Two distinct regions of IFITM3's NTD are required for restriction of IAV.
Previously we and others had shown that loss of IFITM3 resulted in a marked decrease in the antiviral effects of IFN up to the stage of HA protein surface expression (1
). To test the effect of IFITM3 depletion on the entire viral replication cycle, we first treated A549 human lung carcinoma cells expressing a short hairpin RNA (shRNA) targeting IFITM3 (shM3) or a negative-control shRNA (shLuc [Luc stands for luciferase]) with or without IFN-α and then infected them with IAV (H1N1 A/WSN/33 [WSN/33]) ( and ). After 12 h of infection, the A549 cells were washed once and then incubated with fresh medium for an additional 6 h, at which point the resulting viral supernatants were used to infect MDCK cells. Both A549 (18 h postinfection) and MDCK cells (12 h postinfection) were fixed, immunostained for viral HA protein, stained for DNA, imaged, and analyzed to determine the percentage of infected cells under each condition. These data, together with the results of previous studies, demonstrate that depletion of IFITM3 results in a >60% loss of the in vitro
protective effects of IFN-α spanning the entire viral replication cycle.
We next hypothesized that specific amino acids within IFITM3 make critical contributions to its antiviral function, and identifying these molecular determinants will provide a better understanding of IFITM3's protective actions. The CD225 domain is also largely uncharacterized. Therefore, investigating the structural basis of IFITM3's restriction will improve our understanding of a number of related proteins. With this goal in mind, we constructed AS mutants along the entire IFITM3 protein (six amino acids changed to alanine per mutant [ and ]).
We chose this unbiased approach, because no previous work has looked at the function of IFITM3 at this resolution. Furthermore, we were concerned that a deletional analysis might disrupt the protein's architecture. We stably expressed either the wild-type IFITM3 or mutant proteins (designated by AS and the number of the first amino acid of the substituted residues) in A549 cells, which express low basal levels of endogenous IFITM3 (1
). The level of each AS protein was determined by immunoblotting using polyclonal anti-IFITM3 sera raised against the NTD (anti-NTD). We chose not to use an epitope tag because we have found that this results in some loss of function (data not shown).
Next, each AS protein was assessed for its impact on IAV infection. The cells were challenged with WSN/33 at a multiplicity of infection (MOI) of 0.5 to 1.0. The cells were then fixed and stained for viral HA expression and for cellular DNA. Immunofluorescent (IF) images were captured, and the percent infection and cell number were determined (). These studies showed that two portions of IFITM3's NTD, amino acids 19 to 29 and amino acids 43 to 55, were both required for inhibition of IAV infection ( and ). The 55AS protein was expressed at low levels and so was not further evaluated (data not shown).
The anti-NTD sera recognizes the IFITM3 sequence from amino acids 20 to 32 (underlined portion of NTD schematic in ) and was used to determine the expression levels of the wild-type (WT) or AS mutant proteins (). However, because their mutations destroyed its epitope, the anti-NTD sera could not detect expression of 19AS and 25AS. Therefore, the levels of these two proteins were evaluated using an antisera generated against the CIL of IFITM1 (). The two bands seen with the CIL antibody likely arise from it recognizing both the full-length protein (top band) and an IFITM3 species that lacks the first several kilodaltons (bottom band) due to proteolysis. In addition, the higher running band seen with the 25AS mutant may result from it being posttranslationally modified. For both the 19AS and 25AS mutants, we could not assess their intracellular location because the anti-CIL sera does not perform well in IF applications. The intracellular distribution of the 43AS mutant, which also exhibited decreased restriction, was visible at greater levels at the cell's periphery compared to the wild-type IFITM3 (). In support of this observation, the 43AS mutant, compared to IFITM3, was found to colocalize less with the late endosomal- and multivesicular body-associated protein, CD63 (). We conclude that two regions of the NTD (amino acids 19 to 29 and amino acids 43 to 54) are required for viral restriction and that amino acids 43 to 48 are also required for wild-type localization.
Fig 2 IFITM3's CD225 domain is necessary for viral inhibition. (A) Alignment of the IM1 and CIL IFITM3 AS mutant proteins, with the wild-type amino acid sequence (IFITM3) shown in the top row. (B) A549 cells stably transduced with the empty retroviral vector, (more ...) IFITM3's CD225 domain is necessary for viral inhibition.
Next, we assessed the role of IFITM3's IM1 and CIL in viral restriction; these regions represent IFITM3's CD225 domain ( and ). Alterations across a large segment of the CD225 domain (IM1 or CIL), spanning amino acids 67 to 102, lessened IFITM3's restriction of IAV to levels approaching those of the empty vector alone (). One reason for this decrease in activity was the lower levels of expression observed for the 67AS mutant (4-fold lower than IFITM3 based on densitometric analyses of immunoblot autoradiographs ) and the 73AS mutant (3-fold), as well as for the 80AS and 85AS mutants (both 2.5-fold less ). However, while lower expression probably accounts for some lost function, it would not fully explain the observed deficits based on other mutant proteins that restricted equivalently to wild-type IFITM3 when expressed at comparable levels (i.e., 121AS, H57A, and Y80A mutants [ and ]), suggesting that the levels of wild-type IFITM3 are saturating.
Fig 3 Virus-specific determinants of IFITM3-mediated restriction. (A) Alignment of the CIL, IM2, and carboxy-terminal domain (CTD) IFITM3 AS mutant proteins, with the wild-type amino acid sequence in the top row. (B) A549 cells stably transduced with the empty (more ...)
Fig 5 Multiple residues in IFITM3's NTD and CIL are required for restriction. (A) Normalized percent infection of the indicated IFITM3 point mutant cell lines spanning the IFITM3 protein. The cells were challenged side by side with IAV WSN/33 or DENV NGC, and (more ...)
Confocal imaging revealed that both 80AS and 85AS mutant proteins were more centrally located than WT IFITM3, suggesting that these regions influence intracellular distribution (). Consistent with this observation, expression of the 80AS mutant protein resulted in a similarly patterned redistribution of CD63, as well as a diminishment in the overall size and intensity of CD63-containing structures, suggesting that this region of IFITM3 may contribute to this compartment's formation and/or stability (). The 91AS mutant protein could not be expressed. The 97AS mutant protein was present at wild-type levels and localization but exhibited diminished antiviral function. Altering the last portion of the CIL in the 103AS mutant also lowered viral inhibition without perturbing location, albeit with 2-fold-lower expression than IFITM3 ( and to ). Alterations within IM2 (109AS and 121AS), produced moderate changes in antiviral activity ( to ), with the exception of 115AS, which could not be stably expressed (data not shown). Last, mutation of the CTD (127AS) resulted in a 4-fold decrease in restriction.
Virus-specific determinants of IFITM3-mediated restriction.
DENV, like IAV, is inhibited by IFITM3 (1
). To identify the structural determinants needed for DENV restriction, we utilized the same panel of IFITM3 AS-expressing A549 cell lines described above. The cells were challenged with DENV serotype 2 New Guinea C (NGC) strain and then immunostained for DENV E-protein expression as a measure of viral replication. These studies revealed that many of the same regions in IFITM3's NTD (amino acids 19 to 30) and CD225 domain (amino acids 67 to 90 and amino acids 103 to 108) that were necessary for IAV inhibition were also needed for DENV restriction (). However, two regions of IFITM3 that were required for inhibiting IAV were dispensable for curtailing DENV: NTD amino acids 43 to 54 and CD225 domain amino acids 97 to 102. We conclude that many structural determinants are jointly required to stop either IAV or DENV; however, there are also regions of IFITM3 that are specifically required to block the orthomyxovirus.
Multiple residues in IFITM3's NTD and CIL are required for restriction.
The AS mutant proteins revealed several regions of IFITM3 that were required for viral inhibition. To more precisely define the molecular determinants within these regions that are necessary for viral inhibition, we used a panel of proteins each with a single residue changed to alanine. These mutant proteins were stably expressed in A549 cells, assessed for expression by immunoblotting, and then challenged with IAV or DENV. We predominantly mutated charged residues and/or residues that were reported to have posttranslational modifications (PTMs) on PhosphoSitePlus, an open-access resource of PTMs (http://www.phosphosite.org/homeAction.do
). For example, when examining the residues in the NTD altered in the 19AS mutant, we found that phosphorylation of Y20 has been reported repeatedly by investigators using either human (17
) or mouse (19
) cells. These data prompted us to create an Y20A-expressing cell line, which demonstrated that removal of this tyrosine produces a loss of protection against both IAV and DENV (). Furthermore, alteration of Y20 resulted in a mislocalization of the mutant protein to smaller clusters throughout the cellular periphery and surface, as well as its lack of association with CD63 () (12
). The relative colocalization of Y20 with LAMP1 was also lower compared to IFITM3, with the normalized values being 1.0 ± 0.1 for IFITM3 versus 0.3 ± 0.1. for Y20A (data not shown). Our results regarding Y20A are consistent with those recently reported during the completion of this work (21
). Moreover, we note that the altered expression of Y20A is similar to that of IFITM1, suggesting that this region, which IFITM1 does not share, is responsible for the differential localization of these paralogs (; see ). The Y20A mutant protein also migrated slower than IFITM3 and the other mutant proteins in immunoblots, suggesting that this mutation may result in additional PTMs (). Several other point mutations engineered within the NTD, E21A, E25A, and E26A slightly decreased restriction of IAV ( and ).
Fig 4 Multiple residues in IFITM3's NTD and CIL are required for restriction. (A) A549 cells stably transduced with the empty retroviral vector, wild-type IFITM3, or the indicated IFITM3 alanine point mutant proteins were challenged with influenza A virus WSN/33 (more ...)
IM1 comprises the first portion of the CD225 domain, and AS mutants (67AS and 73AS) within this region were defective in intracellular localization and antiviral activity. Consistent with an earlier study, we also found that changing C72 to an alanine produced a decrease in both IAV and DENV inhibition along with a more centralized expression pattern ( and and and ). In contrast, conversion of C71 to an A showed no effect on either localization or IAV or DENV restriction (; data not shown). Further mutational analyses of IM1 will be discussed below.
Within IFITM3's CIL, Y99 has been reported to be phosphorylated on PhosphoSitePlus. However, the number of independent proteomics data sets reporting this PTM was less than Y20 (2 murine studies detected Y99-phosphate [P] versus 16 studies reporting the detection of Y20-P). A549 cells expressing the Y99A mutant protein showed that IAV replication was enhanced compared to wild-type IFITM3 (>12-fold increased replication ). Intriguingly, the cells expressing the Y99A mutant protein still restricted DENV infection, showing only a 1.8-fold increase in viral replication versus wild-type IFITM3. The Y99A mutant protein was expressed to levels similar to those of IFITM3 and has an intracellular distribution indistinguishable from wild-type ( and ). Therefore, two tyrosines, Y20 in the NTD and Y99 in the CIL, each contributed to IFITM3-mediated restriction, with Y99 being preferentially required for the inhibition of IAV over DENV.
Changing a region of mostly charged residues in IFITM3's CIL (85AS [RDRKMV]) strongly decreased restriction, likely due to the mutant protein's relocation to the perinuclear area (). We therefore assayed cells expressing IFITM3 proteins with single-alanine substitutions of R85, D86, R87, and K88 ( to and and ). The D86A protein could not be expressed (data not shown). Interestingly, mutation of R85 to alanine alone was sufficient to diminish restriction of both IAV and DENV and produce the perinuclear localization seen with 85AS ( and ). The location of the R85A mutant protein suggested it may be near the microtubule organizing center (MTOC), and this was confirmed by immunostaining for the kinetochores (). R85A colocalization with CD63 and LAMP1 was somewhat less than that seen between those proteins and IFITM3, with the relative colocalization of either R85A or IFITM3 with LAMP1 being 0.6 ± 0.1 and 1.0 ± 0.2, respectively (; data not shown). In contrast, R87A offered less protection from IAV than against DENV, while maintaining wild-type levels and location ( and and and ).
To assess which individual amino acids in the most C-terminal portion of the CIL (103AS) were needed for restriction, we individually substituted either K104 or C105 with alanine. Alteration of the K104 alone reproduced much of the loss-of-function phenotype seen with 103AS, along with wild-type expression and intracellular distribution ( and and and ). Similar to C71, we found that altering C105 to alanine did not affect IFITM3's location or its inhibition of IAV or DENV infection ( and ; data not shown). Data for additional point mutant proteins are also provided ( and ). In conclusion, these experiments identified or confirmed specific residues within IFITM3's NTD, IM1, and CIL that were needed for antiviral activity.
IM1 within the CD225 domain is required for IFITM protein interaction.
While performing affinity purification mass spectroscopy studies with exogenously expressed HA-tagged IFITM3, we recovered peptides unique to endogenous IFITM2 in addition to the IFITM3 peptides in the spectra, revealing a physical association between the paralogs (data not shown). To explore this finding, we assayed for potential interactions using coimmunoprecipitation (co-IP) studies of lysates from cells expressing HA-tagged IFITM1, IFITM2, or IFITM3 together with untagged IFITM3 (). Interactions were detected between all of the HA-tagged IFITM proteins and IFITM3 when they were cotransfected into 293T cells. Importantly, these interactions persisted in the supernatant fraction after clarification with high-speed centrifugation (100,000 × g), strongly suggesting that these associations are not due to the proteins simply residing in adjacent regions of membrane. Furthermore, the HA-tagged protein, MDA9, did not interact with IFITM3, demonstrating the specificity of this association (, right blots).
To find the residues needed for these interactions, we utilized the IFITM3 AS proteins. These co-IP experiments revealed that amino acids 67 to 85 within IM1 were required for IFITM association, with the most critical residues being amino acids 73 to 78 (LGFIAF [ and ]). Mutating the most hydrophobic residues within the sequence from amino acids 73 to 78 showed that loss of IFITM interactions occurred only when both F75 and F78 were changed to alanines (). The region from amino acids 67 to 85 includes C72, whose palmitoylation is required for restriction (10
). Therefore, we tested C72A, as well as two other amino acids reported to be required for restriction, C71 and C105A (10
), in the co-IP assay. This experiment showed that all three C-to-A mutant proteins interacted with HA-tagged IFITM3 similar to wild-type IFITM3, demonstrating that palmitoylation of any one of these cysteines is unnecessary for these interactions (). In keeping with these biochemical data, when we stably expressed the single mutants (F75A or F78A), double mutant (F75A/F78A), or quadruple mutant (LFIF/AAAA) proteins in A549 cells, only the F75A/F78A- and LFIF/AAAA-expressing cell lines failed to restrict IAV replication (). Confocal imaging showed that the F75A/F78A double mutant was expressed similarly to WT IFITM3, with both proteins partially colocalizing with the MVB- and lysosome-associated protein, CD63 () (12
). In addition, like IFITM3, the F75A/F78A double mutant protein partially colocalized with LAMP1, with the normalized colocalization values being 0.7 ± 0.1 for the F75A/F78A double mutant and 1.0 ± 0.2 for IFITM3 (data not shown). We also detected an interaction between endogenous IFITM3 and a stably expressed IFITM3 containing a single HA tag, and this interaction was enhanced after IFN-γ treatment was used to increase endogenous IFITM3 levels (, bottom bands in top blot). Attempts to test for a direct interaction have unfortunately been complicated by our inability to solubilize the recombinant full-length IFITM3 protein (data not shown).
The NTD of IFITM3 can alter the location of IFITM1 but does not increase its antiviral action.
Among the IFITM members, IFITM3 is the most potent in preventing IAV replication (1
). As noted, IFITM2 and IFITM3 reside in late endosomes and lysosomes, while IFITM1 localizes more to the early endosomes and the cell surface (2
). IFITM3 is 59% identical to IFITM1, with the major difference residing in IFITM3's NTD, which is 22 amino acids longer than the NTD of IFITM1 (). The highly similar IFITM3 and IFITM2 proteins differ in 12 amino acids, six of which reside in the NTD. Therefore, we constructed chimeric IFITM proteins to test whether the respective NTDs can influence restriction and/or protein localization. IFITM3's NTD (amino acids 1 to 50) was fused in frame with IFITM1's amino acids 29 to 245 or IFITM2's amino acids 50 to 118 to generate M3M1 and M3M2 (). The NTD of IFITM1, amino acids 1 to 28, or the NTD of IFITM2, amino acids 1 to 49, was fused to IFITM3, amino acids 51 to 119, to produce M1M3 and M2M3. Stably transduced A549 cell lines were generated for each respective chimera or wild-type protein. The anti-CIL serum was employed to determine the levels of protein expressed (). The cell lines were then infected with IAV (WSN/33) and immunostained for HA expression. Similar to earlier reports (1
), the relative strength of restriction among the wild-type proteins was observed to be IFITM3 > IFITM2 > IFITM1, with the slope of IFITM3's infectivity curve being considerably flatter above a relative MOI of 1.25 (1.9- and 2.4-fold less steep than IFITM2 and IFITM1's respective slopes) (). Moreover, when we replaced IFITM1's endogenous NTD with the longer NTD of IFITM3, the resulting M3M1 chimera was found to be only slightly better at inhibiting IAV than wild-type IFITM1 (). However, because M3M2 expression is 4-fold lower, the activity of the M3M2 protein is estimated to be comparable to the activity of the wild-type IFITM2. Last, since the M2M3 chimera was slightly less effective compared to IFITM3 at preventing viral replication and its levels were 3-fold higher, we infer that the NTD of IFITM2 is not functionally interchangeable with that of IFITM3.
Fig 7 The NTD of IFITM3 can alter the location of IFITM1 but does not increase its antiviral action. (A) Schematic diagrams of the IFITM chimeras. IFITM1 (M1; green), IFITM2 (M2; blue), and IFITM3 (M3; red) are shown with their IM1, CIL, and IM2 domains. The (more ...)
Interestingly, when we looked at the intracellular distribution of the M1M3 and M3M1 chimeras, we saw that their location is partially influenced by their respective NTDs; M1M3 was seen to be more finely distributed in a pattern that extended to the cell's periphery and surface, somewhat similar to the pattern of expression of IFITM1 or that of the Y20 IFITM3 point mutant (). In contrast, M3M1 was situated more centrally with a coarser staining pattern, showing its location in larger intracellular organelles that, based on earlier work, are lysosomes and autolysosomes, akin to IFITM3 () (12
) (data not shown). Together, these results suggest that the intracellular location of the IFITM proteins can be influenced by their respective NTDs. However, a similar transfer of enhanced restriction was not seen with the addition of IFITM3's NTD to the remainder of the IFITM1 protein. We conclude that location alone does not explain the differing efficacies of the IFITM proteins against IAV.
IFITM3 single-nucleotide polymorphisms.
A number of putative single-nucleotide polymorphisms (SNPs) have been identified in IFITM3 including the synonymous rs12252 SNP, which alters a predicted splice acceptor site and was previously reported by us due to its association with severe influenza (7
). We determined the validity of nonsynonymous IFITM3 SNPs in both dbSNP and the 1000GENOMES sequencing project ( and ; see Table S1 in the supplemental material). SNP H3Q/rs1136853 and rs12252 were present in both data resources. As previously reported, rs12252 has variable allele and genotype frequencies in different human populations with the severe influenza-associated alternative C allele being rare in Europeans ( and ) (7
). Similarly, the H3Q/rs1136853 alternative T allele is rare in all populations (minor allele frequency [MAF], 3%). Three SNPs, V31M/rs199582787, P70T/rs199749095, and G95R/rs1744108 were not present in the 1000GENOMES data most likely due to their MAFs of 0.2%, 6.4%, and 1.4%, respectively, in European and African American populations. For 7 SNPs (T4I/cosm42696, FS9S/rs56398316, H27Q/rs55888283, V31M/rs199582787, A34G/rs56188107, T42M/rs55900504, and P70T/rs199749095), the 1000GENOMES project provides evidence for miscalls of the SNPs due to misassignments occurring during human genome assembly and SNP calling; such events may have occurred because the short read 1000GENOMES data can map equally well to both IFITM3 and IFITM2. Indeed, for these 7 SNPs, IFITM2 (NCBI reference sequence NM_006435.2) encodes the alternative amino acid predicted to be a nonsynonymous SNP in its paralog, IFITM3 ( and and Table S1).
Fig 8 IFITM3 single-nucleotide polymorphisms. (A) The relative positions of multiple single-nucleotide polymorphisms (SNPs) reported within the coding region of the human IFITM3 gene are shown. The predicted nucleotide and amino acid changes (if nonsynonymous) (more ...)
However, for 5 SNPs (H3Q/rs1136853, D56G/rs55794999, H57D/rs1553883, N69D/rs12778, and G95R/rs61744108), the amino acids at these positions are identical in IFITM3 and IFITM2, suggesting that these SNPs were accurately called as nonsynonymous in either IFITM3 or IFITM2. To appreciate any functional significance associated with these five SNPs, we individually introduced them into a retrovirus gene-encoded IFITM3, in addition to two of the polymorphisms that effectively change IFITM3 to IFITM2 (T4I/cosm42696 and P70T/rs199749095) and the Q97Stop polymorphism (rs113745243). We then used the same number of CFU of each wild-type- or SNP-encoding retrovirus to stably transduce A549 cells.
Upon challenging these stably transduced cell lines with IAV, we found that the performances of three were indistinguishable from wild-type IFITM3: H3Q, T4I, and P70T (). However, the D56G, H57D, N69D, and Q97Stop cells all expressed the transduced proteins at levels much lower than those of wild-type IFITM3, H3Q, or T4I based on immunoblots and confocal images, and thus not surprisingly, they mounted less resistance to IAV ( to ; data not shown). D56 and H57 both reside among the amino acids altered in 55AS, which was also poorly expressed. Similarly, N69 is one of two conserved asparagines in IM1, a region of IFITM3 whose alteration led to reduced expression or antiviral function (61AS unexpressible) (67AS defective in restriction [ and ]). In contrast, G95R was expressed and localized similarly to both wild-type IFITM3 and P70T, while conferring 3-fold-less protection than either of these proteins ( to ). Therefore, G95R's attenuated restriction of IAV suggests that it may play a role in determining IAV resistance in individuals possessing this rare variant.