The data presented in this chapter are my contribution to the paper published in Nature in 2011: TRIM5 is an innate immune sensor for the retrovirus capsid lattice (Pertel et al., Nature, 2011, Annex II) [113]. Specifically, all the knockdown data performed in MDDCs and MDMs and the retroviral capsid recognition performed in MDDCs (data was obtained in collaboration with J. Guerra). Additional unpublished data connected to the project are shown as well.
As mentioned previously, human TRIM5α is able to restrict N-MLV while rhesus TRIM5α blocks HIV-1 but HIV-1 escaped restriction by human TRIM5α. The mechanism by which TRIM5 blocks retroviruses is not clearly understood but the E3 Ligase activity of the RING domain is essential for restriction [188]. In addition to be a restriction factor, it has been suggested that TRIM5α might have a role in innate signaling [287, 288]. For TRIM25, which shares similarities to TRIM5, it has been demonstrated that its E3 ligase activity is essential for RIG-I induced antiviral activity [289] and indeed the influenza non structural protein 1 (NS1) degrades TRIM25 to evade immune recognition by RIG-I [290]. Later it was discovered that TRIM25 stimulates innate immune signaling by generating free lysine 63 (K63) linked ubiquitin chains [291]. These free ubiquitin chains can serve as a platform to assemble protein kinases and have been shown to activate TAK1 leading to the activation of NF-κB and AP-1 and thus induce the antiviral state [292].
We have shown that ectopic expression of TRIM5α led to the activation of NF-κB and AP-1 reporter constructs with similar magnitude as expression of MAVS - the RIG-I downstream signal transducer - did. Expression of TRIM5α together with IRF3 was able to greatly induce an IFNβ reporter. To understand the impact of TRIM5α on innate immune signaling we created a TRIM5α knockdown in the monocytic cell line THP-1 and analyzed global gene expression after 6 hours of LPS stimulation. 33 genes were less induced by LPS stimulation in the knockdown cells compared to the control cells with the majority of the genes being proinflammatory cytokines and ISGs regulated by NF-κB, AP-1 and IRF3. We showed that TRIM5α interacts biochemically with the TAK1 complex, mainly with TAB2 and TAB3. TRIM5α knockdown in THP-1 cells inhibited TAK1 phosphorylation after LPS stimulation. Knockdown experiments
50
identified the heterodimeric E2 ligase Ubc13-UEV1A as an interacting partner of TRIM5α for AP-1 induction. In vitro assays showed that TRIMCyp or human TRIM5α together with Ubc13 and UEV1A led to the generation of free K63 ubiquitin chains.
The free K63 ubiquitin chains were shown to be able to activate TAK1 autophosphorylation. Addition of HIV-1 in vitro assembled capsid cylinders to the reaction increased TRIMCyp dependent generation of free K63 ubiquitin chains. At the same time we were able to show that MDDCs challenged with N-MLV (restricted by human TRIM5α) led to the induction and secretion of proinflammatory cytokines.
Taken together these data show that TRIM5α is a signal transducer in the TLR4 pathway and at the same time is able to bind and restrict incoming retroviral capsid and induce AP-1 and NF-κB activation via TAK1 interaction. This makes TRIM5α a PRR for the mature retroviral capsid.
As mentioned above TRIM5α knockdown interferes with LPS induced AP-1 and NF-κB signaling, while TRIM5α and IRF3 simultaneous ectopic expression leads to the induction of an IFNβ reporter construct in HEK 293 cells. We wished to test the role of TRIM5α on innate immune signaling in primary innate immune cells. Using the Vpx-VLP method to transduce MDDCs described in chapter 2 we achieved TRIM5α knockdown in MDDCs (Figure 3.2 A). The cells were stimulated with 100 ng/ml LPS and the secreted IFNβ was measured in the collected cell supernatant by titrating the supernatant onto a reporter cell line with a stably inserted luciferase reporter gene under the control of the type I IFN inducible promoter [293]. Alongside the cell supernatant recombinant IFNβ was titrated to obtain a standard curve ranging from 10 pg/ml to 250 pg/ml according to the linear range of the assay (Figure 3.1 A). IFNβ production peaked around 4 hours after LPS stimulation and became undetectable after 24 hours after LPS stimulation. The pattern of secreted IFNβ followed the mRNA expression levels peaking at 1 to 2 hours after LPS stimulation (data not shown and [117]). TRIM5α knockdown cells produced 3 to 8-fold less IFNβ depending on the donor tested (Figure 3.1 B). To asses that the difference in IFNβ production also led to an effect on the induced antiviral of the stimulated cells, TRIM5α knockdown MDDCs were either stimulated with 100 ng/ml LPS or 10 ng/ml recombinant IFNβ for 16 hours and then challenged with SIVMAC239 GFP reporter virus. The Vpx-VLPs knockdown approach did not allow us to use HIV-,1 since treatment of MDDCs with Vpx leads to the rescue of HIV-1 from the antiviral state by
51
Vpx (see chapter 4). SIVMAC on the other hand is not rescued from it by the activity of Vpx. The infected MDDCs were analyzed for GPF expression 72 hours post infection.
TRIM5α knockdown rescued SIV from LPS 3 to 7-fold in MDDCs (Figure 3.1 C left
52
panel and D) and MDMs (data not shown, Pertel et al., Nature 2011, Fig S4f). As expected there was only a small effect on IFNβ stimulation since IFNβ directly activates ISGs via STAT1 and STAT2 signaling independent of the first wave of type I IFN induction – PAMP recognition until the type I IFN induction (Figure 3.1 F). The observed increase of SIVMACinfectivity in IFNβ treated cells was below 0.1 % infected cells (Figure 3.1 C, right panel), making it difficult to judge its significance. But human TRIM5α restricts SIVMAC to some extent [294], resulting in the slight increase of infected cells in the presence of IFNβ. Since TRIM5α acts upstream of IRF3 the knockdown of the latter resembled a TRIM5α knockdown and led to a 5-fold rescue of SIVMAC from the LPS induced antiviral state while there was no effect on IFNβ signaling (Figure 3.1 E).
TRIM5α knockdown MDDCs were compared with control knockdown cells for proinflammatory gene induction in untreated cells (steady state) or after LPS stimulation at the level of mRNA induction and secreted protein (measured by Luminex assays). As it was already seen for IFNβ secretion, TRIM5α knockdown resulted in lower induction of proinflammatory genes after LPS stimulation. The most pronounced effect on the steady state levels was on IL12b with the mRNA being undetectable in the TRIM5α knockdown cells (Figure 3.2 A, top panel). IL1β and CCL8 were both decreased 7-fold. The anti-inflammatory cytokine IL10 remained unchanged showing that the knockdown only disrupted TRL4 signaling without having a general effect on other cellular pathways. 6 hours after LPS stimulation IL12b was 21-fold induced in the control knockdown while it was only 3-fold induced in the TRIM5α knockdown (Figure 3.2 A, bottom panel). CXCL10 responded the strongest to LPS induction with almost 7000-fold induction in control cells but 3-fold less in the knockdown cells.
The difference in mRNA induction was translated into a greater difference in the amount of cytokines secreted into the cell supernatant before LPS stimulation (Figure 3.2 B, top panel) or 24 hours after LPS treatment (Figure 3.2 B, bottom panel). There was 30-fold less IL12b detected in the steady state condition while LPS induced 13-fold less secreted cytokines in the knockdown cells. Soluble IL6 was decreased 26-fold in steady state. CXCL10 and CXCL9 secretion after LPS treatment was decreased in the knockdown cells 18-fold and 80-fold. The greatest effect was
53
observed on CXCL10 secretion in steady state condition with a 300-fold difference.
Similar effects although on a lower levels were observed in MDMs (Figure 3.2 C).
54
We wanted to show that TRIM5α is not only a signal transducer in the innate immune pathway and a restriction factor for retroviral capsid but is able to combine these two activities to function as a PRR capable of binding to its PAMP – in this case the retroviral capsid – and induce innate immune signaling. To that end MDDCs were challenged with either N-MLV, which is restricted by human TRIM5α or with the non-restricted N/B-MLV (Moloney) as a control. To avoid interference from the RNA genome or possible reverse transcription products MDDCs were stimulated with VLPs devoid of the genome. To exclude the possibility of contamination in the transfection reagent used to produce the VLPs the MDDCs were also challenged with medium of cells transfected with only the VSVg envelope protein (VSVg only). To ensure the plasmid, VLP and LV stocks were free from endotoxin (LPS), they were subjected to a Limulus-Amebocyte-Lysate (LAL) endotoxin test (1 EU/ml ≈ 100 pg/ml LPS) and found to be below 150 pg/ml LPS final concentration on the challenged cells. As a positive control plasmid isolated without an endotoxin removal step was used (Figure 3.3 A). MDDC challenged with N-MLV VLPs responded by upregulation of proinflammatory genes 6 hours after challenge, IFIT1 (ISG56) and CXCL9 were 3-fold higher increased compared to unrestricted capsid and on a lower level also IFIT2 (ISG54). The greatest difference was seen on CXCL10 with a 24-fold higher induction of the mRNA after N-MLV challenge (Figure 3.3 B).
As before the increase in mRNA was translated into a 1.5 to 3-fold increase in secreted proteins 24 hours after challenge with restricted capsid (Figure 3.3 B).
When the experiment was performed in MDDC knockdown for TRIM5α, N-MLV increased the production of secreted IFNβ 2-fold but was only slightly decreased in the TRIM5α knockdown cells (Figure 3.3 C). Similarly, the IFNβ induction in LPS treated TRIM5α knockdown cells was also only slightly decrease, making it difficult to judge whether this difference was due to less TRIM5α present in the cells to bind and induce signaling after capsid challenge or less efficient IFNβ induction due to the lack of TRIM5α in its role as a signal transducer (compare Figure 3.1 A).
55
56
Since we have seen repeatedly the greatest effect on CXCL10 mRNA induction and secreted cytokine production in both TRIM5α knockdown MDDC and N-MLV challenge, we wanted to create a CXCL10 luciferase reporter construct to use it to further characterize the role of TRIM5α as a PRR. To achieve that a fragment spanning -1000 bp from the transcription start site to the intron 1 of CXCL10 was cloned in front of a luciferase reporter plasmid from Invitrogen (pGL4.10, see Annex I Table 2 for primer sequences). The CXCL10 transcription start site was deleted and a new one was introduced in front of the luciferase gene. Early test failed because of to high basal level transcription from the CXCL10 promoter resulting in a high background (data not shown). To increase the sensitivity of the reporter plasmid the CXCL10 fragment was cloned into pGL4.12 containing luciferase with a CL1 and PEST destabilization domain (Luc2CP) decreasing the background due to a lowered half life of the luciferase protein (Figure 3.4 A). When HEK 293 cells were transfected with the reporter plasmid and compared to a NF-kB luciferase reporter similar luciferase activity was observed after stimulation with cytokines for 48 hours. 50 ng/ml TNFα led to a 20-fold increase in both reporters (Figure 3.4 B) while they responded only weakly after stimulation with 50 ng/ml IL1β (Figure 3.4 C). The NF-κB promoter did not respond to the stimulation with 10 ng/ml IFNβ but CXCL10 was upregulated 6-fold (Figure 3.4 D).
Unfortunately, ectopic TRIM5α expression induced the CXCL10 reporter activity only 2-fold while AP-1, as expected, increased almost 100-fold (Figure 3.4 E).
Probably AP-1 and NF-κB induction by TRIM5α is not enough to stimulate CXCL10 expression but similarly to IFNβ, TRIM5α has to be expressed in conjunction with IRF3 to potently stimulated IFNβ induction and subsequent second wave signaling leading to the induction of ISGs including CXCL10.
57
58
59