A fairly comprehensive picture of the positive interactions between HIV-1 and host cell proteins has been established in recent years using siRNA screens that reveal how HIV-1 exploits the host cellular machinery [1–4]. Here, we report 114 factors which, in contrast to positive factors, are involved in intrinsic immunity to establish an anti-viral attack in human cells. In this paper we refer to intrinsic immune restriction factors as cellular factors that are constitutively expressed or can be generated in cells in the absence of help from other arms of the immune system. Such factors however are unlikely to act in isolation of the immune system and may indeed be the initial trigger for innate and adaptive immune responses in the host. The proteins identified with known functions are associated with a vast array of cellular processes such as receptor signalling, vesicle trafficking, transcription, mRNA processing, DNA/RNA surveillance, cross-nuclear membrane transport and ubiquitination.
In our screen individual silencing of 114 genes rescued infectivity of HIV-1 by more than 103 FFU/ml and for some genes as many as 2 × 105 infectious units/ml (1.6-12.6 fold). The 52 most potent siRNAs 'hits' were further validated using 4 unpooled siRNAs. Rescue of viral infection by 2 or more of the 4 siRNAs was observed for 28 genes. Of these, AP2M1, DNM2, CTR9, PAF1, RTF1 and SETDB1 were further characterised for inhibition of HIV infection.
At the time of starting this screen only APOBEC3G/F and rhesus macaque TRIM5α proteins were known as factors that restrict HIV-1 replication. APOBECs are packaged into virus particles in the producer cells but act on the next round of infection in the new target cell . The HEK 293T cells used to produce the pseudovirus in our screen do not express APOBECs  so identification of these proteins was not expected. Neither did we identify TRIM5α because it is a species specific restriction factor and the human protein does not significantly affect HIV-1 [7, 31]. Since the screen was completed two other factors have been reported; p21 acts during reverse transcription and is active in stem cells, macrophages and CD4+ lymphocytes [9, 10, 32] while SAMHD1 expression is confined to dendritic and myeloid lineage cells . Tetherin, which acts at the late phase of the viral life cycle prevents viral budding , was not identified in this screen because we probed the siRNA library with an HIV pseudotype that is infectious for only a single round to the integration of provirus and GFP expression.
Using a simple model of viral restriction, intrinsic restriction factors can be loosely divided into; detectors of invasion, messengers, mediators and effectors that either prevent infection of the virus or, in the case of interferon (IFN) signalling, alert neighbouring cells. IFN activation of the JAK-STAT pathway ultimately results in the expression of anti-viral genes. For example the APOBEC and TRIM gene families are induced by type I IFNs .
After entry into the cell cytoplasm HIV-1 uncoating involves the disassembly of the viral matrix (MA) and capsid (CA) and release of the genomic RNA and the associated proteins. Thus these viral components could potentially be seen by the host cell as 'non-self' molecules. Pattern Recognition Receptors (PRRs) or viral sensors recognise distinct pathogen associated molecular patterns (PAMPs). For example, recently TRIM5α was shown to be a PRR that recognises the viral CA eventually resulting in immune signalling and AP-1 and NF-κB activation . Of similar interest, silencing of COX18 (a cytochrome c oxidase assembly protein) resulted in rescue (3.3 fold). COX18 is essential for the production of COX2 (PTGS2)  whose expression is attenuated by TRIM5α knockdown . It seems unlikely that COX18 itself is a direct PRR. It is more plausible that COX18 is an essential downstream component that through its interaction with COX2 may commission or interact with one or more 'effectors' of restriction.
During and after CA disassembly the viral nucleic acids and those produced during reverse transcription, RNA fragments, DNA and DNA:RNA hybrids may become exposed to cytoplasmic PRRs. Proteins that interact with nucleic acids might also play a role in pattern recognition. Our screen did not identify any of the classical RNA-recognising PRRs such as toll like receptors TLR3, 7 or 8 or RIG-1 like receptors RIG-1 or MDA-5 . The best characterised DNA receptor DAI (DNA dependent activator of IFN regulatory factors) responds to adenoviral DNA . In this screen silencing of POLB polymerase (DNA directed) rescued viral infection (3.5 fold). POLB rescue was validated with 2 individual siRNAs. POLB is involved in DNA damage repair and performs base excision repair required for DNA maintenance, replication, recombination . Interestingly, POLB is specifically up regulated after infection of cells with human herpes virus 16.
Once PAMPS are engaged multiple signalling cascades are activated and establish an anti-viral state . One potent hit IL1F9, an interleukin cytokine, is switched on by IFNγ. Interestingly expression of this gene is also induced after herpes simplex virus infection-1 . Signal induced antiviral responses will also use de novo transcription and translation to establish an anti-viral environment. Three transcription factors ALX3, ELF3 and PELP1 (3, 4 and 3.4 fold, validated 2/4 siRNAs) identified in the screen could be involved in this process.
The screen identified a number of genes that are involved in ubiquitination of proteins. Conjugation of the ubiquitin monomer to proteins is mediated by three families of enzymes; E1, E2 or E3 substrate specific ubiquitin ligases . E3 enzymes are critical and interact with both E2 and substrate. Generally, ubiquitination leads to the degradation of proteins by the proteasome. In viral restriction ubiquitination could be involved in the degradation process of either viral proteins or recycling of restriction proteins. The process of ubiquitination is involved in TRIM5α mediated restriction . TRIM5α has E3 ligase activity. When E2 binds to the RING domain E3 ligase activity of TRIM5α transfers the ligase from E2 to TRIM5α . HIV-1 Vif appears to use ubiquitination as a means of preventing the action of APOBEC proteins, by targeting the Cul5-elongin B elongin C-Rbx ubiquitin ligase. This results in the polyubiquitination of APOBEC3G which is then degraded by proteosomes [43–47]. A strong 'hit' in our screen was the E3 ubiguitin ligase MARCH8 (RNF178, MIR)  whose knockdown resulted in rescue (4.6 fold and 2/4 unpooled siRNA). A second strong hit was MKRN3 (markorin 3, ring finger protein, 3) where 2 individual siRNAs rescued infection (4.7 fold). Silencing of RNF19A, TRIM25 and TRIM27, which all belong (or are related) to E3 ligases also rescued infectivity (3-3.5 fold between 1 × 103 and 3 × 103 FFU/ml). Interestingly TRIM25 is essential for RIG-1 mediated IFNβ production and antiviral activity . However of the latter three genes only TRIM27 was validated with individual siRNA with only 1 out of 4 siRNAs confirming the restriction phenotype. Additionally knockdown of WSB2, a bridge-protein which connects substrate-binding domains and E3 ubiquitin protein ligases had a modest rescue (3.2 fold, 1 of 4 individual siRNA).
Once production of the viral cDNA is complete it must be transported to and enter the nucleus through specialised nuclear pores using host cell transport mechanisms . CA appears to remain associated with the PIC for some time following its production [51, 52] but disassociates from the reverse transcription complex prior to nuclear entry . Nuclear trafficking is likely to be another "check point" for the host cell. A truncated protein CPSF6 (cleavage and polyadenylation factor) a member of the S/R family protein impairs nuclear entry of PIC . This restriction can be overcome by CA mutants. In our study invalidation of NPIP (nuclear pore complex interacting protein) resulted in enhanced infection. The function of NPIP is not clear, though it co localises with NUP62, a protein needed for HIV PIC transport from cytoplasm to nucleus .
We might expect that some retroviral elements would be involved in anti-viral defence. Friend-virus-susceptibility factor (Fv-1) a gag region of an endogenous retrovirus , restricts murine leukaemia virus (MLV) at a stage after cellular entry and before integration . MLV CA is the target for Fv-1 [57–59]. We did not identify endogenous retroviral genes in this screen. However BCYRN1 which is believed to be retropositionally generated  was identified (5.8 fold, validated 2/4). BCYRN1 belongs to a family of interspersed repetitive DNA sequences and encodes a neural small untranslated non-messenger RNA. The specific neuronal expression of BCYRN1 is intriguing. Immune privileged sites such as the brain may rely more heavily on innate immune factors to prevent viral infection. The tissue specificity of BCYRN1 implies that it is unlikely to play a role in HIV-1 pathogenesis.
Genes with no obvious role
A number of genes that currently have no tentative functional role in retroviral defence were identified in the primary screen and validated with at least 2 individual siRNAs: SPSC2 (removes signal peptides from proteins as they translocate into the endoplasmic reticulum lumen), R3HDML (a serine protease inhibitor), SPAG16 (a component of microtubules) and RAPGEF3 (Rap guanine nucleotide exchange factor 3). OPTC is a protein that is associated with the extracellular matrix in the eye. ALDH8A1 is an aldehyde dehydrogenase and LOC388955 (PREL1 domain-containing protein 1) is a mitochondrial pseudogene. A functional role cannot be precluded as genes are frequently found to play a practical role in cellular processes far removed from those with which they were originally associated. Several genes, C3orf63, RPRD2, C11orf38, SLC9A3R2 and SHE had no identified function. Further studies on these genes may shed light on currently poorly understood biological processes.
Endocytosis genes AP2M1 and DNM2
Some factors identified in our screen might have been predicted: AP2M1 and DNM2, two components of the endocytic pathway were identified in the primary screen to restrict infection of cells by the pseudovirus but not by wild type HIV-1 viruses. This observation can be explained - we previously showed HIV-1 and 2 viruses can be blocked post-entry by a restriction factor Lv2 but only if the virus was delivered to the cytoplasm using a specific endocytic route mediated by an HIV-2 virus MCR envelope [14, 16, 17]. The entry route is specific to the HIV-2 envelope because pseudoviruses with VSV-G are rescued . We recently showed that blocking entry of HIV-2 MCR virus by an AP2M1/DNM2 dependent route rescued Lv2 restriction. Unlike the wild type HIV-1 viruses, the HIV-1 pseudovirus used in our screen was generated with an Env derived from HIV-2 MCR. So inhibition of the pseudovirus (but not the wild type HIV-1) would be expected to be rescued by siRNAs to AP2M1 and DNM2.
Other factors involved in endocytosis will be interesting to probe with respect to the Lv2 restriction. CCDC53 (coiled-coil domain containing protein), a component of the WASH complex, is present on the surface of endosomes and recruits the Arp2/3 complex for actin polymerisation . Two GTPases (or associated proteins) were identified: RAPGEF3 (Rap guanine nucleotide exchange factor 3) and RAB37 (small GTPase that regulates vesicle trafficking) and knockdown of either resulted in a 3.7 or 2.8 fold rescue. It will be interesting to further evaluate their role in Lv2 restriction. However, apart from vesicle trafficking, GTPases regulate a myriad of cellular functions including signal transduction and cytoskeletal organisation.
We did not predict that silencing of SETDB1 (SET domain, bifurcated 1) would rescue replication of HIV. SETDB1 is involved in TRIM family protein TRIM28 (also known as KAP1) restriction of both exogenous MLV and endogenous retrovirus expression post integration in murine cells . SETDB1, a histone methlytransferase, the NuRD histone deacetlyase complex (HDAC) and heterochromatin associated protein HP-1 are targeted to proviral DNA by TRIM28 to repress transcription. Furthermore, when transfected into human cells, SETDB1 increases HIV-LTR transactivation in conditions where Tat levels were suboptimal  but no function in the viral life cycle prior to integration had been previously described. Our screen with single round pseudotype virus should only identify genes involved in the early phase in HIV replication and not those post integration. Surprisingly, SETDB1 silencing rescued wild type HIV-1 replication by almost 2 × 104 FFU/ml (4.6 fold). Indeed all 4 siRNAs individually silencing SETDB1 resulted in viral rescue and was a gene among our most significant 'hits'. Upon further investigation we found that there was a small increase in reverse transcripts and that there was a relatively strong increase in integrated proviral DNA. Recently, Allouch et al. described a similar phenotype for restriction by TRIM28 through recruitment of HDAC . It will be interesting to probe the molecular details of this interaction in greater depth.
PAF1 complex expression renders cells restrictive to lentiviral infection
Our validation of the PAF1c showed that all of its known components are important for the observed restriction. PAF1c expression restricts HIV-1, 2 and SIV. It restricts infection of a monocytic cell line and is expressed in primary CD4+ T cells, monocytes and macrophages. The restriction results in fewer viral transcripts both in the early and late stage of infection. It also results in less integrated proviral DNA. We are not sure however if this phenotype is due to PAF1's direct action on replicating virus or if its action is through activation of expression of one or more unknown restriction factors.
A role for PAF1 in the cell cycle has been recently demonstrated  and studies using an in vitro disassembly model have suggested that uncoating requires cell cycle dependent host cell factors . However the cell cycle was unperturbed by inhibition of PAF1 expression in the Hela-CD4 cells we used in the screen suggesting that PAF1's cell cycle activity is not involved in this restriction.
There are two possibilities for the action of PAF1c in viral restriction: First, PAF1 interacts with RNA polymerase II and is proposed to be involved in transcription, elongation and stability of mRNA, surveillance and degradation . Indeed a large quantity of PAF1c is located in the nucleus so it is also possible that its activity in gene transcription may enhance expression of anti-viral factors, some of which are already 'hits' in our primary screen. Additionally we have excluded a role for APOBEC3G, TRIM5α, p21, tetherin and the recently described SAMHD1.
Second, it has also been demonstrated that SKI8 (WDR61) interacts with the PAF1c . SKI8 is also part of the SKI complex which is involved in mRNA decay [24, 27]. Given the ability of PAF1/SKI8 to interact with DNA and RNA and the localisation of both these components in the cytoplasm it is possible that PAF1c/SKI8 could act as a PRR. DNA:RNA hybrids or degrading RNA genomes are possible targets. To distinguish between the two possibilities of PAF1's action it may be of use to determine which genes are enhanced in the presence or absence of PAF1 using gene array analysis.