MxB inhibits long interspersed element type 1 retrotransposition

Long interspersed element type 1 (LINE-1, also L1 for short) is the only autonomously transposable element in the human genome. Its insertion into a new genomic site may disrupt the function of genes, potentially causing genetic diseases. Cells have thus evolved a battery of mechanisms to tightly control LINE-1 activity. Here, we report that a cellular antiviral protein, myxovirus resistance protein B (MxB), restricts the mobilization of LINE-1. This function of MxB requires the nuclear localization signal located at its N-terminus, its GTPase activity and its ability to form oligomers. We further found that MxB associates with LINE-1 protein ORF1p and promotes sequestration of ORF1p to G3BP1-containing cytoplasmic granules. Since knockdown of stress granule marker proteins G3BP1 or TIA1 abolishes MxB inhibition of LINE-1, we conclude that MxB engages stress granule components to effectively sequester LINE-1 proteins within the cytoplasmic granules, thus hindering LINE-1 from accessing the nucleus to complete retrotransposition. Thus, MxB protein provides one mechanism for cells to control the mobility of retroelements.

Mx proteins are dynamin-like large GTPases. They have been reported to inhibit distinct groups of viruses by different mechanisms. MxA is cytoplasmic protein, inhibits a plethora of viruses from diverse families, including influenza A virus, vesicular stomatitis virus, LaCrosse virus, hepatitis B virus and others [30]. The antiviral activity of MxA relies on the GTP binding and hydrolysis domain, oligomerization of the stalk domain and the intact BSE (bundle signaling element). In contrast, much fewer viruses have been reported to be restricted by MxB. These viruses include HIV-1 [31][32][33], HBV [34], HCV [35] and herpesviruses [36,37]. MxB is mainly observed at the nuclear envelope, inhibits HIV-1 and herpesviruses through targeting viral capsid and blocking nuclear import of viral DNA [38]. The N-terminal domain (NTD) and the stalk domain are crucial for MxB to intercept HIV-1. Our data demonstrate that MxB restricts LINE-1 retrotransposition by engaging the stress granule pathway.

MxB inhibits LINE-1 retrotransposition
We used the CMV-L1-neo RT reporter to measure LINE-1 retrotransposition. This construct has a neomycin resistance gene inserted into the 3' untranslated region of LINE-1 in such a way that this gene can only be expressed from the reverse transcribed LINE-1 DNA, leading to cell resistance to G418 [39,40]. To test whether Mx proteins affect LINE-1 retrotransposition, we co-transfected HeLa cells with human Mx plasmid DNA and the CMV-L1-Neo RT reporter DNA (S1A Fig), followed by G418 selection. The results of colony assays showed that the ectopically expressed MxB diminished LINE-1 retrotransposition by 4.7-fold (p<0.01), whereas MxA exerted no effect (S1B Fig). We generated a defective L1 control which has the ORF1 RR261/262AA mutations inserted in the CMV-L1-neo RT DNA [40]. As a control for the specificity of MxB inhibition, the pEGFP-N1 vector DNA was tested with MxB overexpression, and no effect on the formation of G418 resistance cell colonies was observed (S1B Fig). A dose-dependent inhibition of LINE-1 by MxB was observed, with the lowest amount (200 ng) of MxB DNA causing significant loss of LINE-1 activity (S1C Fig). Cell viability was not affected by MxB overexpression (S1D Fig). Next, we performed quantitative PCR to measure levels of reverse transcribed LINE-1 DNA in cells transfected with CMV-L1-neo RT [41]. In agreement with the data of colony assay, MxB reduced LINE-1 DNA level by more than 2-fold (S1E Fig).
Although MxB is an interferon (IFN) stimulated gene, MxB has been shown to express at measurable levels in several human tissues such as lymph node, liver, and less expressed in brain and reproductive system (S2A Fig). When we used RT-qPCR and Western blot to measure endogenous MxB RNA and protein levels in HeLa cells, MxB expression in the absence of IFN treatment was detected as previously reported (S2C and S2D Fig) [42,43]. We then generated endogenous MxB knockout HeLa cell line using CRISPR/Cas9, and confirmed MxB knockout by sequencing (S2B Fig) and western blot (Fig 1D). We observed that LINE-1 activity increased by 1.9-fold (p<0.01) in the MxB knockout HeLa cells (Fig 1A). In addition, we measured the viability of MxB knockout cells and found no significant difference between MxB knockout and control cell lines ( We further tested CMV-L1-neo RT transposition in HeLa cells stably expressing MxB, and observed 2.9-fold (p<0.01) decrease compared to that in the control cells ( Fig 1B). Together, these data demonstrate an inhibition of LINE-1 restrotransposition by MxB, in agreement with previous report [28].
IFN has been shown to inhibit LINE-1 replication [44]. Since MxB is an IFN stimulated gene, we investigated the role of MxB in IFN-mediated LINE-1 inhibition. Activity of CMV-L1-neo RT in the control and MxB knockout cells was measured with and without IFN. The results of colony assay showed that LINE-1 retrotransposition was strongly inhibited by IFN in the control cell line, but this inhibition was significantly lost in MxB knock-out cell line (S2G Fig), which indicates that IFN-induced MxB contributes to IFN inhibition of LINE-1.
In support of MxB inhibition of LINE-1 retrotranspotion, the expression level of LINE-1 ORF1p diminished in HeLa cells stably expressing ectopic MxB (Fig 1C), and increased when MxB was knocked out (Fig 1D). We also investigated the effect of MxB on the activity of endogenous LINE-1 by performing quantitative RT-PCR to measure the level of endogenous LINE-1 RNA, using three pairs of primers that specifically amplify regions in the 5'UTR, ORF1 or ORF2 (Fig 1E) [45]. In MxB knockout HeLa cells, levels of endogenous LINE-1 RNA, as measured by all three primer pairs, significant increased, as opposed to the marked decrease seen in HeLa cells stably expressing MxB (Fig 1F and 1G). In summary, these results suggest that MxB is an inhibitor of LINE-1 retrotransposition.

MxB promotes sequestration of LINE-1 ORF1p to cytoplasmic bodies
We next investigated the molecular mechanisms by which MxB inhibits LINE-1 activity. In HeLa cells stably expressing MxB-EGFP, when the CMV-L1-neo RT reporter DNA was transfected, the ORF1p protein was seen to form cytoplasmic granules and co-localize with MxB (Fig 2A), which was also observed when only ORF1p protein was expressed in the absence of ORF2p (Fig 2B). These data were reproduced with transiently expressed MxB (S3A and S3B Fig).
We also investigated the relationship of subcellular location between MxB and LINE-1 RNP which were measured by the LINE-1-ms2x6/MS2-GFP system. As described previously, the expressed LINE-1 RNA contains 6 copies of MS2-binding sites which is bound by MS2-GFP  In the Western blots, the nuclear fraction samples were from cells 5 times more than the cells of the cytoplasmic fraction samples. The intensities of protein bands were quantified and the results are presented in the bar graphs. To calculate the ratio of nuclear/cytoplasmic MxB, the amount of nuclear ORF1p normalized with the nuclear LAMN protein level, the amount of cytoplasmic ORF1p with tubulin, then nucleus/cytoplasm ORF1p ratio was calculated by dividing nuclear ORF1p value by the cytoplasm ORF1p value. At the end, the nucleus/cytoplasm ORF1p ratio of the vector control is arbitrarily set at "1". N, nucleus; C, cytoplasm (mean ± SEM; paired t-test) (J) The Operetta High-Content Screen system (PerkinElmer) was utilized to detect the ORF1p fluorescence signals within the nuclei and the cytoplasm in HeLa cells which were transfected with MxB-EGFP and ORF1p-Flag. Ratios of the nuclear for visualization of LINE-1 RNA [46]. When LINE-1-ms2x6 and MS2-GFP DNA were coexpressed with MxB-Flag, LINE-1 RNA was seen to co-localize with MxB in the cytoplasm (Fig 2C). The association of MxB with LINE-1 RNP was also supported by the data of coimmunoprecipiation experiments performed with ectopic ORF1p (Fig 2D and 2E) or endogenous ORF1p (Fig 2F). The association of LINE-1 ORF1p and MxB was reduced when RNase was added during co-immunoprecipiation to remove RNA ( Fig 2G). These results suggest that MxB associates with ORF1p in an RNA-dependent manner.
MxB has been shown to inhibit the nuclear import of HIV-1 DNA [43,47,48]. We asked whether MxB also impairs the nuclear import of LINE-1 RNP. To answer this question, we performed nucleocytoplasmic fractionation experiments to measure the levels of ORF1p in the cytoplasm and the nucleus. Indeed, MxB reduced the nuclear ORF1p which was either ectopically expressed or endogenous (Fig 2H and 2I). This observation was supported by the data of high-content imaging analysis (Fig 2J), which calculates the ratios of ORF1p signals in the nucleus and cytoplasm in HeLa cells. We could also observe nuclear ORF1p in control cells, which disappeared with MxB overexpression (Fig 2A and 2B, indicated by white arrows).
Once in the nucleus, LINE-1 ORF2p cleaves cellular DNA by its endonuclease activity to initiate reverse transcription. This DNA cleavage event causes DNA damage response, shown by the increase of γH2AX foci. If MxB prevents the nuclear import of LINE-1 RNP, we expected to detect fewer LINE-1-induced γH2AX foci with MxB expression. When we immunostained γH2AX, we detected a great number of γH2AX foci in cells transfected with the CMV-L1-neo RT DNA. As expected, MxB stable expression dramatically reduced the number of these foci (Fig 3A, 3B and 3C). In agreement with the MxB overexpression data, greater numbers of γH2AX foci were scored in MxB knockout HeLa cells with CMV-L1-neo RT transfection than in the control cells, concurrent with increased LINE-1 ORF1 expression (Fig 3D,  3E and 3F). When LINE-1 DNA was transfected into MxB knockout cell lines, γH2AX level and foci number increased compared with control cells (S4A, S4B and S4C Fig). Ectopic MxB decreased the γH2AX level induced by CMV-L1-neo RT (Fig 3G). The defective L1-ORF1 (RR261/262AA) did not increase γH2AX level. We also observed that MxB overexpression alone diminished the number γH2AX foci, and MxB knockout itself led to an increase in γH2AX foci, which is likely a result of the suppression of endogenous LINE-1 by MxB. These data further suggest that MxB reduces the formation of double-strand DNA breaks associated with LINE-1 retrotransposition.

Knockdown of stress granule marker proteins abrogates MxB inhibition of LINE-1
LINE-1 ORF1p has been reported to associate with stress granules [49][50][51]. Several LINE-1 inhibitors have also been reported to co-localize with LINE-1 ORF1p in stress granules, including SAMHD1 [26], ZAP [28], APOBEC [52,53] and MOV10 [51,53]. We therefore asked whether the MxB-and ORF1p-containing cytoplasmic foci also bear stress granule markers. To this end, we performed immunofluorescence staining experiment to detect stress granule marker G3BP1 or TIA1. Indeed, the MxB/ORF1p foci contained G3BP1 and TIA1 in HeLa cells that stably expressed MxB and were transfected with CMV-L1-neo RT reporter or ORF1p plasmid (Figs 4A, 4B, S5A and S5B). We have scored the number of ORF1p-containing stress granules with and without MxB expression, observed no significant difference between the and cytoplasmic ORF1p signals were calculated. Results from the control group, which was transfected with the pEGFP-N1 vector, were arbitrarily set as 1. The results are summarized in the bar graph (mean ± SEM; paired t-test). ��  In support of the localization of MxB, ORF1p with G3BP1 and TIA1, these four proteins were co-immunoprecipitated with each other, which was not observed with MxA ( Fig 4D).
We next tested whether MxB inhibition of LINE-1 depends on stress granule formation. We thus depleted endogenous G3BP1 or TIA1, and examined whether MxB still inhibits LINE-1. Results of LINE-1 reporter assays showed that MxB lost inhibition of LINE-1 when either G3BP1or TIA1 was knocked down (Fig 4E and 4F). We also observed 50% increase in LINE-1 retrotransposition with knockdown of either G3BP1 or TIA1 in the absence of MxB expression (Fig 4E and 4F) [26]. In agreement with these functional data, knockdown of endogenous G3BP1 or TIA-1 led to the loss of ORF1p/MxB-EGFP granules in the cytoplasm (S6C and S6F Fig), further supporting the stress-granule nature of the ORF1p/MxB foci. Arsenite (AS) is commonly used to induce canonical stress granules. Interestingly, MxB did not localize to stress granules that were induced with arsenite, neither the endogenous MxB induced by IFN-α in HeLa cell (S7A Fig) nor ectopically expressed MxB (S7B Fig), which suggests that MxB itself is not recruited to arsenite-induced stress granules, but rather is located to G3BP1-and TIA1-bearing foci through interaction with LINE-1 ORF1p. Together, these data demonstrate that MxB associates with LINE-1 RNP in the stress granules, and enhances cytoplasmic sequestration of LINE-1 RNP. We treated both the control and MxB knockout cell lines with AS. The number of SGs increased significantly with arsenite treatment, yet there was no difference between MxB knockout and control cell lines (S7C Fig). When the CMV-L1-neo RT

MxB requires its NLS, GTPase activity and oligomerization to inhibit LINE-1
MxB has a GTPase globular domain and a stalk domain which mediates MxB oligomerization [55][56][57][58][59][60][61]. In addition, the N-terminal sequence of MxB has a nuclear localization signal (NLS) [31,62]. To understand the contribution of the different MxB domain to LINE-1 inhibition, we generated a series of MxB mutants (Fig 5A). We first deleted the N-terminal 25 amino acids which contain the NLS, and generated mutant MxBdel25. We also replaced this 25-amino acid sequence with the NLS from the SV40 large T antigen, this mutant is named MxBdel25+NLS. Lastly, we replaced the first 42 amino acids of MxA with the first 25 amino acids of MxB, to generate the MxA+B25 chimeric protein (Fig 5A). The subcellular localization of these Mx proteins were examined with confocal microscopy (Fig 5B). MxA was detected in the cytoplasm, while MxB was localized at the nuclear envelope as previously reported [43,63]. When the NLS in N-terminal 25 amino acids of MxB was deleted, the MxBdel25 mutant changed subcellular location to the cytoplasm, the distribution was the same as the natural short isoform of MxB [64]. Results of LINE-1 reporter assays showed that MxBdel25 lost the ability of inhibiting LINE-1 (Fig 5C). MxBdel25+NLS inhibited LINE-1 as strongly as the wild type MxB, and the MxBdel25+NLS mutant was detected at the peri-nuclear region, a distribution pattern not exactly the same as MxB itself (Fig 5B and 5C). Interestingly, the MxA+B25 protein, which has its first 42 amino acids changed to the first 25 amino acids of MxB, was localized to the peri-nuclear region, and did not inhibit LINE-1 (Fig 5B and 5C). These data suggest that the NLS is essential but not sufficient for Mx proteins to inhibit LINE-1.
It was reported that MxB is localized at cytoplasmic face of nuclear core complex (NPC) and interact with NPC protein NUP214, as well as transport receptor transportin-1 (TNPO-1) [47,65]. We thus investigated whether NUP214 or TNPO-1 are required for LINE-1 inhibition by MxB. We observed that TNPO-1, but not NUP214, was localized to the MxB/ORF1p cytoplasm foci (S8C and S8D Fig). When we knocked down endogenous NUP214 or TNPO-1, we found that NUP214 knockdown led to moderate decrease of LINE-1 retrotransposition, whereas TNPO-1 knockdown moderately increased LINE-1 retrotransposition (S8E and S8F  Fig). Knockdown of neither NUP214 nor TNPO-1 changed MxB inhibition of LINE-1. These data suggest a modest role of NUP214 and TNPO-1 in LINE-1 retrotransposition, but neither of these two proteins affect MxB inhibition of LINE-1.

PLOS GENETICS
We next generated more MxB mutants, including K131A which is defective in GTP binding ability [43], M574D/Y651D which disrupt MxB dimerization, as well as mutation R449D/ F495D that prevent MxB oligomerization [59,60]. When these MxB mutants were tested in LINE-1 reporter assays, none of them affected LINE-1 activity (Fig 5D), in agreement with the decreased association of these MxB mutants with G3BP1 or ORF1p (S9A and S9B Fig), and none of these mutants affected the nuclear import of LINE-1 RNP (S9C and S9D Fig). These MxB mutants exhibited dispersed cytoplasmic localization (Fig 5E). Together, these results suggest that both the GTPase activity and oligomerization are essential for MxB to inhibit LINE-1.

Inhibition of LINE-1 by Mx2-like proteins
We next tested whether MxB homologs of different species can also inhibit LINE-1. We first established a phylogenetic tree of Mx proteins from mammals (S10 Fig). The results showed that Mx proteins from most mammals segregate into two lineages, MxA/Mx1 and MxB/Mx2, with the exception of the mouse Mx proteins, mouse Mx1 and Mx2 are orthologous with human MxA. Human MxB ortholog in mouse was lost in evolution. We then tested the anti-LINE-1 activity of several Mx proteins from the MxA-like and MxB-like groups. The results showed that MxB protein from Macaca mulatta (Mac) and African green monkey (AGM) strongly inhibited LINE-1 (Fig 6A), whereas murine Mx1 and Mx2 did not affect LINE-1 activity (Fig 6B). The Equus caballus Mx2 also inhibited LINE-1 retrotransposition, the Ovis aries Mx2 exhibited week inhibition of LINE-1 (Fig 6B). To understand the variable effect of MxBlike proteins on LINE-1, we examined their subcellular localization. MacMx2 and AGMMx2 showed similar localization to the nuclear envelope as human MxB (Fig 6C). In the case of Equus and Ovis MxB proteins, their N-terminal sequences significantly differ from human MxB NLS. Nonetheless, Equus MxB was detected at the nuclear envelope as opposed to the Ovis MxB that was diffused in the cytoplasm (Fig 6C). These data suggest a correlation of localization to the nuclear envelope and the inhibition of LINE-1 by MxB-like proteins.
Finally, we tested whether human MxB can inhibit LINE-1 in non-human cells such as in murine BALB/3T3 clone A31 cell line. We stably expressed human MxB in the BALB/3T3 clone A31 cell line, and then transfected with human LINE-1 reporter CMV-L1-neo RT DNA. A strong inhibition of human LINE-1 retrotransposition was observed (Fig 6D). Levels of endogenous mouse LINE-1 subfamilies (L1A, L1Gf and L1Tf) and SINE RNA were measured by RT-qPCR with promoter specific primers in BALB/3T3 clone A31 cells that stably express human MxB. Marked inhibition of these mouse LINE-1 subfamilies (L1A, L1Gf and L1Tf) and SINE was observed ( Fig 6E). IAP (Intracisternal A-type Particle elements) and MusD are active mouse ERVs (endogenous retroviruses) (also named LTR (long terminal repeat) retrotransposons) [66]. When we transfected IAP-neo TNF and MusD-neo TNF reporter DNA into human MxB-expressing BALB/3T3 clone A31 cells, the retrotranspostion of mouse IAP and MusD was strongly inhibited (Fig 6F). These results suggest that Mx2-like (orthologous to human MxB) proteins inhibit a wide range of retrotransposons in different species, thus play an important role in safeguarding the genome.

Discussion
LINE-1 has approximately 100 full-length copies in human genome. Because of the mutagenesis nature of its retrotransposition into new loci, LINE-1 activity is tightly regulated by the host. Among the multi-layered mechanisms humans have evolved to control LINE-1 retrotransposition, it is not surprising that the innate antiviral factors, which defend against exogenous viruses, also have the ability of inhibiting LINE-1. This concept is strongly supported by a study from Goodier and colleagues who systematically tested a group of known interferoninduced antiviral proteins and found several of these factors with anti-LINE-1 function, including BST2, ISG20, MAVS, ZAP, and MxB (also called Mx2) [28]. In this study, we confirmed LINE-1 inhibition by MxB and further showed that MxB does so by sequestering LINE-1 RNP within cytoplasmic bodies containing stress granule marker proteins G3BP1 and TIA1. We further showed that the LINE-1 inhibition ability of MxB depends on its nuclear envelope subcellular location, the GTPase activity and oligomerization.
Clearly, there are differences in the detailed mechanisms by which MxB inhibits different viruses and LINE-1. For example, the requirement for the GTPase activity differs. The GTPase defective MxB still inhibits HIV-1 [31,32,54] but not herpesviruses [36,67]. Our data showed that MxB needs its GTPase function to suppress LINE-1. Although MxB depends on the cellular peptidylprolyl isomerase, cyclophilin A (CypA), to inhibit HIV-1 [33,54] and HCV [35], MxB inhibition of LINE-1 is CypA-independent (S8A and S8B Fig). These different requirements may reflect the different natures of the viral proteins which MxB targets, as well as the different replication strategies that the inhibited viruses employ. Regardless of these mechanistic variations, MxB strongly depends on its oligomerization and N-terminal sequence to inhibit its target viruses and LINE-1. It has been reported that MxB N-terminal sequence participates in the recognition of HIV-1 capsid structure and this activity is subject to regulation by phosphorylation [71,72]. It is possible that MxB interaction with LINE-1 ORF1p also requires its N-terminal sequence. We also noted that the N-terminal sequence was also reported to be dispensable for MxB to inhibit LINE-1 [28]. This discrepancy with our observation may be a result of the different experimental systems used in different studies including cell lines and protein tags. More studies are warranted to resolve this discrepancy.
Our data showed that MxB inhibition of LINE-1 depends on stress granule marker proteins G3BP1 and TIA1 (Fig 4E and 4F). LINE-1 proteins ORF1p and ORF2p as well as LINE-1 RNA have been reported to associate with cytoplasmic granules containing stress granule markers and p body markers [49][50][51]. This suggests that sequestering RNA to stress granules and p bodies serves as an intrinsic cellular strategy to control the activity of LINE-1, which operates through sequestering LINE-1 RNP within cytoplasmic bodies and therefore preventing LINE-1 from accessing the nucleus and completing retrotransposition. However, this strategy by itself may not be quite effective, as knockdown of G3BP1 or TIA1 only marginally increased LINE-1 activity (Fig 4E and 4F). MxB appears to enhance this intrinsic anti-LINE-1 mechanism, since we observed much greater LINE-1 ORF1p localization to G3BP1 or TIA1/MxBpositive cytoplasmic granules compared to the relatively more dispersed cytoplasmic distribution of ORF1p in the absence of MxB expression (Figs 4A, 4B, S5A and S5B). MxB is not the only factor stimulating sequestration of LINE-1 RNP in cytoplasmic granules, we and others have reported that ZAP [29] and SAMHD1 [26] employ a similar mechanism to assist cellular control of LINE-1 retrotransposition. It appears that cells have evolved means to regulate the stress granule-based anti-LINE-1 mechanism. One possible scenario is that high-level LINE-1 activity may stimulate the expression of MxB and other anti-LINE-1 factors that in turn suppress LINE-1 retrotransposition through engaging the stress granule pathway. In retrospective, sequestration of MxB to stress granules as a result of interaction with ORF1p alters subcellular localization of MxB, and may potentially influence its other cellular functions.
Stress granules are a typical membraneless organelles (MLOs) which are formed by liquidliquid phase-separated in cytoplasm. MLOs regulate diverse cellular function, such as protein turnover, mitosis mRNA storage and translation, virus replication, and antiviral activity. Murine Mx1 and Mx2 were observed in nuclear bodies and granular cytoplasmic structures [73]. Coincidentally, exogenously expressed as well as IFN induced human MxA protein also forms as membrane-less condensates in the cytoplasm, and interacts with and sequesters viral nucleocapsid proteins. Endogenous human MxB is localized to the cytoplasmic face of nuclear pore which also has phase-separated nature. Therefore, we speculate that Mx family proteins all have the propensity of forming MLO, which contributes to their antiviral function. MxB is not an integral component of stress granules, since we did not observe MxB localization with stress granules that were induced with arsenite treatment (S7A and S7B Fig). Expression of MxB itself did not cause stress granule formation either. Although there are MxB-containing cytoplasmic foci with MxB overexpression as seen in S3A Fig, these likely result from MxB aggregation. We suspect that MxB association with ORF1p stimulates G3BP1/TIA1-dependent aggregation of ORF1p, either through the GTPase activity of MxB and/or the ability of MxB to oligomerize. This mechanism allows MxB to inhibit LINE-1, also sends MxB to the ORF1p/ G3BP1/TIA1 positive cytoplasmic granules.
Our data showed that the GTPase mutation K131A prevents MxB localization to the nuclear envelope and impairs its interaction with ORF1p. Furthermore, Betancor and colleagues reported that the GTPase domain cooperates with the N-terminal domain in Mx2 binding to the HIV-1 capsid [61]. We speculate that the GTPase activity may influence MxB conformation which is involved in interaction with LINE-1 ORF1p (S9C Fig) and inhibiting LINE-1 retrotransposition (Fig 5D).
We further investigated Mx proteins from different species for their ability to inhibit LINE-1. Mx2 proteins from non-human primates, including Macaca mulatta (Mac) and African green monkey (AGM), which are homologous to human MxB, also effectively restrict LINE-1 (Fig 6A), whereas Mx1 and Mx2 proteins from rodents, both of which share homology with human MxA, exhibit no effect (Fig 6B). Equus caballus Mx2 strongly inhibited LINE-1, whereas the Ovis aries Mx2 exhibited modest inhibitory effect (Figs 6B and S10). We also observed that anti-LINE-1 Mx2 or MxB proteins are located to nuclear envelope. Furthermore, classical nuclear localization signal of the SV40 large T antigen enables the MxBdel25 mutant to inhibit LINE-1 (Fig 5C). These results suggest a correlation of MxB-like protein localization to nuclear envelope and their anti-LINE-1 function.
Furthermore, the retrotranspostion of mouse SINE, IAP and MusD was strongly inhibited by stably expressed human MxB in BALB/3T3 clone A31 cells. This suggests that MxB regulates a wide spectrum of retroelements. SINE does not encode proteins, is unable to retrotransposition autonomously, its activity depends on LINE-1 ORF1p and ORF2p [14,15]. MxB is thus expected to inhibit SINE through sequestering LINE-1 ORF1p as we observed in this study. IAP and MusD are endogenous retroviruses, they encode Gag protein as HIV-1 does [74]. It is likely MxB targets Gag and inhibits IAP and MusD, similar to MxB inhibiting HIV-1 by targeting viral capsid. Since MxB also inhibits retrotransposons in murine cells, it can be envisioned that MxB either exerts its inhibition by itself or is assisted by factors shared by human and murine cells In conclusion, our data support MxB as an anti-LINE-1 factor which functions by sequestering LINE-1 RNPs within the cytoplasmic granules through engaging the stress granule marker proteins G3BP1 and TIA1. This finding further substantiates the importance of the stress granule pathway in controlling LINE-1 and potentially other transposable elements. It will be interesting to investigate whether MxB uses similar mechanisms to restrict the infection of particular viruses.

Plasmids and antibodies
CMV-L1-neo RT reporter DNA contains the complete human LINE-1 DNA and a neomycin resistance gene as a reporter of LINE-1 retrotransposition [26]. The ORF1-Myc DNA was cloned into the pCMV-Tag 3B expression vector. The L1-ORF1 (RR261/262AA) plasmid was generated by inserting ORF1 261/262RR to AA mutations into the CMV-L1-neo RT plasmid, as reported in [40]. IAP-neo TNF and MusD-neo TNF reporter plasmids contain mouse LTR retrotransposon and a neomycin resistance cassette in reverse orientation [23]. The L1-ms2x6 contains six tandem MS2 CP binding sites near the 3' end of L1-RP in 99-PUR L1-RP vector. MS2-GFP plasmid was generated by inserting MS2 sequence in pEGFP-N1 vector. The MxB-Flag DNA was inserted into the pQC-XIP vector. The MxB mutations del25, K131A, R449D/F495D and M574D/Y651D were created using PCR-based mutagenesis method. The wild type and mutated MxB-EGFP were generated by fusing EGFP sequence to C-terminal end of MxB. The MxB sequence was inserted to the N-terminus of EGFP in pEGFP-N1 vector between the cleavage sites of restriction enzymes BamH1 and Age1, to generate MxB-EGFP  [33]. Rabbit anti-ORF1p antibody was generated by immunizing rabbits with recombinant ORF1p [26]. Anti-Flag M2 affinity gel (A2220) was purchased from Sigma.
MxB KO cells were generated using the CRISPR-Cas9 system. Cells were transfected with lentiCRISPRv2 (52961, Addgene) [75] carrying single guide RNAs (sgRNAs) that target MxB. The gRNA sequence was listed in S1 Table. Following selection with puromycin (0.8 μg/ml), the resistant cells were serially diluted in 96-well plates to obtain single cell clones.
HeLa and BALB/3T3 clone A31 cells stably expressing MxB were generated using the retroviral vector system pQC-XIP (Clontech) which expresses MxB.

Cell viability assay
Cell viability was measured using the cell-counting kit-8 (96992, Sigma) according to the protocol from the manufacturer. MxB/EGFP stably expressing cell lines or MxB knockout cell lines were cultured in 96-well plates with 2×10 3 cells per well. The MxB DNA was transfected into HeLa cells using PEI before cell viability was assessed. 10 μl cell-counting kit-8 was added into each well. After 3 hours, OD450 was measured with a microplate reader (Multiskan FC, ThermorFisher).

Cell colony assay
The LINE-1 retrotransposition assay was performed as described previously [23,26]. In brief, 1.5 × 10 5 cells/well were seeded in 6-well plates. After 18 hours, cells were transfected with the CMV-L1-neo RT reporter DNA with or without MxB DNA, using PEI (sigma). 48 hours posttransfection, G418 (0.75 mg/ml) was added to select for resistant cells because of LINE-1 retrotransposition. Twelve days after selection, the G418-resistant colonies were fixing with 4% paraformaldehyde, stained with 0.5% crystal violet, and scored. IAP-neo TNF or MusD-neo TNF reporter assays were similarly performed as the CMV-L1-neo RT reporter assay.

Quantification of LINE-1 cDNA by PCR
HeLa cells were transfected with MxB or empty vector and CMV-L1-neo RT reporter DNA.72 hours after transfection, total cellular DNA was extracted using the QIAamp DNA Mini kit (QIAGEN). The same amounts of DNA (250 ng) were subjected to PCR with primers L1cDNAF/L1cDNAR to amplify the reverse transcribed LINE-1 cDNA [46]. The forward primer L1cDNAR crosses neomycin resistance gene junction, thus only amplifies the spliced and reverse transcribed DNA. Levels of β-globin DNA were determined in PCR using primers β-globinF/ β-globinR. The results were used to normalize the levels of LINE-1 DNA [76]. The PCR products were separated in 1% agarose gels, and stained with Ethidium Bromide. All primer sequences are listed in S1 Table.

Quantification of LINE-1 RNA by RT-qPCR
HeLa cells were transfected with CMV-L1-neo RT reporter DNA, with or without MxB DNA. 36 hours later, cellular total RNA was extracted using Trizol reagent (Invitrogen). The same amount of total cellular RNA was treated with DNase. The RNA was dissolved in RNase free water and reverse transcribed with SuperScript III Reverse Transcriptase (Invitrogen) according to the manufacturer's instruction with the oligo dT primer. Endogenous LINE-1 RNA levels were determined by qPCR using the Luna Universal qPCR Master Mix (NEB, M3003), and analyzed with the ΔΔCT method. The primers were named 5'UTRF/5'UTRR, ORF1pF/ ORF1pR, ORF2pF/ ORF2pR [45]. The GAPDH RNA was measured as the internal control. RNA levels of endogenous retrotransposition elements of mouse LINE-1 and SINE in BALB/ 3T3 clone A31 cells was measured by RT-qPCR with primers: mL1_TfF/mL1_TfR, mL1_GfF/ mL1_GfR, mL1_AF/mL1_AR, mSineF/mSineR. The internal control is mRrm2 [77]. All primer sequences are listed in S1 Table.

Gene silencing
To knock down target gene by siRNA, HeLa cells was transfected with siRNA oligos (30 nM) by Lipofectamine RNAiMAX (Invitrogen) before plasmid transfection in 6-well plates. The knockdown efficiency was examined by Western blot. The siRNAs used were listed in S1 Table. The expression of CypA was silenced by shRNA targeting CypA (TRCN0000049228; Sigma).

Immunoprecipitation
Transfected HEK293T Cells were harvested and lysed in the RIPA buffer (0.1% SDS, 1% Triton X-100, 1% sodium deoxycholate, 150 mM NaCl, 10 mM Tris [pH 7.5], 1 mM EDTA). The cell lysates were clarified by centrifugation at 12,000 rpm for 10 min at 4˚C. One milligram of the cell lysates was incubated with 50 μl of anti-Flag M2 affinity gel (A2220 Sigma) for overnight at 4˚C. The beads were then washed three times with the RIPA buffer, followed by incubation with the 2X loading buffer at 95˚C for 8 minutes to elute the bound proteins. The eluted materials were separated in SDS-12% PAGE, and further analyzed in Western blotting.

Operetta high-content screening
To quantify the distribution of LINE-1 RNP complex in the nucleus and the cytoplasm, cells were visualized and analyzed on an Operetta High-Content Screen system (PerkinElmer). HeLa cells were transfected with ORF1-Flag and pMxB-EGFP. 48 hours after transfection, cells were seeded in a Cell-Carrier-96 plate (6005550, PerkinElmer) in triplicates at 20000 per well. 24 hours later, cells were fixed and permeablilized, and then incubated with DAPI to stain the nuclear DNA for 10 min at room temperature. After washing with PBS, plates were scanned using the Opertta HTS imaging system (PerkinElmer) to collect images which were further analyzed using the harmony software to quantify the percentage of ORF1p in the nucleus the cytoplasm in each cell [26].

Cytoplasmic and nuclear fractionation
Subcellular fractionation was performed as described previously [78] Briefly, cells were harvested and lysed in buffer A (20 mM Tris, pH 7.6, 0.1 mM EDTA, 2 mM MgCl 2 , 0.5 mM NaF, 0.5 mM Na 3 VO 4 supplemented with protease inhibitors (Sigma-Aldrich, S8830)) for 2 min at room temperature and for another 10 min on ice. Nonidet P-40 (NP-40) was then added at a final concentration of 1% (vol/vol). The lysates were homogenized by gently vortexing or inverting the tube. Cytoplasmic fraction was collected by centrifugation in a pre-chilled centrifuge at 500xg for 3 min at 4˚C. The pelleted nuclei were washed three times in buffer A containing 1% NP-40, then suspended in one to two pellet volumes of the extraction buffer B (20 mM HEPES (pH 7.9), 400 mM NaCl, 25% (vol/vol) glycerol, 1 mM EDTA, 0.5 mM NaF, 0.5 mM Na 3 VO 4 and 0.5 mM DTT), with vigorous vortex. The mixture was snap-frozen twice in liquid nitrogen and incubated for 20 min on ice. Lastly, the soluble nuclear extracts were collected by centrifugation at 20,000xg for 20 min at 4˚C. To visualize nuclear ORF1p, we examined in the Western blots nuclear fraction samples from cells 5 times more than the cytoplasmic fraction samples.
To calculate the ratio of nuclear/cytoplasmic MxB, we first normalized the amount of nuclear ORF1p with the nuclear LAMN protein level, the amount of cytoplasmic ORF1p with tubulin, then nucleus/cytoplasm ORF1p ratio was calculated by dividing nuclear ORF1p value by the cytoplasm ORF1p value. At the end, the nucleus/cytoplasm ORF1p ratio of the vector control is arbitrarily set at "1".

Phylogenetic analysis
Orthologous Mx2 sequences were downloaded from Genbank. The protein sequences were aligned using Clustal Omega and converted to a codon alignment using MEGA 7. The Neighbor-Joining tree was calculated in MEGA 7 with 1,000 bootstrap replicates.

Statistics
All experiments were performed three or more times independently under similar conditions. All data were plotted as mean values, with variation as SEM. Statistical significance was calculated by Student's two-tailed t test. P values of statistical significance are represented as ��� P<0.001; �� P <0.01; � P<0.05. The MxB-EGFP stably expressing HeLa cell lines were transfected with siRNAs targeting G3BP1 and then transfected with CMV-L1-neo RT DNA (500ng). Immunofluorescence was performed to detect ORF1p, G3BP1 and MxB 24 hours after transfection. Expression of G3BP1 was examined by Western blot. Scale bar, 10 μm. (D, E) Immunofluorescence was performed to detect ORF1p, TIA1 and MxB in HeLa cells co-transfected with MxB and CMV-L1-neo RT DNA (500ng) (C) or the ORF1p-Flag DNA (500ng) (D). The white arrows indicate nuclear ORF1p. (F) MxB-EGFP stably expressing HeLa cells were transfected with siRNAs targeting TIA1 and then transfected with CMV-L1-neo RT DNA. Immunofluorescence was performed to detect ORF1p, TIA1 and MxB 24 hours post transfection. Expression of TIA1 was examined by Western blot. ORF1p-containing SGs were scored in more than 50 cells for each treatment. The average number of ORF1p-containing SGs per cell is presented in the bar graph (mean ± SEM; paired t-test) in (C and F). ��� , P<0.001. Scale bar, 10 μm. Neighbor-Joining method was used for the evolution analysis. Orthologous Mx2 sequences were acquired from Genbank. The protein sequences were aligned using Clustal Omega and converted to a codon alignment using MEGA 7. The Neighbor-Joining tree was calculated in MEGA 7 with 1,000 bootstrap replicates. Bootstrap values (>70) were also tested. Mx genes were selected from MxA-like and MxB-like major clades. " � " indicates the Mx genes that were investigated in this study for inhibiting LINE-1 retrotransposition. (TIF) S1 Table. Primers and RNAs used in this study. (DOCX)