Identification of HTLV - human protein interactions
To identify retroviral PPIs with the human proteome we adapted our well-established HT-Y2H system [12
]. Using Gateway-based ORFeome libraries encoding HTLV-1 and HTLV-2 proteins (HTLV-1 Gag, Pol, Rex, Tax, Env, p12, p13, p30 and HTLV-2 Gag, Pol, Rex2, Tax2, Env and APH-2 - Additional file 1
: Table S1) in a Y2H screen against the ~12,000 proteins expressed from Human ORFeome v3.1 [17
], we identified 1028 diploid colonies representing 286 potential interactions between human proteins and HTLV viral proteins. These interactions were independently confirmed by pairwise Y2H retesting [12
HTLV structural and regulatory proteins have significant sequence or functional similarity (for example HTLV-1 Tax and HTLV-2 Tax share 77% of sequence similarity, and both are transcriptional activators of viral expression). These homologous viral proteins might share one or more interacting partners amongst the human proteins, interactions that were not identified in initial screens because (i) highly overlapping or similar viral ORFs may be misidentified with BLAST, and (ii) interactions can be missed in a single screen [12
]. We retested all homologous HTLV proteins for interaction with each human protein found in our initial screen with at least one homologous viral protein. For instance, all human proteins identified as HTLV-1 Tax interactors were also retested against HTLV-1 and HTLV-2 Tax and Rex proteins (Additional file 1
: Table S1). This strategy combines the advantages of pooling [14
] with individual testing, to reduce the cost and workload of the initial screen while keeping the ability to differentiate similar proteins, overcome sensitivity and specificity issues and permits comparison of negative results. The final data set contained 166 interactions between 10 viral proteins and 122 human proteins (Figure and Additional file 1
: Table S2). Among the 166 PPIs identified 87 and 79 interactions involved HTLV-1 and HTLV-2 -encoded proteins, respectively. Twenty-eight out of the one hundred and twenty-two human proteins were found to interact with both viruses (Figure ).
Figure 1 Pipeline of the HT-Y2H experiment. (A) Retroviral ORFeome screened against Human ORFeome 3.1 in both configurations (DB-hORF AD-rvORF and DB-rvORF AD-hORF). Interactions found in the primary screen were subjected to homologous individual retest, where (more ...)
In addition to applying stringent internal controls and retests, to eliminate artifacts of the assay [19
], we verified the quality of our HT-Y2H results by applying a binary interactome evaluation [12
]. This evaluation employs independent protein-protein interaction assays to measure how any PPI dataset performs relative to a positive reference set (PRS) of high confidence manually curated interactions from the literature versus a random reference set (RRS) and position our dataset compared to these controls [12
]. We tested 158 Y2H-identified binary interactions by mammalian protein-protein interaction trap assay (MAPPIT) [20
]. MAPPIT is a forward mammalian two-hybrid strategy based on the activation of type I cytokine-signaling pathway. To perform a MAPPIT assay, we used as bait and prey, interacting partners fused to a STAT recruitment-deficient homodimeric cytokine receptor or to the C-terminal STAT3 recruitment portion of the gp130 receptor, respectively. Interactions between bait and prey proteins result in a functional cytokine receptor monitored by a STAT3-responsive promoter. The verification rate of our host-pathogen interactome data set by MAPPIT was 29% (40/137 testable pairs, Additional file 1
: Table S2), which compares favorably to PRS detection rates [18
]. As for other PPI assays tested so far, only a fraction of verifiable interactions detected by one PPI method will retest positive with another [18
]. Previous studies show that MAPPIT detects about 20%-25% of PRS pairs under conditions that minimize the detection of RRS pairs [18
]. As a control for specificity, a random set of 40 proteins from the human ORFeome 3.1 was also tested by MAPPIT for their interaction with HTLV proteins, and only 3 out of 40 (7.5%) were found positive. The MAPPIT retest rate of our HTLV-human PPIs represents ~80-100% of the maximum number of interactions expected to be recovered by MAPPIT, with an estimated false positive rate of 0-20% [12
Human proteins interacting with viral proteins apparently have significantly different topological properties compared to random proteins in the human PPI network [15
]. Viral proteins seem to preferentially target "hubs," i.e. highly connected proteins in the human-human PPI network. Preferential targeting of hubs is also observed in our HTLV network (Table ). Human targets of HTLV proteins have higher connectivity (degree k
= 13.75) compared to the whole network (k
= 3.79), are more centrally located as measured by higher betweenness centrality (BC), 12443 for viral targets vs 2208 for random proteins, and have lower characteristic path length (CPL), 3.09 for viral targets vs. 4.38 for random proteins.
Topological features of viral targets
Degree, Characteristic path length (CPL) and betweenness centrality (Betweenness) for the 131 human proteins identified in our screen (Viral Targets), the whole human PPI network (Whole Network), and for human proteins interacting with 19 random human proteins (Random Sources). P-values assess the difference between viral targets and the whole network
As previously demonstrated and again confirmed here, our Y2H methodology delivers high quality, reproducible biophysical interactions [12
], but there is no guarantee that biophysical interactions are functionally relevant in vivo
. To functionally validate our PPI dataset, we reasoned that some human proteins interacting with viral transactivators are likely to influence Tax transcriptional activities and thus contribute to viral replication and expression of cellular genes.
Many HTLV-human interactions in our data set (106/166) involved the retroviral transactivator proteins HTLV-1 Tax (57/166) or HTLV-2 Tax2 (49/166). To examine the functional consequences of these associations, HEK293T cells were cotransfected with expression vectors for Tax-1 and Tax-interacting proteins, together with a firefly luciferase reporter driven by the HTLV-1 LTR promoter. As determined by normalized luciferase reporter assays, we identified 31 proteins (37% of the 83 Tax-interacting proteins) that regulated HTLV-1 LTR promoter activation by Tax (Figure and Additional file 1
: Table S3). There were 8 host factors that significantly enhanced Tax transactivation activities suggesting their potential implication in viral replication and persistence in infected cells. Another group of 23 cellular proteins down-regulated HTLV-1 LTR viral promoter activation and as such may be implicated in the viral latency which allows viruses to escape immune surveillance (Figure and Additional file 1
: Table S3). We selected two Tax1-cellular partners, SPG21, involved in the repression of T cell activation [22
], and FANCG, a DNA damage response activated protein [23
], for further validation in a T lymphocyte cell line. We used Jurkat T cells harboring a HTLV-1 LTR luciferase reporter (Jurkat-LTR-Luc) to confirm potential roles of SPG21 and FANCG in viral replication. We transduced Jurkat-LTR-Luc cells with a control shRNA and three validated shRNAs directed against SPG21 or FANCG and measured luciferase reporter-expression and cell viability. In accordance with regulation of Tax-transactivation data (Figure and Additional file 1
: Table S3), knockdown of SPG21 increased HTLV-1 LTR promoter activity while depletion of FANCG decreased HTLV-1 LTR promoter activity (Figure ).
Figure 2 HTLV 1 and 2 virus-host Y2H PPI network. Big diamonds: viral ORFs, with HTLV-1 and HTLV-2 in blue and light blue, respectively. Small circles: Human ORFs, thin links: Y2H interactions; thick links: MAPPIT confirmed interactions. Human ORFs are colored (more ...)
Figure 3 Effect of SPG21 and FANCG knockdown on viral promoter activation. Jurkat-LTR-Luc cells were transduced with lentiviral particles expressing a control shRNA and three validated shRNAs targeting various sequences of the SPG21 and FANCG mRNAs. Cells were (more ...)
In summary, we identified 166 interactions between 10 viral proteins and 122 human proteins and verified their overall quality through an independent assay. We functionally validated our dataset by showing involvement of 31 human proteins in viral transcriptional regulation.
Analysis of the HTLV-1 and-2 interactome maps
Our standardized experimental conditions, which combine stringent, high-throughput Y2H for a defined search space with systematic retesting of all homologous proteins, permit comparisons between interacting protein pairs. Network views of our data identify shared and distinct PPIs between HTLV-1 and HTLV-2. (Figure ).
We found 34 human proteins that bind HTLV-1 Tax protein, but not the HTLV-2 Tax homolog (Figure and Additional file 1
: Table S4). Consistent with its intrinsically disordered conformation and pleiotropic activities [26
], specific HTLV-1 Tax interactors include proteins associated with a range of distinct cellular functions such as transcription regulation (ETV4, RFX4, MyEF2, ZNHIT4, ZMAT1 and HOXB9), cell apoptosis (TRIP6 and CRADD), protein degradation (WDFY3 and PSMA1), and microtubule cytoskeleton (KIF9, KRT6A and KTR8).
We also found 26 HTLV-2 Tax interactors that did not interact with HTLV-1 Tax, including cell cycle proteins (Cep70, MAD1L1 and SSX2IP), transcription factors (NFKB activating protein, ZBTB16 and SOX5) and proteins involved in the endosomal-lysosomal system (AP4M1 and GCC1) (Figure and Additional file 1
: Table S4). Considering the differential oncogenic potential of the two HTLV viruses [9
] and the central roles of their Tax proteins, these PPIs could shed light on mechanisms of cellular transformation by the Tax oncoprotein.
We have identified 10 novel HBZ binding proteins (Figure ) including the homeobox transcription factor HOXD3; two RNA binding proteins, PCBP1 involved in restricting viral infections [27
] and RNPS1, that can induce genomic instability when overexpressed [28
]. Consistent with its association with transcriptional repression, we also found that HBZ interacts with MYST2, a member of the largest family of histone acetyltransferase enzymes, implicated in the regulation of DNA synthesis [29
]. We also identified 8 novel APH-2 interactors (Figure and Additional file 1
: Table S2) including USF2, a member of the basic helix-loop-helix (bHLH) leucine zipper family of transcription factors that may play a role in late viral mRNA transcription [30
]; VPS37A, a subunit of the mammalian endosomal sorting complex ESCRT-1 that have been shown to play a role in HIV-1 budding [31
]; and NP54, a member of the nucleoporin complex that have been shown to bind HIV-1 Vpr and to play a critical role in the nucleocytoplasmic transport of viral preintegration complex [32
]. Interestingly, we did not find any common interactor between HBZ and APH-2. The functions of these new HBZ and APH-2 associations with cellular factors remain to be further characterized.
Enrichment of viral targets for biological pathways
The immediate human targets of HTLV proteins found here were not significantly enriched for annotated pathways in the Kyoto Encyclopedia of Genes and Genomes (KEGG) [35
], i.e. the number of proteins belonging to a specific pathways is not significantly higher than random expectation, probably because of the limited number of human targets. To improve sensitivity, we also analyzed second-degree interactors, those human proteins in the human-human PPI network [14
] that interact with human targets of viral proteins. Proteins associated with apoptotic pathways, Notch signaling, cell cycle, ubiquitin mediated proteolysis, as well as proteins involved in several human cancers including chronic myeloid leukemia, were overrepresented compared to random expectation (Table ).
KEGG pathways enriched in secondary viral interactors
For each enriched KEGG pathway is given the pathway identifier in the KEGG database (Pathway ID), the number of observed proteins belonging to the considered pathway (Observed), the number of proteins in the pathway expected at random (Random), the ratio between the number of observed proteins and the expected number (Odds Ratio), the false discovery rate (FDR), and the corrected FDR (FDR-Corr)
In an apoptotic pathway sub-network, KEGG analysis highlighted the tumor necrosis factor (TNF) receptor and the AKT/PI3K signaling pathways as potential targets for HTLV proteins. In this network HTLV Tax and Rex proteins are closely linked to the Akt/PI3K and mitochondrial apoptotic pathways. We identified interactions between HTLV Tax proteins and nitric oxide synthase 3 (NOS3), hepatocyte growth factor-regulated tyrosine kinase substrate (HGS), Ewing sarcoma breakpoint region 1 (EWSR1) and glucose transporter-4 (SLC2A4) proteins. KEGG analysis indicated that phosphatidylinositol-3-kinase (PI3K), BCL2-antagonist of cell death (Bad), and DNA fragmentation factor alpha (DFFA) proteins are second-degree targets of HTLV Tax proteins (Figure ). We also found that the HTLV Rex proteins interact with DLC2 (for dynein light chain 2), able to regulate cell death-inducing functions of pro-apoptotic proteins Bim (Bcl-2-interacting mediator of cell death) and Bmf (Bcl-2-modifying factor). HTLV Rex proteins are nuclear-localizing proteins well known to drive post-transcriptional export of viral mRNAs from the nucleus to the cytoplasm [36
]. Besides its interaction with the cellular export factor CRM1 [39
], functional relationship between Rex proteins and their cellular partners have not been fully investigated. Interaction between Rex proteins and DLC2 may shed light on a new role of Rex in the apoptotic pathway. To assess the subcellular localization of Rex1 and DLC2, we transfected HeLa cells with expression vectors for Rex1-GFP and Flag-tagged DLC2. Cells were stained by anti-flag antibody followed by Alexa546-conjugated secondary antibody and a far-red fluorescent DNA dye (DRAQ5) for nuclear staining. Consistent with previous reports [40
], DLC2 was found exclusively in the cytoplasm (Figure , DLC2); and Rex-GFP was localized in nucleolar foci (Figure , Rex1-GFP). Co-expression of Rex1-GFP and Flag-DLC2 provoked a change in the localization of DLC2 with two patterns being observed. DLC2 was localized in the cytoplasm as well as in nuclear foci (Figure , DLC2 + Rex1-GFP, Alexa546). It thus appeared that coexpression with Rex1 directs DLC2 in nucleolar foci as revealed by the good match of the green (Rex1-GFP) and orange (Flag-DLC2) fluorochromes. We conclude that HTLV Rex proteins might interfere with the anti-apoptotic activities of DLC2 in HTLV infected cells.
Figure 5 Targeting of apoptotic pathway by viral proteins. (A) Schematic representation of PPIs. Big diamonds: viral ORFs, with HTLV-1 and HTLV-2 in blue and light blue, respectively. Small circles: human ORFs with green representing membership of the apoptotic (more ...)
Figure 6 HTLV-1 Rex and DLC2 co-localize in nucleolar foci. (A) HeLa cells were transfected with expression vectors for Rex1-GFP and Flag-DLC2 as indicated. Twenty-four hours post-transfection, cells were labeled with anti-flag M2 mouse antibody followed by alexa546-conjugated (more ...)
We also identified TNF receptor-associated factor type 2 (TRAF-2) as a central protein mediating interactions between HTLV proteins, TNF receptor (TNFR) signaling, and the Akt/PI3K survival pathway (Figure ). We found that TRAF2 directly binds HTLV-2 Gag and is also a second-degree interactor of HTLV Tax and Rex proteins. Depending on its interacting partners, TRAF2 signals drive contradictory cellular responses. Direct binding to the cytoplasmic domain of TNFR2, which does not contain a death domain, can trigger NFκB and JNK activation, but TRAF2 also indirectly mediates the signal from a death domain containing receptors such as TNFR1 via interaction with FADD and TRADD pro-caspases adaptor factors [43
]. Retroviral infection is frequently associated with elevated TNFα, and cell lines derived from ATL patients show sensitivity to TNF-related apoptosis [44
]. Gag protein could target TRAF2 for proteasomal degradation, thereby facilitating sensitivity to TNFα-induced cell death. To investigate this possibility we co-expressed GFP tagged HTLV-2 Gag, Flag tagged TRAF2 and a Myc-Ubiquitin expressing vectors. The presence of HTLV-2 Gag reduced TRAF2 protein levels (Figure , αFlag compare lanes 1 and 2; and lanes 3 and 4), and degradation of TRAF2 correlated with a reduction of Myc-ubiquitylated proteins (Figure , αMyc compare lanes 3 and 4) suggesting that the TRAF2-E3 ubiquitin ligase activity was also affected by the presence of HTLV-2 Gag protein. The degradation of TRAF2 could be blocked by preincubating cells with proteasome inhibitor MG132 (Figure ). Together these data indicate that HTLV-2 Gag induces proteasomal degradation of TRAF2.
Figure 7 Gag induces proteasomal degradation of TRAF2. (A) Western blot of HEK293T cell extracts transfected with expressing vectors for Flag-TRAF2, HTLV-2Gag-GFP and Myc-ubiquitin. Cell extracts were immunoblotted with anti-Flag, anti-Myc, anti-GFP and anti-actin (more ...)
Ubiquitin-mediated proteolysis pathway
We identified cellular E2 ubiquitin-conjugating enzymes UBE2I and UBE2N or UBC13; and E3 SUMO-protein ligases PIAS (protein inhibitor of activated STAT) 1, 2 and 4. Both types of enzymes have been previously shown to play a role in Tax-mediated NF-kB activation [50
]. KEGG analysis also highlighted E3 ubiquitin ligases (CDC23, TRAF2 and TRAF6), which interact with HTLV proteins and which may play important roles in induced perturbations of the proteasomal pathway. CDC23 is a member of the anaphase promoting complex/cyclosome (APC/C, including CDC23), an E3 ubiquitin ligase that controls metaphase to anaphase transition [52
]. TRAF proteins contain a RING finger domain, a domain that can simultaneously bind ubiquitination enzymes and their substrates [55
] (Figure ). HTLV-1 Tax might also provide a bridge to the proteasome by disrupting the interaction between an E3 ubiquitin ligase and its substrate, illustrated by the inactivation by Tax of the A20-Itch E3 ligase complex, potentially leading to a permanent activation of tumor necrosis factor (TNF) receptor (TNFR) signaling [57
Figure 9 Targeting of the Ubiquitin-mediated proteolysis pathway by viral proteins. Schematic representation of PPIs. Big diamonds: viral ORFs, with HTLV-1 and HTLV-2 in blue and light blue, respectively. Small circles: human ORFs with green representing membership (more ...)
Most eukaryotic cellular proteins are selectively degraded by the ubiquitin-proteasome system [58
]. Numerous infectious and cancer agents induce aberrations in the proteasomal pathway, and several inhibitors have been proposed as promising therapies [59
]. Effective therapy faces challenges, as the activity of the proteasome is subjected to multiple regulation, and the selection of precise targeted proteins involves highly specific E2 and E3 ubiquitin enzymes [63
The highly conserved Notch signaling pathway regulates diverse cell fate decisions, including differentiation, proliferation, communication and specification. Several members of the Notch signaling pathway, including Numb [64
], dishevelled (Dvl) proteins [65
], cAMP-response element-binding protein (CREB)-binding protein (CREBBP or CBP) [66
], and p300 [68
], are targeted by HTLV Tax, Rex, Hbz, Gag and Pol proteins (Figure ). It has been recently shown that the γ-secretase inhibitor (GSI) reduced tumor cell proliferation and tumor formation in an Adult T-cell Leukemia animal model [69
]. To directly assess the involvement of the Notch pathway in viral infection, we treated an HTLV-1 transformed cell line (MT4) with a γ-secretase inhibitor (GSI) (L-685,458) [70
] and tested whether inhibition of the Notch pathway could affect HTLV-1 expression in MT4 cell line. Interestingly, we showed by quantitative RT-PCR, that inhibition of the Notch pathway significantly lowered HTLV-1 HBZ (p
< 2.1E-5), Gag (p
< 0.04) and Tax1 (p
< 0.003) expression in MT4 cells (Figure ), suggesting that GSI could be a new class of retroviral replication inhibitors.
Figure 10 Targeting of the Notch signalling pathway by viral proteins. (A) Big diamonds: viral ORFs, with HTLV-1 and HTLV-2 in blue and light blue, respectively. Small circles: human ORFs with green representing membership of the Notch pathway. Grey links: human-human (more ...)