Targeting microglia with lentivirus and AAV: Recent advances and remaining challenges

2.1. Design and production of lentiviral vectors

The HIV-1 genome serves as a starting point for assembling a virion containing the transfer vector. The first generation of lentiviral vectors encoded all the packaging elements on a single vector, which posed a high biosafety risk [7]. Removing the virulence factors and separating the envelope from the packaging vector improved safety and enhanced viral replication without interfering with the production of functional HIV-1-based virions (Fig. 2b) [11]. Moreover, the 3’LTR contained a U3 deletion to abolish promoter activity for viral genome replication, and to self-inactivate the vector without affecting the virion titer or transgene expression [[12], [13], [14]]. The third, and currently preferred, generation singled out rev from the packaging vector, and replaced tat with a constitutive promoter upstream of the transfer vector (Fig. 2c) [15].

The viral envelope glycoproteins determine the lentiviral target site on the host membrane [10]. Instead of using the HIV-1 envelope gp160, which would only target cells expressing the CD4 receptor, HIV-1 can be pseudotyped with envelope glycoproteins from other viral species such as the rabies or vesicular stomatitis viruses. This increases cellular tropism and virus stability [16]. Commonly, the HIV-1 viral core is pseudotyped with the vesicular stomatitis virus glycoprotein (VSV-G), which binds to the widely expressed low-density lipoprotein (LDL) receptor [17,18].

2.2. Lentiviral transduction of microglia

The first attempts to achieve lentiviral transduction of microglia were in primary microglial cultures [[19], [20], [21], [22]]. Balcaitis et al. generated an HIV-1/VSV-G with a transfer vector encoding for GFP under the murine stem cell virus promoter [20]. Within 72 h after applying the virus to a mouse microglial culture, the authors found improved transduction as the ratio of virus to target cells (MOI) increased. Jiang et al. reported similar transduction efficiencies of 25% with MOI = 2.5, 55% with MOI = 5, 10, or 20, and 99% with MOI = 50 or 100 in rat microglia [21]. Microglial morphology was indistinguishable between control and MOI = 5, but MOI > 50 resulted in apoptosis starting from day 4.

The first in vivo lentiviral microglial transduction study was by Åkerblom et al., who transduced microglia in the rat striatum using self-inactivating-HIV-1 VSV-G-pseudotyped virions (SIN-HIV-1/VSV-G) [23]. They designed a transfer vector that encoded GFP under the ubiquitous phosphoglycerate kinase (PGK) promoter (PGK-GFP). To increase microglial specificity, the authors took advantage of an observation that there is no microRNA-9 expression in microglia, but it is commonly found in other cell types. By introducing four tandem-repeats of the microRNA-9 binding site (4miR.T) in the 3’UTR of GFP, the endogenous microRNA-9 sequesters GFP mRNA in all cells except microglia. Indeed, the authors observed 75% of Iba⁺-microglia to be GFP⁺; however, the transduction efficiency was not quantified, and there was no field-of-view image of the striatum to support the high transduction rate reported [23].

In a follow-up study, Brawek et al. performed calcium imaging of microglia in the mouse somatosensory cortex [24]. The authors used the previously described SIN-HIV-1/VSV-G but substituted the cytomegalovirus (CMV) for the PGK promoter and the calcium sensor Twitch-2B (CMV-Twitch-2B-4miR.T) for GFP. The transduction efficiency of Iba1⁺-microglia was 37% close to the injection site and 63% in the periphery. The authors also reported transduction of neurons and astrocytes near the epicenter and suspect that the endogenous microRNA-9 pool was insufficient to counteract the strong transgene transcription by the CMV promoter.

The latest study, by Nie et al., focused on how the toll-like receptors (TLR) 2/4 in microglia of the prefrontal cortex contribute to social avoidance behavior induced by repeated social defeat stress [25]. The authors generated a transfer vector based on the PGK-GFP-4miR.T construct and replaced the GFP with a double-floxed inverted open reading frame (DIO) encoding mCherry and RNAi for TLR2 and TLR4 (PGK-DIO-mCherry-TLR2/4RNAi). Upon Cre-mediated recombination of the DIO, mCherry and RNAi are expressed. The authors injected the virus into the Cx3cr1^{tm2.1(cre/ERT2)Litt}/WganJ mouse model [26], which resulted in microglia-specific YFP and transgene expression after injection with tamoxifen [27]. After four weeks, microglial transduction efficiency was 25% and 45% for control and RNAi, respectively, and up to 90% specific for microglia. The discrepancy between efficiencies in control and RNAi transduced mice was not further addressed. This study shows that microglial specificity can be increased by combining Cre-mouse models with viral strategies.

In summary, all lentiviral in vivo studies thus far have taken advantage of the SIN-HIV-1/VSV-G system and adapted the promoter or transgene cargo (Table 1).

3.1. Design and production of AAV vectors

An estimated twelve naturally occurring AAV serotypes have been identified, of which AAV serotype 2 (AAV2) is commonly used in research [[35], [36], [37]]. To generate a recombinant AAV, a packaging cell line expressing the adenoviral gene E1+ must be transfected with a packaging vector, which supplies the rep and cap genes, a transfer vector containing the transgene flanked by ITRs, and a helper vector encoding for E2a, E4, and VA (Fig. 4b). Importantly, the transgene size is limited to 4.5 kb because the AAV capsid is only 20 nm in diameter, which constrains the size of the genome [38]. AAV pseudotyping increases tropism and efficiency: For example, AAV2/10 contains the ITR and rep from AAV2 and cap from AAV10 [[39], [40], [41], [42], [43], [44], [45], [46]].

The host cell replication machinery is required to synthesize the complementary DNA strand. This is a rate-limiting step in the AAV transduction process. Self-complementary AAVs (scAAVs) were developed to overcome this, although it further reduces the packaging capacity to approximately 2.2 kb, making it challenging to package a cell type-specific promoter and a transgene-of-interest [47,48].

3.2. AAV transduction of microglia

Bartlett et al. were the first to propose that AAVs are a viable option for targeting microglia in vivo. They injected Cy3-labeled AAV2/2 driving GFP expression under the CMV promoter into the inferior colliculus and hippocampus. Cy3 was detectable in microglia after 24 h, but microglia never expressed GFP, suggesting that the virus was either degraded or prevented from expressing the transgene [49].

In 2003, Cucchiarini et al. sought to compare AAV2/2 and AAV2/5 transgene expression levels of RFP driven by promoters from the macrophage lineages. In primary rat microglial culture, the authors describe efficiencies of 25% for F4/80, 10% for CD68, and only one cell per field-of-view for CD11b. They then injected AAV2/5-F4/80-RFP into the rat striatum and observed F4/80-RFP⁺-microglia cells, along with other unidentified cell types [50].

To find a more effective serotype, Su et al. generated several pseudotyped AAVs using a CMV-GFP transfer, and packaging vectors with capsids 2, 5, 6, 8, and 9. In primary murine microglial cultures, AAV2/6 was the most effective for transducing microglia in vitro as reported by the greatest fold change in GFP mRNA [19].

Rosario et al. modified the AAV6 capsid (AAV6^™) through site-directed mutagenesis of two tyrosine residues to phenylalanine and a threonine to valine (Y731F/ Y705F/ T492V) [51]. These modifications prevent proteasomal degradation when AAV escapes the endosomal compartment [[52], [53], [54], [55]], and has been shown to increase transduction efficiency in monocyte-derived dendritic cells [[52], [53], [54]]. The authors report a consistent 95% transduction efficiency for the scAAV2/6^™-F4/80-GFP and scAAV2/6^™-CD68-GFP in primary microglia or mixed glial cultures, using quantitative real-time PCR. Iba⁺/GFP⁺ microglia were detectable with AAV2/6^™ injected either into the ventricle of P0 pups or into the adult hippocampus. The authors quantified specificity, and emphasize in the text the exclusive labeling of microglia. However, the graph contradicts this statement, indicating a specificity of only 75% for F4/80 and 20% for CD68. In addition, they report a high efficiency in vitro, but the in vivo efficiency is not quantified. Overall, this study provides one of the first attempts to rationally engineer AAVs to improve microglial transduction in vivo.

The most recent study focused on manipulating microglia with DREADDs (Designer Receptor Exclusively Activated by a Designer Drug) to reverse pain response [56]. Grace et al. intrathecally transduced microglia in the spinal cord with AAV2/9 containing DREADD driven under the CD68 promoter. The authors had previously validated this strategy and shown that DREADDs colocalize with Iba1⁺ cells, but not with NeuN⁺ neurons or GFAP⁺ astrocytes [57]. Unfortunately, the high GFP background fluorescence levels make it difficult to evaluate this conclusion.

In summary, less than 20% AAV transduction efficiency can be achieved in vivo (Table 2), confirming that AAV transduction of microglia is challenging. This observation aligns with other publications that have focused on improving in vivo AAV transduction for other CNS cell types. Most approaches report a lack of microglial transduction upon different delivery strategies, including intrastriatal injection [44], systemic delivery [58,59], or neonatal intracerebroventricular delivery [40,60]. Only a few studies describe minimal microglial transduction without further quantification [41,61]. Although AAVs seem to be the preferred option for targeting neurons, extensive research is still needed to improve in vivo transduction efficiency in microglia.

4.1. Viral entry

Viruses exhibit high affinity for defined host cell receptors, providing access to the cell upon attachment. For example, the HIV-1 envelope glycoprotein gp160 binds the CD4 receptor and co-receptors Cxcr4 and Ccr5 [63], the latter two are transcribed in microglia [64,65]. Co-receptors such as the integrins Itgav and Itgb5 [66,67], hepatocyte growth factor receptor (c-MET) [68], and fibroblast growth factor receptor 1 (FGFR1) [69,70] have been implicated in AAV attachment and internalization, from which the integrins are expressed in microglia.

Therefore, the first step towards optimizing transduction is to screen for microglial receptors, which are targeted by lentiviral envelopes or AAV capsids. This would provide the building blocks for generating pseudotyped virions. To improve attachment of the microglia-trophic AAV2/6 and 2/9, the screening must also consider membrane glycans on the microglial surface. The icosahedral AAV6 capsid binds to the glycan moieties heparin sulfate proteoglycan (HSPG) and α2-3 and α2–6 N-linked sialic acid [51,[71], [72], [73]], while AAV9 binds terminal N-linked galactose [57,74]. Such knowledge would facilitate the development of further targeted mutations of HSPG binding sites on the capsid, which has been shown to alter viral particle spread and efficiency in tissue tropism using AAV2/2 [44,75,76].

Although this strategy appears straightforward, a word of caution must be raised: microglia may elicit an unexpected response similar to macrophages. Macrophages express pattern recognition receptors (PRRs) such as toll-like receptors or sialic acid-binding immunoglobulin-type lectin 1 (Siglec1). These PRRs specialize in recognizing pathogen-associated molecular patterns (PAMPs) like viral glycoprotein structures [77]. Once PRRs recognize PAMPs, they induce a type 1-interferon response, which converts macrophages to a prophagocytic state. This leads to release of pro-inflammatory cytokines and, eventually, destruction of the virus [78]. Whether these mechanisms are also found in microglia is still under investigation, but several studies have suggested parallels [79].

Thus far, it has been shown that microglia change their receptor expression when they are in a prophagocytic state [80,81]. The current viral brain delivery systems result in local tissue damage, which leads to microglial activation [82]. On the one hand, this could be beneficial for microglial transduction since changes in the receptor distribution could enhance viral binding. On the other hand, microglia might be locally alerted to foreign viral particles and activate their internal viral defense mechanism. To circumvent this, alternative viral delivery strategies should be investigated, such as peripheral injection in the tail vein [58] or nasal administration [83].

4.2. Viral host defense-escape mechanisms

Once the virus successfully enters the cell, the virus must control enzymes in the host cell to ensure viral replication. Host cells, however, have developed strategies to detect and counteract a viral invasion. Lentivirus packages critical enzymes within the viral core, enabling immediate reverse transcription and integration into the genome (Fig. 1). In addition, after shedding its envelope the wildtype lentivirus releases the virulence factors Vpx and Vif, which target the host restriction factors SAMHD1 and APOBEC3 that interfere with reverse transcription and modify the reverse-transcribed viral DNA, respectively [84,85]. These virulence factors have been removed from the lentiviral vector system for biosafety reasons (Fig. 2). It would be interesting to see whether microglia express SAMHD1 and APOBEC3, and whether partially restoring the virulence factors would increase microglial transduction efficiency. Alternatively, drugs that prevent SAMHD1 or APOBEC3 activation could be developed.

Specific host defense mechanisms against AAVs have not been described. AAVs enter the cell via the endosomal pathway, but have to escape from the endosome before its fusion with lysosomes (Fig. 3) [[86], [87], [88]]. When and how AAVs leave the early or late endosome is still debated. Similarly, the mechanism for AAV entry into the nucleus is not known [87,89]. Only 20% of AAV2 entering the cell reaches the nucleoplasm, suggesting that several unknown mechanisms exist that prevent nuclear import [90]. Several groups have shown AAV transduction efficiency can be enhanced in vitro by applying drugs like bafilomycin, which prevent endosomal acidification by inhibiting the proton pump [91,92]. However, such drugs cannot be used in vivo because they do not cross the blood-brain barrier, and cause severe side effects [93]. Overall, a better knowledge of the host inhibitory factors, as well as the viral pathway to the nucleus, will significantly improve microglial transduction strategies.

4.3. Microglia-specific transgene expression

Specificity is another challenge faced when transducing microglia because often, additional cells are transduced (Table 1, Table 2). Therefore, identifying alternative constitutively expressed microglia-specific promoters is required to optimize transgene expression. Until now, the research field has focused on promoters from the macrophage lineage like CD11b, CD68, and F4/80, which also label monocytes, and show varied expression depending on the microglial activation state. For example, CD68 is more strongly expressed in proinflammatory microglia [94]. Now, with the availability of next-generation sequencing data [[95], [96], [97], [98]], it should be possible to identify genes and their corresponding promoters that are constitutively and exclusively expressed in microglia.

An alternative strategy would be to use mouse models that express Cre in a defined cell type. The transfer vector would then include the promoter and a double-floxed inverted orientation (DIO) sequence which is inverted upon Cre activation [99]. The tamoxifen-inducible Cx3cr1^CreERT2 mouse model has frequently been used to target microglia [100]. Administering tamoxifen any time after embryonic day E14.5 induces Cre recombinase activity in microglia, blood-derived macrophages, and monocytes [101]. Due to their rapid turnover, macrophages and monocytes will be replaced by non-recombined cells within one month, leaving microglia forming more than 90% the enriched recombined population [27,99,[101], [102], [103]]. Nie et al. were one of the first to combine the mouse model and DIO-expressing lentivirus to specifically target microglia [25]. They found high specificity, suggesting that this is a valuable strategy for studying effects in adult microglia.