivo, also when used to transduce hematopoietic stem cells [94–96].

A schematic of the lentiviral vector components is presented in Fig. 6a. Basically this represents the third-generation self-inactivating lentiviral vector [97–99]. The vector plasmid termed JS1 encodes the transgene from a promoter (P), which could be a Polymerase III unit encoding a shRNA transcript. The HIV-1 sequences that ensure packaging of the encoded RNA transcript into virion particles, reverse transcription and subsequent integration into the chromosomes of the target cell are marked by black boxes. The Green Fluorescence protein (GFP) reporter is made from the PGK promoter. A helper plasmid is needed to synthesize the structural Gag subunits for assembly of virion particles and the Pol enzymes that execute critical replication steps. A second helper plasmid encodes the viral Rev protein that is needed for Gag-Pol and vector gene expression through interaction with the RRE responsive element. Finally, the VSV-G glycoprotein is expressed to provide the virion particles with a broad target cell specificity. These four plasmids are co-transfected into 293T cells to initiate the production of lentiviral vector particles with the vector RNA genome, which can stably transduce a wide variety of target cells (Fig. 6b).

Fig. 6
figure 6

Lentiviral vectors for stable shRNA expression. (a) The four plasmids needed for lentiviral vector production. The vector genome is expressed from the Rous Sarcoma Virus (RSV) promoter. Transcripts start with the HIV-1 R and U5 regions, the packaging signal (ψ). The deletion introduced in the U3 region of the 3′LTR will be copied into the 5′ LTR promoter during reverse transcription, resulting in self-inactivation. To improve transduction efficiency and transgene expression in target cells, the HIV-1 central PPT (cPPT) was inserted upstream of the transgene promoter and the posttranscriptional regulatory element (PRE) was inserted downstream of the transgene, respectively. The enhanced green fluorescent protein (GFP) reporter is expressed from the phosphoglycerate kinase promoter (PGK). Transcription of the vector genome and the GFP reporter terminates at the HIV-1 polyA signal within the 3′LTR. The HIV-1 Gag-Pol gene was expressed from pSYNGP as a human codon-optimized sequence without RRE. The Rev protein was expressed from pRSV-rev. The viral vector was pseudotyped with the vesicular stomatitis virus G protein (VSV-g) expressed from pVSV-g. (b) Scheme of lentiviral vector transduction by co-transfection of 293T cells with the four plasmids. Vector particles are then used to transduce target cells, in which the vector genome is stably integrated into the DNA genome

Full size image

Because the lentiviral vector is actually based on the HIV-1 genome, one may expect some problems when anti-HIV shRNAs are introduced. We previously discussed these potential problems and presented protocols to use lentiviral vectors for an RNAi-based attack on HIV-1 [100–102]. Briefly, it is important to avoid the targeting of HIV-1 sequences that are also present in the lentiviral vector system. This is relatively easy because all HIV-derived sequence elements in the four components of this system (marked in black in Fig. 6a) have been codon-optimized such that they lose similarity to the HIV-1 genome. Other potential issues are self-targeting of the vector construct, which seems to be suppressed by the hairpin structure of shRNA constructs [103]. Constructs with miRNA-like antiviral inserts may face other specific problems like Drosha-mediated cleavage and inactivation of the vector RNA genome. We discussed strategies to avoid such adverse effects [104, 105].

Preclinical Efficacy Tests

It is important to realize that a gene therapy will likely reach only a fraction of the T cells in the human body. It is thus pivotal to study viral escape in the presence of unprotected T cells that will support ongoing HIV-1 replication and thus potentiate the risk of viral escape [106, 107]. Unhindered HIV-1 replication in unmodified T cells will generate many HIV-1 variants or a viral quasispecies, e.g., variants that acquire resistance to the shRNA inhibitor by a point mutation in the target sequence (Fig. 7). We studied virus inhibition and evolution in pure cultures of shRNA-expressing cells versus mixed cell cultures of protected and unprotected T cells [106]. The addition of the unprotected T cells indeed accelerated HIV-1 evolution and consequently triggered viral escape from a gene therapy with a single shRNA inhibitor. But the expression of three antiviral shRNAs from a single lentiviral vector prevented escape, also in the presence of unprotected cells. These results confirm the increased inhibitory capacity and more durable effects of a combinatorial RNAi approach against HIV-1.

Fig. 7
figure 7

HIV-1 escape facilitated by unmodified bystander cells. In vitro setting: pure cultures of shRNA-expressing cells (green) are protected against HIV-1 infection. HIV-1 variants will be generated at a very low rate and only shRNA-resistant variants (red virus) will be able to spread. In vivo setting: mixed cell cultures of protected (green) and unprotected (pink) cells will allow virus replication in the unprotected cells, leading to the rapid generation of a viral quasispecies by spontaneously acquired mutations. This quasispecies may also contain one or more shRNA-resistant variants that can replicate in the protected cells. Modified from ref. [106]

Full size image

It is also important to test the combinatorial RNAi regimen against a wide range of HIV-1 variants, including all circulating subtypes. Please note that the shRNA targets were selected based on sequence conservation among the different subtypes, but this obviously does not result in 100 % coverage due to the considerable genetic variability among HIV-1 strains [30]. The survey of clinical HIV-1 isolates should include drug-resistant HIV-1 variants, but we tried to avoid targets that correspond to protein domains in which drug-resistance mutations are located [16]. We have recently performed such tests, which indicate the broad effectiveness of the triple shRNA regimen [108].

Preclinical Safety Tests

Gene silencing by means of RNAi can have adverse effects on cell physiology, metabolism, and growth. The shRNAs can affect unintended mRNAs in addition to the intended target mRNA. Although complete base pairing complementary with such secondary mRNA targets can be avoided using in silico screens, silencing may occur with partial sequence complementarity [109]. Overexpression of the shRNA from powerful Polymerase III systems may trigger saturation of RNAi components like Exportin-5, Dicer, or Ago2 [110] or trigger innate immune responses [111, 112]. As mentioned, silencing of a cellular cofactor can affect cell growth. Although there are several possibilities to score cell growth over time, we realized the need for a simple assay to score subtle cell growth effects and developed the Competitive Cell Growth (CGG assay) [113]. This method is based on the difference in proliferation rate of transduced (GFP-positive) and untransduced cells in the same culture. One only has to maintain the transduced cell culture, that is the mixture of transduced and untransduced cells, and perform fluorescence-assisted cell sorting (FACS) staining for the percentage of GFP-positive cells. This percentage will go down if the transgene has a negative impact on the cell (Fig. 8). This internally controlled assay is able to detect very small differences in cell replicative capacity.

Fig. 8
figure 8

The competitive cell growth (CCG) assay. (a) Percentage of transduced GFP positive cells (green) among unmodified cells (pink) as a function of the culturing time. This culture can simply be initiated with the transduction mixture, followed by regular FACS analysis. (b) A T cell line was transduced with a shRNA-encoding lentiviral vector and passaged. The Gag5-transduced cells are gradually lost, indicating a growth deficit

Full size image

The Humanized Mouse Model

The safety and efficacy of a gene therapy protocol can be tested in a humanized mouse model (Fig. 9a). In particular, we used the BRG-HIS mouse model in which immunodeficient newborn mice are injected with human hematopoietic progenitor cells. These mice build a fairly complete human immune system consisting of different cell lineages, including mature T cells. This complex process of hematopoiesis can be monitored to screen for a negative impact of the gene therapy on cell development. For instance, we tested the impact of lentivirus-transduced anti-HIV shRNAs cells and initially reported the absence of severe adverse effects [114]. Using the very sensitive cell competition concept as described above [113], we noticed a transient negative effect of one of the four shRNAs on T cell development (Fig. 9b), which allowed us to reformulate the shRNA cocktail [115].

Fig. 9
figure 9

In vivo safety studies in the humanized immune system (HIS) mouse model. (a) Human CD34+ stem cells are transduced with a lentiviral vector expressing the indicated single or combinatorial shRNA(s) against HIV-1. The mixture of transduced cells (green) and unmodified cells (pink) are injected into newborn immunodeficient mice, followed by hematopoietic development for two months and subsequent cell analysis. (b) The bone marrow and thymus were analyzed by FACS for the percentage of transduced GFP+ human cells, which was compared to that of the input cell mixture. A ratio around 1.0 indicates that transduced cells have no growth deficit or benefit over unmodified cells. Each dot represents an individual mouse. The Gag5 shRNA and the R4 combinatorial regimen (Gag5, Pol47, Pol1, and RT5) show a detrimental effect on cell recovery. Removal of Gag5 from R4 to create R3 restored cell recovery. Modified from data in ref. [115]

Full size image

Gene Therapy Strategies for HIV-AIDS

We previously discussed the potential dangers of a gene therapy based on a lentiviral vector [100]. Although the number of treated patients is still relatively low, there is growing evidence that these vectors can be used safely for the ex vivo transduction of hematopoietic stem cells [94–96]. In fact, safety was demonstrated in an anti-HIV trial with a third-generation lentiviral vector encoding a triple RNA payload of anti-HIV shRNA, ribozyme against the CCR5 mRNA and TAR RNA decoy [49]. Similarly, a first-generation retroviral vector was safely used to deliver an anti-HIV ribozyme [116]. Thus far no therapeutic effects were scored in these trials, perhaps because of the low number of cells that were transduced in these initial studies.

We propose to develop an ex vivo gene therapy as illustrated in Fig. 10. The therapy will likely be offered to HIV-infected individuals that fail on regular drug regimens. The patient will be pretreated with granulocyte colony stimulatory factor (GCSF) to mobilize the hematopoietic precursor cells from the bone marrow into the periphery. These stem cells will be purified from the blood and transduced ex vivo with the lentiviral vector that encodes the anti-HIV RNAi arsenal. Upon infusion back into the patient, the modified cells should resist productive HIV-1 infection and thus preferentially survive over the unmodified cells that are infected by HIV-1 and removed by the immune system. This should lead to a repair of the immune system, which may be a slow and partial process, but with a durable impact.

Fig. 10
figure 10

RNAi gene therapy against HIV-1. An HIV-infected patient that fails on regular drug therapy will undergo apheresis for the collection of hematopoietic stem cells after pretreatment with granulocyte colony stimulatory factor (GCSF). The hematopoietic stem cells will be purified and transduced ex vivo with the therapeutic lentiviral construct. Transduced cells will be infused back into the patient and the antiviral shRNA(s) will protect these cells against HIV-1. The HIV-resistant immune cells will survive preferentially and should prevent disease progression

Full size image

RNAi Against Other Chronic Infections

Other chronic virus infections may be targeted by RNAi-mediated gene therapy [54, 117–123]. Virus eradication may be easier to achieve for the hepatitis B and C viruses (HBV, HCV) because chromosomal integration is not an intrinsic part of their replication cycle. In fact, recent trials with novel anti-HCV drugs indicated that complete viral clearance can be achieved fairly easily [124, 125]. This contrasts with the HIV-1 situation, where powerful combinatorial drug regimen can nearly completely suppress replicating virus, but drug treatment cannot be stopped because the viral reservoirs will reignite HIV-1 spread. This reservoir is established in different cell types, mostly resting T cells [126], but possibly also activated T cells [127]. There are likely multiple molecular mechanisms to impose HIV-1 latency, but the stably integrated HIV-1 DNA genome is central in most scenarios [128]. Much attention is given to the formulation of therapeutic strategies that will lead to a complete cure [126]. The best current option would be to start therapy as early as possible to avoid establishment of the viral reservoir, such that the immune system can control the virus once therapy is stopped [129, 130].