TRIM5α requires Ube2W to anchor Lys63-linked ubiquitin chains and restrict reverse transcription

TRIM5α is an antiviral, cytoplasmic, E3 ubiquitin (Ub) ligase that assembles on incoming retroviral capsids and induces their premature dissociation. It inhibits reverse transcription of the viral genome and can also synthesize unanchored polyubiquitin (polyUb) chains to stimulate innate immune responses. Here, we show that TRIM5α employs the E2 Ub-conjugating enzyme Ube2W to anchor the Lys63-linked polyUb chains in a process of TRIM5α auto-ubiquitination. Chain anchoring is initiated, in cells and in vitro, through Ube2W-catalyzed monoubiquitination of TRIM5α. This modification serves as a substrate for the elongation of anchored Lys63-linked polyUb chains, catalyzed by the heterodimeric E2 enzyme Ube2N/Ube2V2. Ube2W targets multiple TRIM5α internal lysines with Ub especially lysines 45 and 50, rather than modifying the N-terminal amino group, which is instead αN-acetylated in cells. E2 depletion or Ub mutation inhibits TRIM5α ubiquitination in cells and restores restricted viral reverse transcription, but not infection. Our data indicate that the stepwise formation of anchored Lys63-linked polyUb is a critical early step in the TRIM5α restriction mechanism and identify the E2 Ub-conjugating cofactors involved.


Introduction
TRIM5a is an innate immune effector that belongs to the tripartite (TRIM) protein superfamily. The conserved tripartite architecture comprises a really interesting new gene (RING) E3 ubiquitin (Ub) ligase domain, one or two B-box domains, and a coiled-coil, followed by variable protein interaction domain(s). Although TRIM proteins can have wide-ranging biological roles, anti-pathogen defense functions are common (Nisole et al, 2005;McNab et al, 2011). The expression of many TRIM proteins is induced by type I interferons, and several have been shown to manipulate immune signaling pathways by ubiquitinating signal transducing proteins, such as TRIM23, TRIM25, and TRIM56 (Gack et al, 2007;Arimoto et al, 2010;Tsuchida et al, 2010). Certain TRIM proteins, including TRIM5a , TRIM21 (Mallery et al, 2010), and TRIM56 (Wang et al, 2011), have also been shown to target viral replication specifically. An emerging theme is that antiviral TRIM proteins both inhibit viral infection and initiate innate immune signaling cascades that lead to inflammatory cytokine production and an antiviral state (Pertel et al, 2011;McEwan et al, 2013;Uchil et al, 2013;Versteeg et al, 2013). The molecular mechanisms that underlie these multiple activities are not yet well defined and are of great interest for understanding the innate immune detection of viruses.
TRIM5a represents an important barrier to zoonotic retroviral infection Wu et al, 2013Wu et al, , 2015. Multimers of homodimeric TRIM5a bind susceptible retroviral capsids in the cytoplasm, inducing premature capsid dissociation and inhibiting reverse transcription Diaz-Griffero et al, 2009;Black & Aiken, 2010;Ganser-Pornillos et al, 2011;Zhao et al, 2011;Goldstone et al, 2014). Several observations indicate the involvement of the Ub-proteasome system in these processes: (i) proteasome inhibitors restore slow capsid dissociation kinetics and rescue formation of integration-competent reverse transcripts, but do not restore viral infectivity Wu et al, 2006;Kutluay et al, 2013); (ii) TRIM5a and proteasomal components co-localize in cells (Campbell et al, 2008;Lukic et al, 2011;Danielson et al, 2012); and (iii) the short half-life of TRIM5a is decreased even further upon recognition of restriction-sensitive capsids (Rold & Aiken, 2008), and lengthened by proteasome inhibition . Thus, one plausible model for restriction is that TRIM5a-capsid complexes are diverted into a constitutive, proteasome-dependent unfolding/degradation pathway (Towers, 2007), although an alternative autophagosomal unfolding/degradation model has also been suggested (Mandell et al, 2014). In addition to proteasome recruitment, TRIM5a has been shown to generate unanchored Lys63-linked polyubiquitin (polyUb) chains upon binding to susceptible retroviral capsids. These unanchored polyUb chains are postulated to activate transforming growth factor-b-activated kinase 1 (TAK1), which in turn stimulates NF-jB nuclear translocation and induces antiviral gene expression (Shi et al, 2008;Pertel et al, 2011).
The Ub enzymology underlying different TRIM5a activities remains to be defined. Toward this end, we have employed genetic and biochemical approaches to identify and characterize two E2 Ubconjugating enzymes required for TRIM5a restriction of retroviral reverse transcription: Ube2W and the heterodimeric Ube2N/Ube2V2. We find that Ube2W can attach single Ub molecules directly to TRIM5a, which can then act as substrates for Ube2N/Ube2V2 in the synthesis of TRIM5a-anchored Lys63-linked polyUb chains, both in vitro and in cells. We find that TRIM5a is aN-acetylated in cells, a modification that blocks N-terminal ubiquitination, and that Ube2W targets specific TRIM5a lysine residues with Ub in vitro. Furthermore, Ube2W, Ube2N/Ube2V2, and the Ub residue Lys63 are each required for TRIM5a restriction of viral DNA synthesis, but not infection. Thus, our data indicate that TRIM5a employs multiple E2 enzymes to synthesize protein-anchored Ub chains that are required for inhibition of retroviral reverse transcription.

Results
TRIM5a requires two non-redundant E2 enzymatic activities to inhibit retroviral DNA synthesis We performed an RNAi screen to identify non-redundant E2 Ub-conjugating enzymes required for TRIM5a antiviral activity. Because proteasome inhibitors restore retroviral reverse transcription without rescuing viral infectivity Wu et al, 2006;Kutluay et al, 2013), we reasoned that abrogating functionally important ubiquitination activities associated with TRIM5a should rescue reverse transcription, but might not rescue viral infectivity. Human TRIM5a (TRIM5a hu ) acts differently on two closely related variants of murine leukemia virus (MLV) that differ in their capsid sequences: TRIM5a hu restricts N-tropic MLV (N-MLV) but not B-tropic MLV (B-MLV) (Towers et al, 2000;Besnier et al, 2003;Keckesova et al, 2004;Perron et al, 2004;Yap et al, 2004). B-MLV therefore serves as a convenient control for differentiating between loss of TRIM5a hu -dependent E2 activities, which should specifically affect N-MLV infection, versus alterations in cell viability or metabolism, which should affect both N-MLV and B-MLV. We used singleround, VSV-G-pseudotyped N-and B-MLV vectors expressing GFP to assess endogenous TRIM5a hu activity. As expected, these two viral vectors were equally infectious in feline CrFK cells, which do not express a restricting TRIM5a protein (McEwan et al, 2009) (Supplementary Fig S1A), but were differentially infectious in human HeLa cells, which endogenously express TRIM5a (Fig 1A,  Mock). siRNAs targeting 38 different human E2 or E2 variant (UEV) proteins, which collectively encompass the majority of known human E2 enzymes (Markson et al, 2009), were then tested for their ability to restore N-MLV DNA synthesis (Fig 1A). The three strongest hits were Ube2W, Ube2N (Ubc13), and Ube2V2 (Mms2); individual depletions of any of these E2 proteins abrogated TRIM5a hu -dependent restriction of N-MLV DNA synthesis completely. These three proteins likely represent just two different E2 enzymatic activities because Ube2N and Ube2V2 function together in a heterodimeric E2 enzyme complex comprising an active Ube2N E2 subunit and a catalytically inactive Ube2V2 subunit that provides specificity for synthesis of Lys63-linked polyUb chains (Hofmann & Pickart, 1999;McKenna et al, 2003).
The strongest hits from the siRNA screen were validated by individually depleting Ube2W, Ube2N, and Ube2V2 from restrictive human TE671 cells using shRNA stably expressed from retroviral vectors (Fletcher et al, 2010). E2-depleted cells were infected with equivalent doses of N-or B-MLV, and viral DNA synthesis and infectivity were measured 6 and 48 h postinfection, respectively. Cells expressing a scrambled control shRNA restricted N-MLV efficiently, as measured by strong reductions in DNA synthesis ( Fig 1B) and infectivity ( Fig 1C). However, depleting either Ube2W, Ube2N, or Ube2V2 substantially rescued N-MLV DNA synthesis, without major effects on unrestricted B-MLV DNA synthesis ( Fig 1B). As predicted, E2 depletion rescued DNA synthesis more effectively than it rescued viral infectivity, validating the logic of the screen readout [compare Fig 1C (infection) with B (DNA synthesis)]. Specific depletion was confirmed by RT-qPCR for Ube2W and Ube2V2 (Fig 1D and Supplementary Fig S1B) or immunoblot (IB) for Ube2N ( Fig 1E). Because Ube2N can function with either Ube2V2 or Ube2V1 (Uev1a) (Hofmann & Pickart, 1999), we confirmed that depletion of Ube2V1 did not rescue restricted N-MLV DNA synthesis in TE671 cells ( Supplementary Fig S1C and D). Furthermore, depletion of Ube2V2 had no impact on the expression of either Ube2V1 ( Fig 1D) or Ube2N (Fig 1E). These experiments indicate that both Ube2W and Ube2N/Ube2V2 are required for efficient restriction of retroviral reverse transcription by TRIM5a hu .

TRIM5a can be ubiquitinated and degraded in a proteasomedependent fashion
We next analyzed the ubiquitination state of TRIM5a hu under conditions of proteasome inhibition. We treated TE671 cells exogenously expressing HA-TRIM5a hu with the proteasome inhibitor MG132, and HA-TRIM5a hu species were detected by IB against the HA-tag (Fig 2A). A 2-hour MG132 treatment leads to accumulation of higher molecular weight (HMW) HA-TRIM5a hu species (Fig 2A, lane 2). The HMW products continued to increase with time, ultimately becoming the predominant TRIM5a species after 10 h of MG132 treatment (Fig 2A, lane 5). Measurement of the HA band densities corresponding to the unmodified and HMW forms of HA-TRIM5a hu demonstrated that it is not the unmodified form that accumulated in these experiments (Fig 2A); rather it is the HMW species that accumulate, consistent with post-translational modification, such as ubiquitination. We assume their accumulation reflects a lack of degradation owing to proteasome inhibition.
To control for non-specific effects of overexpressing a RING domain-containing protein in cells, we also performed parallel control experiments on TE671 expressing HA-TRAF6, another RING E3 ligase involved in innate immune signaling (Xia et al, 2009). In contrast to TRIM5a hu , only very small amounts of HMW HA-TRAF6 species accumulated upon MG132 treatment, whereas the unmodified form accumulated (Fig 2B) is not a general property of overexpressing an active RING E3 ligase. Our observations are consistent with a model in which TRIM5a hu is post-translationally modified, leading to accumulation of HMW species that are normally turned over in a proteasome-dependent fashion. This result should be treated with some caution, however, because proteasome inhibition, and consequent Ub depletion, can also affect other Ub-dependent cellular processes.
To confirm that the HMW TRIM5a hu species were ubiquitination products, we co-transfected 293T cells with 6×His-tagged TRIM5a hu (His-TRIM5a hu ) and HA-tagged Ub (HA-Ub). His-TRIM5a hu species were enriched by Ni 2+ affinity purification, and the HMW His-TRIM5a hu species were confirmed as ubiquitinated products by IB, detecting the HA-tag (Fig 2C, lanes 4 and 6). Importantly, in this case, ubiquitinated His-TRIM5a hu species were detectable in the absence of MG132 treatment (lane 4). We assume that these species were enriched by the affinity purification step and could therefore be detected by IB. Ubiquitinated TRIM5a hu ((Ub) n -TRIM5a) was also detected in a reciprocal affinity purification (pulldown of His-Ub, IB detection of HA-TRIM5a hu ), again in the absence of MG132 treatment (Fig 2D). These experiments demonstrate that A, B Proteasome inhibition reveals TRIM5a hu high molecular weight (HMW) species. TE671 cells that expressed HA-TRIM5a hu (A) or a control RING protein, HA-TRAF6 (B), were treated with the proteasome inhibitor MG132, lysed at the indicated time intervals, and analyzed by IB detection of the HA-tag (HA), or b-actin (loading control). The band intensity ratios (ImageJ) (HA:b-actin) of unmodified HA-TRIM5a hu (labeled "1") and HMW HA-TRIM5a hu (labeled "2") are displayed beneath and normalized to the 0-h time point and plotted. C, D Constitutive TRIM5a hu autoubiquitination can also be detected in the absence of MG132. 293T cells were transfected with the designated combinations of expression constructs for 6×His-TRIM5a hu and HA-Ub. After 48 h, 6×His-TRIM5a hu was pulled down (PD) with a Ni 2+ affinity matrix, followed by analysis of the input and affinity-purified samples by IB with the designated antibodies (C). TE671 cells that stably expressed HA-TRIM5a hu were mock-transfected or transfected with a construct expressing 6×His-Ub. Cells were not treated with MG132. After 48 h, 6×His-Ub was affinity-purified and the samples were analyzed by IB (D). Antibodies were against the HA-tag (HA), the His-tag (His), or b-actin (loading control). Note some unmodified TRIM5a hu precipitates with the Ni 2+ affinity matrix in a 6×His-Ub-independent manner.  (Rold & Aiken, 2008).

Depletion of Ube2W or Ube2N/Ube2V2 suppresses TRIM5a ubiquitination in cells
We next tested whether the E2 enzymes required for TRIM5a hu restriction of N-MLV DNA synthesis also participate in TRIM5a hu ubiquitination. As shown in Fig 3, single depletions of Ube2W, Ube2N, or Ube2V2 each strongly inhibited the formation of HMW TRIM5a hu Ub adducts that otherwise accumulated upon proteasome inhibition. However, the resulting distributions of TRIM5a hu species varied in the different cases. Ube2W depletion abrogated all detectable TRIM5a hu ubiquitination (Fig 3A), whereas depletion of either Ube2N or Ube2V2 prevented polyUb chain formation but did not prevent formation of a monoubiquitin (monoUb) TRIM5a hu species (Ub-TRIM5a hu , Fig 3B and C). The similar accumulation of Ub-TRIM5a hu following depletion of either Ube2N or Ube2V2 is consistent with the idea that these two E2 activities function together as a single Ub E2 conjugating enzyme. Together, our observations are consistent with a model in which Ube2W mediates TRIM5a monoubiquitination and the Ube2N/Ube2V2 enzyme then extends the polyUb chain from the anchoring Ub.

TRIM5a is modified with Lys63-linked polyUb chains in cells
The observations that HMW TRIM5a hu species accumulate in the absence of proteasome activity and that depletion of either Ube2W or Ube2N/Ube2V2 blocks this accumulation were surprising because proteasome recruitment and protein degradation are typically mediated by Lys48 (K48)-linked polyUb chains, whereas Ube2W typically modifies proteins with monoUb and the Ube2N/ Ube2V2 heterodimer typically builds Lys63-linked polyUb chains (K63-Ub) (VanDemark et al, 2001;Christensen et al, 2007;Scaglione et al, 2013;Tatham et al, 2013). We therefore characterized the nature of the polyUb linkages within the TRIM5a hu HMW species by co-expressing and purifying mutant Ub proteins that either could not form K63-Ub (Ub K63R) or could not form K48-linked chains (Ub K48R). As shown in Fig 4A, formation of (Ub) n -TRIM5a hu species was normal after expression of Ub K48R, but was efficiently blocked by expression of Ub K63R, consistent with the idea that Ube2N/Ube2V2 catalyzes formation of K63-Ub on TRIM5a hu ( Fig 4A). Importantly, Ub K63R species were present at normal levels in cell lysates as verified by detecting the 6×His tag ( Fig 4A, panel 2). The Ub K63R-mediated suppression of TRIM5a hu HMW species was as efficient as the suppression seen upon depletion of Ube2N ( Fig 4B). These observations suggest that TRIM5a is specifically labeled with K63-Ub by Ube2N/Ube2V2. In further support of this hypothesis, an antibody specific for K63-linked Ub chains reacted well with affinity-purified His-TRIM5a hu HMW species ( Fig 4C). In this case, K63-linked (Ub) n -TRIM5a hu species were detected in the absence of exogenous Ub expression ( To test the functional importance of K63-linked polyUb modification of TRIM5a for restriction, we overexpressed Ub K63R in TE671 cells and assayed restriction of N-MLV DNA synthesis. Like Ube2N depletion (Fig 1), Ub K63R overexpression restored restricted N-MLV DNA synthesis to nearly the same level as the unrestricted B-MLV control ( Fig 4D). Once again, viral infectivity was not substantially restored ( Fig 4E), as measured in a luciferasebased infectivity assay. Together, these results indicate that Ube2N/Ube2V2 adds K63-linked polyUb chains to TRIM5a and that these modifications are required for restriction of N-MLV DNA synthesis.

The N-terminus of TRIM5a is quantitatively acetylated in cells
Ube2W has been shown to catalyze formation of a conventional peptide bond between the C-terminus of Ub and the free N-terminus of several proteins in vitro (Scaglione et al, 2013;Tatham et al, 2013;Vittal et al, 2015). However, most eukaryotic proteins are cotranslationally acetylated at their N-termini (aN-acetylation), often following removal of the initiator methionine (  because the resulting amide nitrogen is no longer nucleophilic. The TRIM5a N-terminal sequence, (M)ASGIL, should be a good substrate for aN-acetylation (Polevoda & Sherman, 2003), and TRIM5a aN-acetylation has been reported previously (Bienvenut et al, 2012). We confirmed this observation by analyzing C-terminally FLAG-One-Strep (FOS)-tagged human or rhesus macaque (rh) TRIM5a proteins expressed in human 293T cells. The exogenously expressed TRIM5a-FOS proteins were purified by Strep-Tactin affinity chromatography, separated by SDS-PAGE, and digested with trypsin, and the tryptic peptides were analyzed by liquid chromatography coupled with tandem mass spectrometry (LC/MS/MS). In both constructs, we identified N-terminal peptides that lacked the terminal methionine and were acetylated on the penultimate Ala residue, whereas non-acetylated or methionylated peptides were not detected in either case (Supplementary Fig S2A for human and  Supplementary Fig S2B for rhesus). Hence, human and rhesus TRIM5a proteins are N-terminally acetylated in human cells, which raises the question of how they might be monoubiquitinated by Ube2W.

Ube2W monoubiquitinates full-length TRIM5a in vitro
To characterize how Ube2W and Ube2N/Ube2V2 can function together with TRIM5a, we tested their activities in reconstituted systems using pure recombinant proteins. These experiments employed full-length rhesus TRIM5a proteins (TRIM5a rh ), which are expected to utilize the same set of functionally important Ub E2 enzymes as TRIM5a hu proteins because TRIM5a rh efficiently restricts retroviral DNA synthesis when expressed in human cells . The C-terminally tagged TRIM5a rh -FOS was expressed in insect cells, purified by Strep-Tactin affinity chromatography, treated with precision protease to remove the FOS tag, and purified to homogeneity by gel filtration chromatography. Electrospray mass spectrometric analysis of the intact purified recombinant TRIM5a rh protein indicated that demethionylation and aN-acetylation were complete because species representing methionylated or demethionylated and non-acetylated forms of TRIM5a rh were not detected (Supplementary Fig S3). These data indicate that our recombinant TRIM5a rh protein had the same N-terminal sequence and modification pattern as TRIM5a proteins expressed in human 293T cells. Ub (Pickart & Raasi, 2005), E1 (UBA1), and E2 enzymes were all expressed in E. coli and purified to homogeneity for use in the in vitro ubiquitination assays, see Materials and Methods section.
To assay Ub ligase activity, TRIM5a rh was incubated with UBA1, Ub, and ATP, together with the different E2 enzymes that were identified in our cellular assays (Ube2W or Ube2N/Ube2V2 or the closely related Ube2N/Ube2V1), or with one of five different control E2 enzymes (Ube2B, Ube2C, Ube2H, Ube2L3, or Ube2D3). We verified that each purified E2 was active and efficiently charged with Ub by UBA1 (Supplementary Fig S4). TRIM5a rhdependent Ub ligase activity was monitored by IB, with an anti-Ub antibody used to detect formation of both unanchored and TRIM5a rh -anchored ubiquitination products (Fig 5A), and an anti-TRIM5a rh antibody used to monitor formation of TRIM5a rhanchored ubiquitination products alone ( Fig 5B). Ube2D3 was used as a positive control in these experiments because this E2 enzyme is highly promiscuous and has previously been shown to be capable of ubiquitinating TRIM5a rh in vitro (Yamauchi et al, 2008;Kim et al, 2011). As expected, Ube2D3 catalyzed the formation of TRIM5a rh -linked polyUb chains, and possibly also unanchored chains (Fig 5A and B, lane 9). However, Ube2D3 lacked chain linkage specificity, because individual Ub lysine mutations could not perturb TRIM5a rh autoubiquitination (Supplementary Fig S5A  and Table 1). Moreover, Ube2D3 does not function in TRIM5a restriction because its efficient depletion did not significantly abrogate TRIM5a restriction of N-MLV DNA synthesis or infectivity ( Supplementary Fig S5B-D).
As shown in Fig 5, the closely related Ube2N/Ube2V1 and Ube2N/Ube2V2 E2 enzymes could collaborate with TRIM5a rh to synthesize polyUb chains (Fig 5A, lanes 6 and 7) that were not covalently anchored to TRIM5a rh (Fig 5B, lanes 6 and 7). This observation agrees well with a previous report that Ube2N/Ube2V1 and TRIM5a rh synthesize unanchored K63-Ub (Pertel et al, 2011). Ube2W was the only other assayed E2 that exhibited any observable enzymatic activity (Fig 5, lane 5). In this case, however, the enzyme predominantly attached a single Ub molecule directly onto TRIM5a rh , and also made a small amount of TRIM5a rh with two attached Ub molecules (Fig 5B, lane 5). The ability of Ube2W to transfer a single Ub onto TRIM5a rh in vitro is consistent with the pattern of ubiquitination seen in cells, where depletion of Ube2N/ Ube2V2 activity led to accumulation of Ub-TRIM5a hu (Fig 3B and C) and depletion of Ube2W blocked all detectable TRIM5a hu ubiquitination ( Fig 3A). Thus, Ube2W can catalyze the addition of a single

A
(Ub) n -TRIM5a hu synthesis is inhibited by expression of K63R but not K48R mutant Ub. IB of PD products on Ni 2+ affinity matrices from lysates of TE671 cells expressing HA-TRIM5a hu transfected with empty vector or 6×His-tagged Ub WT, Ub K48R, or Ub K63R. TRIM5a hu species were detected by IB for the HA-tag (HA), and equivalent purifications of His-labeled ubiquitinated proteins for each Ub variant were confirmed by IB detection of the 6×His tag (His). Input fractions were analyzed to confirm that similar levels of TRIM5a hu were present before purification. B Ube2N depletion prevents accumulation of HMW (Ub) n -TRIM5a species. IB of PD products on Ni 2+ affinity matrices from lysates of TE671 cells expressing HA-TRIM5a hu and the designated combinations of 6×His-Ub and a Ube2N-specific shRNA. Note that some unmodified TRIM5a hu associated non-specifically with the Ni 2+ affinity matrix in all cases. C Affinity-purified (Ub) n -TRIM5a hu contains Lys63 linkages. IB of PD products from lysates of 293T cells expressing the designated combinations of 6×His-TRIM5a hu and HA-Ub. Ubiquitinated products were detected with an antibody specific for Lys63-linked Ub chains (denoted K63-Ub). Input levels of 6×His-TRIM5a hu (His) and b-actin (loading control) are shown below. Asterisk denotes a non-specific band in the input fraction. D, E K63R mutant Ub expression rescues TRIM5a-restricted N-MLV DNA synthesis but not infectivity. TE671 cells that stably expressed WT or K63R 6×His-Ub-GFP, or parental TE671  The EMBO Journal Ube2W primes TRIM5a autoubiquitination Adam J Fletcher et al Ub onto TRIM5a in cells, and this activity can be recapitulated in vitro using pure recombinant proteins.

Ube2N/Ube2V2 builds Lys63 polyUb chains on Ub-TRIM5a in vitro
Previous studies have shown that Ube2N/Ube2V2 and Ube2N/ Ube2V1 can extend a polyUb chain from an anchored Ub that is attached to a substrate by a different E2, such as Ube2W (Christensen et al, 2007;Tatham et al, 2013). We therefore tested whether Ube2N-containing enzymes could use Ub-TRIM5a rh proteins as substrates for further ubiquitination. Ub-TRIM5a rh proteins were created in situ by incubating Ube2W with TRIM5a rh (Fig 6A, lane 2), and a panel of six different E2 enzymes was tested for the ability to use these Ub-TRIM5a rh proteins as substrates to create TRIM5a rh -anchored polyUb chains (Fig 6A  lanes 3-8). Ube2N/Ube2V2 and Ube2N/Ube2V1 enzymes both exhibited this activity ( Fig 6A, lanes 6 and 7, respectively)-note significant depletion of the band representing Ub-TRIM5a rh with concomitant appearance of TRIM5a rh -anchored polyUb-whereas none of the control E2 enzymes tested showed any observable activity ( Fig 6A, lanes 3-5 and 8). As expected, the TRIM5a rhlinked polyUb chains synthesized by Ube2N/Ube2V2 contained K63 linkages, as analyzed both by testing the activities of single K/R Ub mutants ( Fig 6B) and by trypsin digestion/mass spectrometry mapping experiments (Table 1). Furthermore, we observed a noticeable difference in the pattern of TRIM5a rh polyUb products produced by Ube2N/Ube2V2 in the presence versus absence of Ube2W (Supplementary Fig S6A, compare lane 3 to lane 4). In the absence of Ube2W, darker staining of the IB was observed in a region representing very high MW polyUb products, which likely represent longer chains. In contrast, the presence of Ube2W led to more intense staining over a broader region of the IB, indicating an overall increase in the number of Ub linkages. This overall increase was confirmed by quantifying the overall staining intensity of ubiquitination products (lane 3 intensity was 354, lane 4 intensity was 954). The increase in K63 linkages was also supported by an eight-fold increase in mass spectrometric spectral counts (Table 1).
Collectively, these results are again in excellent agreement with the patterns of TRIM5a ubiquitination seen in cells, where Ube2N/ Ube2V2 adds K63-Ub, but can only do this when Ube2W is present (Figs 1 and 3). We therefore conclude that Ube2W and Ube2N/ Ube2V2 can modify TRIM5a in a sequential reaction in which Ube2W first adds a single Ub molecule, and Ube2N/Ube2V2 then builds K63-Ub onto the anchoring Ub molecules.

Ube2W transfers ubiquitin on to internal TRIM5a lysine residues
Mass spectrometric analyses were performed to map the site(s) of TRIM5a rh autoubiquitination. In this case, the Ube2W/TRIM5a rh ubiquitination reaction was performed on a preparative scale to drive the reaction to completion and to produce sufficient quantities of products for analysis. Mono-and di-ubiquitinated TRIM5a rh products were purified by SDS-PAGE (Supplementary Fig S7A, lanes 4 and 5), and digested with trypsin, and ubiquitinated tryptic peptides were then identified by LC/MS/MS based upon the presence of remnant isopeptide-linked Gly-Gly dipeptides (Peng et al, 2003;Flick et al, 2004). This analysis identified sites of TRIM5a rh ubiquitination on Lys residues 45 and 50 ( Supplementary Fig S7B; D.E. Christensen, C. Nelson, W.I. Sundquist, unpublished observations). Lys45 and Lys50 are located within the RING domain, near the expected Ube2W binding site. As expected, peptides corresponding to the acetylated protein N-terminus were detected, and no other free or modified N-terminal peptides were detected. A peptide corresponding to K48-linked diUb was also detected in these reaction mixtures, but this product may arise owing to the high levels of ubiquitinating enzymes used in this preparative reaction because this product was not detected in other Ube2W ubiquitination reactions where lower enzyme levels were used (see Table 1).
The observation that Ube2W could add isopeptide-linked monoUb modifications to internal TRIM5a rh Lys residues was noteworthy because Ube2W has previously been reported to preferentially couple Ub to the free aN-termini of proteins via conventional peptide bonds (Scaglione et al, 2013;Tatham et al, 2013). We therefore tested the idea that the TRIM5a rh aN-acetyl group was blocking Ube2W-mediated N-terminal ubiquitination, using a different recombinant TRIM5a rh protein that had a free, unacetylated N-terminus (denoted *TRIM5a rh ). *TRIM5a rh was obtained by expressing an N-terminally tagged FOS-TRIM5a rh protein in insect cells, purifying the protein by Strep-Tactin affinity chromatography, and removing the N-terminal affinity tag with PreScission protease to produce a protein with a free aN-terminus on the resulting non-native, N-terminal Gly-Pro extension. Ube2W added 1-3 Ub molecules onto *TRIM5a rh in preparative in vitro ubiquitination assays, and in this case, the di-ubiquitinated products were the most prevalent species (Supplementary Fig S7A, lanes 2 and 3). Thus, Ube2W added approximately one additional Ub molecule to the non-acetylated *TRIM5a rh protein than to the acetylated TRIM5a rh protein. In this case, mass spectroscopic mapping experiments detected peptides that corresponded to aN-terminal ubiquitination products, as well as isopeptide-linked Ub molecules at internal Lys residues 45, 50, and less frequently to Lys residues 85, 218, 284, and 372 (see Supplementary Fig S7B, caption). These experiments demonstrate that Ube2W can conjugate Ub to the free N-terminus of the artificial *TRIM5a rh construct. However, when the N-terminus is blocked by acetylation, as is the case in cells, Ub can instead be linked to internal Lys residues.
To validate the identity of the TRIM5a rh Lys residues ubiquitinated in the in vitro assay, we prepared mutant TRIM5a rh protein with lysine to arginine substitutions at the two major sites of ubiquitination located in the RING domain (Lys45, Lys50) and at the adjacent Lys44 (TRIM5a rh 3KR). TRIM5a rh 3KR displayed only a weak ability to accept Ub from Ube2W (Fig 7A, compare lanes 2 and 6). Moreover, the HMW TRIM5a rh smear produced by incubation with Ube2W and Ube2N/Ube2V2 was much lighter for the TRIM5a rh mutant presumably because there is less monoubiquitinated TRIM5a rh to act as a substrate for polyubiquitination (Fig 7A,   1   compare lanes 4 and 8). Although the K44R, K45R, and K50R mutations are found within the RING domain which is responsible for E2 binding, detection of Ub on the IB revealed that both TRIM5a rh mutants produced unanchored HMW Ub species on incubation with Ube2N/Ube2V2 at similar levels to WT TRIM5a rh , suggesting the mutations did not negatively affect E2 binding (Fig 7B, compare lanes 3, 7, and 11). Therefore, while mutation of target lysine residues reduces the specific auto-ubiquitination of TRIM5a rh in vitro, it does not prevent formation of Ub chains. Furthermore, these results confirm our observation that TRIM5a rh residues Lys45 and Lys50 are the primary sites of Ub attachment by Ube2W in vitro and that other Lys residues are only weakly ubiquitinated by Ube2W, as evidenced by the weak band corresponding to mono-ubiquitination of TRIM5a rh 3KR (which is then lost on addition of Ube2N/Ube2V2). Interestingly, Ube2W appears to enhance the polyubiquitination activity of Ube2N/ Ube2V2 in a TRIM5a rh -dependent manner (Fig 7B, compare lanes  4, 8, 12, and 14). This enhanced polyubiquitination activity by Ube2W is likely the result of aberrant transfer of Ub to the Ube2W N-terminus or lysines, as is observed in the E2 charging assay (Supplementary Fig S4B, lane 2, band  ‡), which can then function as the anchoring Ub for polyUb synthesis by Ube2N/Ube2V2. We then asked whether mutation at these specific lysine residues was able to similarly diminish TRIM5a ubiquitination in cells. As A TRIM5a rh ubiquitination reactions used the same conditions as those shown in Fig 5, but with multiple E2 enzymes. Ube2W (lanes 2-8) was used to attach anchoring monoUb molecules to TRIM5a rh in situ, and the indicated E2 enzymes were tested for polyUb chain elongation from the anchored Ub of Ub-TRIM5a rh . B Mutant Ub proteins with the designated Lys to Arg mutations were used to identify the residue(s) used for polyUb chain extension. K0 Ub has all seven lysine residues mutated to arginine.
before, we used TRIM5a hu restriction assays to readout restriction of infection and viral DNA synthesis as a measure of ubiquitination and proteasome recruitment. We prepared a TRIM5a hu mutant construct (TRIM5a hu 3KR), with an authentic N-terminus, bearing lysine to arginine substitutions at positions 44, 45, and 50 and tested its ability to restrict N-MLV infection and DNA synthesis upon expression in feline CRFK cells. Interestingly, TRIM5a hu 3KR was able to restrict N-MLV DNA synthesis and infection as effectively as WT TRIM5a hu ( Supplementary Fig S8A and B). Treatment of TRIM5a hu -expressing CrFK cells with MG132 also revealed robust HMW modification of the TRIM5a hu mutant ( Supplementary  Fig S8C), suggesting that additional TRIM5a lysine residues are able to accept Ub in cells. In support of this, we could specifically purify ubiquitinated species of TRIM5a hu 3KR from cells, at levels comparable to WT protein (Supplementary Fig S8D). Therefore, mutation of these three lysine residues was unable to preclude TRIM5a hu ubiquitination in cells, suggesting that ubiquitination of TRIM5 hu at positions additional to these is able to support TRIM5a-mediated inhibition of viral replication in cells. We therefore examined a second pair of TRIM5a mutants. TRIM5a rh 8KR (K44R, K45R, K50R, K85R, K218R, K284R, K371R, K372R) is mutated at all of the Lys residues found to be ubiquitinated by MS analysis as well as at lysine residues 44 and 371, selected for mutation based on close proximity to lysine residues 45 and 372. In vitro, TRIM5a rh 8KR was very substantially impaired in auto-ubiquitination with Ube2W ( Fig 7A, compare lanes 2 and 10,  and 4 and 12). Importantly, as for the WT protein, the small amount of monoubiquitination seen on TRIM5a rh 8KR (Fig 7A, lanes 9 and  10) was lost on addition of Ube2N/Ube2V2 and converted a faint HMW smear, consistent with polyubiquitination of the small The EMBO Journal Ube2W primes TRIM5a autoubiquitination Adam J Fletcher et al amount of monoubiquitinated TRIM5a rh 8KR. We also prepared an equivalent mutant in TRIM5a hu (TRIM5a hu 7KR). This protein has lysine to arginine substitutions at positions 44,45,50,84,282,367,368. Note that TRIM5a rh K218 is not conserved in the human orthologue, and TRIM5a rh K85, K284, K371, and K372 are equivalent to TRIM5a hu residues K84, K282, K367, and K368, respectively. We did not expect TRIM5a hu 7KR to restrict infection when expressed in cells because TRIM5a hu residues Lys367 and Lys368 are located in the surface-exposed PRYSPRY loop 5 ( Supplementary Fig S8E) (Biris et al, 2012), and their simultaneous mutation can impair restriction activity by decreasing binding to N-MLV capsid protein (Sebastian et al, 2009). Consistent with this expectation, TRIM5a hu 7KR was unable to restrict either N-MLV DNA synthesis or infection (Supplementary Fig S8A and B). However, TRIM5a hu 7KR still produced HMW species upon MG132 treatment ( Supplementary Fig S8C). Similarly, ubiquitinated TRIM5a hu 7KR species could be purified from cells, at levels similar to WT protein, even in the absence of proteasome inhibitor (Supplementary Fig S8D). We conclude that TRIM5a hu 7KR is probably inactive owing to mutations in the PRYSPRY domain. Nevertheless, this mutant reveals that even when all of the identified in vitro ubiquitination sites are mutated, TRIM5a hu 7KR can be ubiquitinated at additional sites in cells.

Discussion
In this study, we demonstrate that the anti-retroviral restriction factor TRIM5a is labeled with Ub in cells in a Ube2W-, Ube2N-, UbeV2-, and K63-Ub-dependent manner. Depletion of Ube2W abolished all conjugation of Ub to cellular TRIM5a, whereas depletion of Ube2N or Ube2V2 abolished poly-but not monoubiquitination (Fig 3). Reconstitution of ubiquitination reactions in vitro using purified full-length TRIM5a revealed that TRIM5a efficiently catalyzed auto-monoubiquitination using Ube2W (Fig 5) and that this modification was converted into anchored K63-Ub by addition of Ube2N/Ube2V2 (Fig 6). Our data add TRIM5a to the growing list of E3 Ub ligases shown to catalyze Ube2W-dependent monoubiquitination in vitro (Christensen et al, 2007;Alpi et al, 2008;Scaglione et al, 2011;Zhang et al, 2011;Tatham et al, 2013;Vittal et al, 2015). The cooperation between Ube2W and Ube2N/Ube2V2 has also been observed in vitro (Christensen et al, 2007;Tatham et al, 2013). To the best of our knowledge, however, TRIM5a represents the first instance in which the sequential ubiquitination mechanism has been dissected in cells using assays for protein function. In experiments where the relevant E2 enzyme was depleted, or Ub K63R was expressed, we observed that TRIM5a ubiquitination activity correlated with its restriction of retroviral DNA synthesis by reverse transcription, but not with its inhibition of viral infection. Interestingly, inhibition of proteasomes with small molecules Wu et al, 2013) also rescues restricted reverse transcription but not infection. Our data therefore suggest that the role of TRIM5a ubiquitination is not to prevent viral infection per se, but to block viral reverse transcription, through a mechanism that also requires proteasome activity. This model is consistent with those in which viral infectivity is suppressed when TRIM5a forms a complementary super lattice on the viral capsid and prevents progression to proper uncoating and nuclear entry Wu et al, 2006; Ganser-Pornillos et al, 2011); however, our data cannot formally exclude a role for viral capsid ubiquitination in the TRIM5a restriction mechanism. Pertel et al (2011) have postulated that the role of TRIM5a Ub ligase activity is to stimulate an innate immune response through the synthesis of unanchored K63-Ub and that this activity is necessary for TRIM5a to restrict viral infection. While we do not find a requirement for ubiquitination activity in TRIM5a restriction of viral infection (Figs 1 and 4), our data are compatible with a model in which TRIM5a synthesizes unanchored polyUb, as we too observed unanchored polyUb chain synthesis in vitro using recombinant TRIM5a and either Ube2N/Ube2V2 or Ube2N/Ube2V1 (Fig 5). However, our findings that: (i) Ube2W was absolutely required for restriction of viral DNA synthesis, and not infection, and (ii) that Ube2W causes the K63-Ub to be anchored to TRIM5a itself suggest a novel mechanism whereby anchored, and not free, K63-Ub is the necessary signal to trigger downstream effector mechanisms that result in a block to viral reverse transcription in a proteasomedependent manner. We also note that addition of Ube2W to reactions between TRIM5a and Ube2N/Ube2V2 results in a greater abundance of polyUb, and the produced chains are also shorter ( Supplementary Fig S6). How K63-Ub could underlie proteasome recruitment remains unclear, although we envisage either a mechanism of Ub chain branching whereby the introduction of additional Ub chain linkages confers additional function (Nakasone et al, 2013) or via the recruitment of K63-Ub-specific Ub binding domains that facilitate transfer of ubiquitinated TRIM5a, bound to a viral capsid, to the proteasome. A potential role for autophagy in TRIM5a antiviral activity was recently described (Mandell et al, 2014), lending support to the notion that additional cofactors of TRIM5a remain to be discovered. Moreover, an alternative autophagic degradation pathway could explain how K63-Ub induces degradation, with the effects of proteasome inhibition explained by concomitant disturbances in Ub homeostasis.
Mass spectrometry mapping experiments have helped us understand how Ube2W mediates TRIM5a polyUb "anchoring" to TRIM5a. Analyses of trypsin-digested TRIM5a expressed in cells revealed that the N-terminus of TRIM5a is quantitatively acetylated -a modification that blocks N-terminal ubiquitination. In vitro ubiquitination of aN-acetylated-TRIM5a followed by mass spectrometry analyses demonstrated that Ube2W can catalyze monoubiquitination of wild-type TRIM5a at multiple lysine residues, predominantly Lys45 and Lys50 in the TRIM5a RING domain. In lower abundance, we also detected ubiquitination of lysines 85, 218, 284, and 372. While aN-acetylated N-terminal peptides were detected in our mass spectrometry experiments, no N-terminally ubiquitinated peptides were detected (Supplementary Fig S2). Ube2W is reported to target protein N-termini preferentially (Scaglione et al, 2013;Tatham et al, 2013;Vittal et al, 2015) and, indeed, we found that Ube2W could add Ub to the N-terminus when this site was (artificially) unblocked. Thus, our data show that although Ube2W can target protein N-termini, it can also target internal lysine residues, as is the case for native, WT TRIM5a rh .
We were able to validate the importance of these six (and two adjacent) Lys residues in vitro, as TRIM5a rh mutated at eight lysine residues was a poor substrate for Ube2W-mediated monoubiquitination. However, mutation of the corresponding seven lysine residues in TRIM5a hu did not substantially diminish TRIM5a hu ubiquitination in cells, suggesting that additional lysine The EMBO Journal residues can be ubiquitinated, at least when the primary targets are mutated. Similarly, while TRIM5a rh mutated at lysine residues 44, 45, and 50 was also impaired for in vitro ubiquitination with Ube2W, these mutations did not diminish TRIM5a hu ubiquitination or restriction activity in cells. One possible explanation for this discrepancy is that TRIM5a is not itself the functional target of the K63-Ub that is essential for restriction of viral DNA synthesis. However, the loss of ubiquitinated TRIM5a species after either E2 enzyme depletion, or Ub K63R expression in cells, correlates with the loss of restriction of viral DNA synthesis. This observation supports the notion that TRIM5a itself is the functional target of ubiquitination (Figs 1, 3 and 4). Another possible explanation is that we have failed to mutate all of the possible target lysines in TRIM5a.
Our mass spectrometry demonstrated that TRIM5a ubiquitination is not highly specific. In vitro, the RING domain lysines at positions 45 and 50 are the major targets for autoubiquitination, possibly because these lysines are the closest to the E2 binding site in the RING domain. However, conformations adopted by TRIM5a assemblies in cells, perhaps aided by additional cofactors, might orientate the protein such that additional lysines become favored ubiquitination targets. We observed a hint of Ube2W-catalyzed monoubiquitinated TRIM5a rh bearing three or eight lysine-arginine substitutions in vitro, which was lost on Ube2N/Ube2V2 addition (Fig 7A,  compare lanes 6 and 7), suggesting that in the absence of these amine groups, TRIM5a might ubiquitinate lysines more distant to the E2 binding site, albeit less efficiently in our in vitro reactions. Finally, it is possible that other E3 ligases functionally ubiquitinate TRIM5a via the E2 enzymes identified here. However, our observation that TRIM5a is able to autoubiquitinate in vitro using these enzymes makes TRIM5a itself the most likely E3 candidate for TRIM5a-dependent restriction. Although our data do not prove that TRIM5a is the functional target for Ub-dependent restriction, we favor a model in which TRIM5a autoubiquitination is necessary for subsequent inhibition of viral DNA synthesis. In principle, TRIM5a-anchored Ub molecules offer the advantage that they would presumably remain associated with the incoming viral capsid and thus could "mark" that capsid for inactivation. Furthermore, redundancy of target lysines for ubiquitination is consistent with a protein whose effective function depends on combating rapidly evolving pathogens by evolving changes in antiviral specificity. Plasticity in the lysines used as ubiquitination targets would reduce the likelihood that TRIM5a sequence changes would impact its ability to autoubiquitinate. Redundancy of target lysines may therefore be an advantageous, evolved property of the TRIM5a protein.
In summary, we have identified a novel series of reactions catalyzed by TRIM5a that result in the anchoring of K63-Ub to TRIM5a itself. Perturbation of the enzymatic activities involved in the formation of this modification prevents TRIM5a from restricting retroviral infection prior to viral reverse transcription. This hitherto unidentified mechanism appears to lie in the same pathway as proteasomedependent disassembly of retroviral capsids. Also, the observed K63-Ub specificity suggests that a relationship may exist between TRIM5a restriction of retroviral reverse transcription and stimulation of innate immune signaling cascades. Our study highlights the value of combining in vitro assays for ubiquitination with cellular assays for Ub ligase function. The ability to measure TRIM5a ubiquitination activity functionally in cells, using different steps in the retroviral lifecycle as the readout, makes this an attractive approach for further studies of TRIM protein-mediated ubiquitination and for analyzing the contribution of K63-Ub to proteasome recruitment and innate immune signaling.

RNAi-mediated E2 ubiquitin-conjugating enzyme library screen
The siRNA library screen for 38 human E2 Ub-conjugating enzymes constituted 38 siGENOME SMARTpools (Dharmacon), one pool per well in a 96-well plate. Pools were resuspended in Optimem and Oligofectamine. 8 × 10 3 HeLa cells were added to each pool, with siRNA concentrations of 30 nM in final volumes of 100 ll. The cells were incubated at 37°C, washed after 24 h, and incubated for a further 24 h. Cells were then infected with N-tropic or B-tropic MLV-GFP vectors (MOI~0.2) in a final volume of 200 ll.
shRNA-mediated depletion of Ube2W, Ube2N, and Ube2V2 shRNA-encoding oligonucleotides were designed, annealed, and ligated into pSIREN Retro Q (Clontech) according to manufacturer's instructions (http://www.clontech.com/GB/Support/Online_Tools). The 19mer target sequences are listed in Supplementary Table S1. The shRNA sequence designed to deplete Ube2V2 corresponded to the most effective 19mer target sequence from a screen of four independent siRNA oligonucleotides. The Ube2N 19mer sequence was previously described (Duncan et al, 2006). The Ube2W 19mer sequence targeting the 3 0 UTR was designed according to Clontech instructions. The Ube2V1 22mer sequence was previously described (Pertel et al, 2011). TE671 cells were transduced at an MOI~1 and selected in 2.5 lg/ml puromycin (Fletcher et al, 2010).
Infectivity and quantitative polymerase chain reaction (qPCR) assays 10 5 TE671 cells per well were seeded in 12-well plates and transduced 24 h later (MOI~0.2) in quadruplicate with RQ1 DNasetreated (30 U/ml, 2 h, 37°C, in buffer containing 400 mM Tris (pH 8.0), 100 mM MgSO 4 , and 10 mM CaCl 2 ) N-MLV or B-MLV GFPexpressing viral vectors, in the presence of 8 lg/ml Polybrene. Heat-treated viral supernatants (95°C, 5 min) served as controls for plasmid contamination. 6 h p.i., duplicate wells were washed in PBS, trypsinized, pelleted, washed again in PBS, pelleted, and stored at À80°C. Total DNA was extracted using QIAamp DNA Mini Kit (Qiagen), eluting DNA in 50 ll water. Parallel duplicate wells were scored for GFP expression by flow cytometry (BD Accuri C6). Concentration of each DNA sample was quantified using a Nano-Drop spectrophotometer ND-1000 (Thermo Fisher Scientific). TaqMan qPCR reactions contained 5 ll of eluted DNA, 300 nM forward and reverse primers (listed in Supplementary Table S1), 150 nM probe, 20 lg salmon sperm DNA, and 1× TaqMan Universal Master Mix II (Life Technologies). Standard curves generated from GFP cDNA plasmids were created in duplicate for each qPCR assay. Cycle parameters for GFP gene detection were 50°C for 2 min and 95°C for 10 min, followed by 40 cycles of 95°C for 15 s, and 60°C for 1 min using a Prism 7000 thermocycler (Applied Biosystems). SYBR Green qPCR assays were used to detect transcript abundance of E2s, using 1 lM primers as listed in Supplementary Table S1, and 200 ng cDNA. Cycle parameters were 5 min at 95°C, followed by 40 cycles of 95°C for 10 s and 60°C for 30 s. Standard curves generated from E2 cDNA plasmids were created in duplicate for each qPCR assay. The specificity of E2-specific primer pairs was assessed in control PCRs (Supplementary Fig S1).
cDNA synthesis for quantitation of Ube2W, Ube2V2, and Ube2V1 depletion Total RNA was purified using the RNeasy RNA Extraction Kit according to manufacturer's instructions (Qiagen). 1 lg RNA was mixed with 42 ng oligo-dT primer and 0.8 mM dNTPs, heated at 65°C for 5 min, and then chilled on ice. The mixture was then supplemented to a final concentration of 250 mM Tris (pH 8.3), 375 mM KCl, 15 mM MgCl 2 , and 10 mM DTT and incubated at 42°C for 2 min. 200 units of SuperScriptTM II reverse transcriptase were then added and the mixture incubated at 42°C for 50 min. The reaction was inactivated by heating at 70°C for 15 min. RNA was degraded by addition of two units E. coli RNase H (Life Technologies) and incubated at 37°C for 20 min. cDNA was stored at À20°C.
Lysates were sonicated for 15 s and rotated for 3 h at room temperature with 50 ll equilibrated NiNTA agarose (Qiagen). The agarose matrix was washed twice with 1 ml lysis buffer, twice with 1 ml 3:1 wash buffer:lysis buffer, once with 1 ml wash buffer (25 mM Tris, 20 mM imidazole, pH 6.8), resuspended in 2× SDS sample buffer supplemented with 300 mM imidazole to elute bound His-tagged proteins, and 10% b-mercaptoethanol as a reducing agent, and heated for 10 min at 95°C prior to SDS-PAGE.

Expression and purification of recombinant proteins
Recombinant Ub, UBA1, and E2 proteins were expressed in Rosetta (DE3) pLysS bacteria (EMD Millipore) grown in LB media and were purified to homogeneity. Briefly, human UBA1, Ube2B, Ube2C, Ube2D3, Ube2H, Ube2L3, Ube2N, Ube2V1, Ube2V2, and Ube2W were purified by affinity chromatography on Ni-NTA agarose (Qiagen) and eluted with 0.5 M imidazole in buffer containing 50 mM HEPES (pH 7.0) and 300 mM NaCl. The eluted E2 proteins were treated further with precision protease, and the solutions were incubated at 4°C overnight to remove the 6×His tags (leaving nonnative Gly-Pro dipeptides at the N-termini). The UBA1 and E2 enzymes were then purified further by size exclusion chromatography in 50 mM HEPES pH 7.2 and 100 mM NaCl. Ub was purified as described previously (Pickart & Raasi, 2005).
Pure rhesus TRIM5a (TRIM5a rh ) protein was a generous gift from Vish Chandrasekaran (Li et al, manuscript in preparation). Briefly, Flag-One-Strep (FOS)-tagged TRIM5a rh was expressed from a baculoviral vector in SF21 insect cells, purified by affinity chromatography on streptactin agarose, treated with precision protease to remove the FOS tag, and then purified to homogeneity by anion and gel filtration chromatography. The protein corresponds to wild-type TRIM5a rh except that it has a non-native LEVLFQ peptide at the C-terminus.

Mass spectrometry
ESI/MS analysis of TRIM5a as an intact protein Protein samples were purified using a C18 Ziptip TM (Millipore) to remove salts and other small molecule contaminants. The protein was loaded onto a Ziptip TM in 5% acetonitrile and then washed with 200-500 ll of 5% acetonitrile with 1% formic acid. Proteins were eluted from the Ziptip TM and collected in a pre-rinsed microcentrifuge tube with three consecutive 0.75-ll aliquots of 60% acetonitrile with 1% formic acid, then one 1-ll aliquot of 98% acetonitrile with 2% formic acid was added, and then 1 ll of 5% formic acid was added prior to ESI/MS analysis.
ESI/MS analysis of intact TRIM5a was performed using a Q-ToF-2 mass spectrometer (Waters). 3 ll of eluent was infused into the instrument with using a nanochip autosampler (Triversa Nano-Mate TM , Advion) with nanoflow static infusion at a rate of~100 nl/min. Electrospray ionization was used in the positive-ion mode, but with the ion-source spray voltage of the Q-ToF-2 instrument set to zero. Ionization spray voltage was applied with the Triversa Nano-Mate at 1.51 kV at the nozzle of the nanochip device. Spectra were acquired with a cone voltage of 45 eV, and the instrument was scanned from 600 to 1,400 m/z in 2 s. Scans were accumulated for about 1 min. Spectra were combined, and multiply-charged molecular ions were deconvoluted into a molecular-mass spectrum processed into the average-isotope neutral molecular weight using MaxEnt TM software (Waters).

LC/MS/MS analysis of TRIM5a and ubiquitin
TRIM5a was analyzed by LC/MS/MS following tryptic digest of gel bands and also from solution samples. Peptides were analyzed using a nano-LC/MS/MS system comprised of a nano-LC pump (2D-ultra, Eksigent) and a LTQ-FT mass spectrometer (ThermoElectron Corporation, San Jose, CA). The LTQ-FT is a hybrid mass spectrometer in which the linear-ion trap was used for MS/MS sequencing of peptides (collision-induced dissociation (CID) fragmentation with helium collision gas), and the Fourier transform ion-cyclotron resonance (FT-ICR) part of the instrument was used to acquire MS data with high mass accuracy and high resolution. All TRIM5a peptides identified were within 3 ppm mass error compared to theoretical values, but most peptides were within 2 ppm mass error. The LTQ-FT was equipped with a nanospray ion source (ThermoElectron) at 2.3-kV spray voltage. The LTQ-FT mass spectrometer was operated in data-dependent acquisition mode controlled by Xcalibur 1.4 software, in which the "top 10" most intense peaks observed in an FT primary scan (i.e., MS survey spectrum) are subsequently trapped for MS/MS fragmentation and analysis in the LTQ linear-ion trap part of the instrument. Spectra in the FT-ICR were acquired from m/z 350 to 1,400 at 50,000 resolution at a rate of about three full scans per second. The LTQ linear-ion trap was operated with the following parameters: precursor activation time 30 ms and activation Q at 0.25; collision energy was set at 35%; dynamic exclusion width was set at low mass of 0.1 Da and high mass at 2.1 Da with one repeat count and duration of 10 s.

Nano-LC
Approximately 5-20 fmoles of tryptic digest was dissolved in 5% acetonitrile with 0.1% formic acid and injected onto a C18 nanobore LC column for nano-LC/MS/MS and identification of peptides. The nanobore column was packed and assembled homemade [C18 (Atlantis, Waters Corp); 3-lm particle; column: 75 lm ID × 100 mm length; unpacked column from New Objective]. A linear gradient LC profile was used to separate and elute peptides with a constant total flow rate of 350 nl/min. The gradient consisted of 5-80% solvent B in 78 min (solvent B: 80% acetonitrile with 0.1% formic acid; solvent A: 5% acetonitrile with 0.1% formic acid).
In-gel digest of proteins SDS-PAGE gel slices were destained in 50% methanol, sliced into small pieces, and dehydrated in acetonitrile, and then the acetonitrile removed prior to digestion. The gel pieces were digested with TPCK-modified sequencing grade trypsin (Promega). 10 ll of trypsin (i.e. 20 ng/ll in 50 mM ammonium bicarbonate) was added and 20 more ll of 50 mM ammonium bicarbonate was added to the gel pieces. Digestion was allowed to continue for 2 h (at 37°C), and then two more aliquots of 20 ll of trypsin was added at 2-h intervals and then allow to continue digesting overnight. This relatively aggressive digest protocol was needed to improve the yield of peptides from Ub and TRIM5a proteins. The digestion was quenched by the addition of 20 ll 1% formic acid. This solution was allowed to stand and peptides that dissolved in the 1% formic solution were extracted and collected. Further extraction of peptides from the gel material was performed twice by the addition of 50% acetonitrile with 1% formic acid and sonicated at 37°C for 20 min, and these solutions were also collected and combined. A final complete dehydration of the gel pieces was accomplished by addition of 20 ll of 100% acetonitrile and incubation at 37°C for 20 min. The combined supernatant solutions of extracted peptides were combined and dried in a vacuum centrifuge (Speed-Vac). The peptides were reconstituted in 100 ll of 5% acetonitrile with 0.1% formic acid for LC/MS/MS analysis. After further dilution, 2% of the total digest volume was injected on the nano-LC column for LC/MS/MS analysis in a 5 ll volume.
Digests of proteins in solution A 15-ll aliquot of~1-10 pmoles of protein in solution was digested with trypsin. Proteins in solution were digested with TPCK-modified trypsin (Promega) following the trypsin protocol outlined above. 6% of the total digest volume was injected on the nano-LC column for LC/MS/MS analysis in a 5 ll volume. All solution samples within a set of samples were analyzed identically and in a quantitative manner (i.e. sample volume, digest protocol, injection amount, and LC/MS/MS analysis).

Protein and peptide identifications
Database searches All identified proteins and peptides from protein digest samples were assigned from protein database searches of the LC/MS/MS data, using in-house processing with MASCOT search engine (in-house licensed, ver. 2.2.7, Matrix Science, Inc.). Mascot searching was performed from an in-house computer, in which NCBI and "custom" protein databases were searched. Searches were also performed using Proteome Discoverer (Thermo, version 1.4), in which Sequest and Mascot searched were conducted. Peptides were searched assuming the following criteria: 1 Trypsin-specific cut sites, allowing for two missed cleavages. 2 Accurate mass measurement of peptide molecular ions by FTMS with search window 5 ppm (but peptides typically will have < 2 ppm mass error). Molecular ions with +1, +2, or +3 charge states were considered. 3 Peptide sequence information from MS/MS; CID fragmentation of the parent ion of each peptide was obtained in the linear-ion trap region of the LTQ-FT instrument. 0.5 Da mass error tolerance was allowed for peptide fragment ion masses in the search (but MS/MS fragment ions typically have errors < 0.3 Da). 4 Mass data peak lists for the Mascot searches were generated using Sequest in Qual Browser software (Bioworks Browser 3.2, Excalibur, Thermo) or with Proteome Discoverer. 5 Peptide modifications were included in the search as variables (e.g. oxidation on methionine, histidine, or tryptophan; ubiquitination on lysine and N-terminus amino acid; acetylation on N-terminus amino acid).
Supplementary information for this article is available online: http://emboj.embopress.org