Transcription factor LSF facilitiates lysine methylation of α-tubulin by microtubule-associated SET8

Microtubules are critical for mitosis, cell motility, and protein and organelle transport, and are a validated target for anticancer drugs. However, tubulin regulation and recruitment in these cellular processes is less understood. Post-translational modifications of tubulin are proposed to regulate microtubule functions and dynamics. Although many such modifications have been investigated, tubulin methylations and enzymes responsible for methylation have only recently begun to be described. Here we report that N-lysine methyl transferase KMT5A (SET8/PR-Set7), which methylates histone H4K20, also methylates α-tubulin. Furthermore, the transcription factor LSF binds both tubulin and SET8, and enhances α-tubulin methylation in vitro, countered by FQI1, a specific small molecule inhibitor of LSF. Thus, the three proteins SET8, LSF, and tubulin, all essential for mitotic progression, interact with each other. Overall, these results point to dual functions for both SET8 and LSF not only in chromatin regulation, but also for cytoskeletal modification.


Introduction 1
Microtubules (MTs), the polymerized heterodimers of α-tubulin and β-tubulin, are major 2 cytoskeletal components that play important roles in key cellular processes such as structural 3 support, localization of organelles, and chromosome segregation (Janke, 2014;Verhey and 4 Gaertig, 2007). A number of post-translational modifications (PTMs) of tubulins have been 5 reported, which contribute to the functional diversity of MTs and affect MT dynamics and 6 organization (Song and Brady, 2015). This led to the hypothesis of a tubulin code (Verhey et al., 7 2007), where tubulin modifications specify biological outcomes through changes in higher-order 8 microtubule structure by recruiting and interacting with effector proteins. Notably, tubulin 9 methylation has been less studied than other types of tubulin modification, such as tyrosination, 10 glutamylation, glycylation, acetylation, and phosphorylation, although in the parallel histone 11 code hypothesis, methylation is the most common and well-understood modification. 12 SET8/PR-Set7 is a N-lysine methyltransferase responsible for the monomethylation of 13 both histone and non-histone proteins in higher eukaryotes (Dillon et al., 2005). It is functionally 14 characterized as a histone H4 lysine 20-specific monomethyltransferase (Fang et al., 2002); this 15 modification is a specific mark for transcriptional repression and is also enriched during mitosis 16 (Nishioka et al., 2002;Rice et al., 2002). SET8 is required for cell proliferation, chromosome 17 condensation, and cytokinesis, since deletion or RNAi mediated depletion of the enzyme impairs 18 all these functions. Previous findings, in particular, suggested that SET8 and H4K20me1 are 19 required for mitotic entry (Wu and Rice, 2011). SET8 also mediates monomethylation of other 20 substrates, including p53, which results in repression of p53 target genes (Shi et al., 2007). 21 However, how H4K20me1 is regulated and how it functions to promote cell cycle progression 22 remains an open question, including the possibility that other non-histone substrates may be 23 involved. 24 LSF, previously characterized widely as a transcription factor, is an oncogene in 25 hepatocellular carcinoma, being signficantly overexpressed in hepatocellular carcinoma cell lines 26 and patient samples (Fan et al., 2011;Gu et al., 2015;Kim et al., 2017;Seol et al., 2016;Yoo et 27 al., 2010;Zhang et al., 2017), as well as in other cancer types (Kotarba et al., 2018). LSF is also 28 generally required for cell cycle progression and cell survival (Hansen et al., 2009;Powell et al., 29 2000;Rajasekaran et al., 2015). Initially, LSF was described as a regulator of G1/S progression 30 (Powell et al., 2000) and essential for inducing expression of the gene encoding thymidylate 31 synthase (TYMS) in late G1. The additional involvement of LSF in mitosis resulted from 1 characterization of the effects of Factor Quinolinone Inhibitor 1 (FQI1), a specific small 2 molecule inhibitor of LSF (Rajasekaran et al., 2015). FQI1 not only abrogates the DNA-binding 3 and corresponding transcriptional activities of LSF (Rajasekaran et al., 2015), but also specific 4 LSF-protein interactions (Chin et al., 2016). Finally, FQI1 inhibits growth of hepatocellular 5 carcinoma tumors in multiple mouse models, and causes cell death via mitotic defects in 6 hepatocellular carcinoma cell lines (Grant et al., 2012;Rajasekaran et al., 2015). 7 In this study, we demonstrate that these three regulators of mitosis, SET8, LSF, and 8 tubulin, all interact with each other both in vitro and within cells. Furthermore, we demonstrate 9 that SET8 is a microtubule-associated methyltransferase that methylates lysines on α-tubulin. In 10 parallel to how transcription factors stimulate histone modification by interacting both with the 11 chromatin writers and the DNA, LSF stimulates methylation of tubulin by SET8. These results 12 suggest that LSF and SET8 have biological implications beyond gene transcription and histone 13 methylation, respectively. 14 15 Results 16

SET8 directly interacts with tubulin 17
Analysis of two commercial tubulin preparations (>97% and >99 % pure, respectively) by mass 18 spectroscopy identified anticipated associated proteins (e.g. MAP1, MAP2), but surprisingly also 19 peptides covering SET8 (Appendix Fig S1A; Appendix Table S1). The presence of SET8 in 20 these preparations was indicated by immunoblotting, although it is a minor component 21 (Appendix Fig S1B). This raised the question of whether SET8 might target  substrates. Indeed, initial overexpression of GFP-SET8 in a COS7 cell line indicated that the 23 majority of GFP-SET8 was localized in the cytoplasm (Fig 1A). Upon screening for association 24 with various cytoplasmic structural features by staining with relevant fluorescence dyes or 25 antibodies along with GFP-SET8 expression, GST-SET8 significantly co-localized only with α-26 tubulin, indicating MTs. MT co-localization was observed at stages through the cell cycle (Fig  27   1A, Appendix Fig S1C). Most obvious was in G1 phase, when SET8 exhibited the same pattern 28 as the filamentous tubulin distributed throughout the cytoplasm, emphasized by the yellow in the 29 merged image. In S phase, as expected from previous studies, some SET8 was also nuclear (Fig  30   1A). Although SET8 is often considered to be a nuclear protein, localization of SET8 also in the 31 cytoplasm of human cells has been documented by others, in a cell type-specific manner (Thul et 1 al., 2017). Furthermore, biochemical fractionation confirmed localization of endogenous SET8 2 in both the nucleus and the cytoplasm of human HEK293T cells (Fig 1B), which were used in 3 subsequent experiments. 4 To determine whether endogenous cellular SET8 associates with tubulins, we 5 immunoprecipitated protein complexes from HEK293T cell extracts. Using antibody against 6 SET8, α-tubulin was also precipitated, as well as β-tubulin, although to a considerably lesser 7 extent ( Fig 1C). Conversely, upon expression of Flag-tagged α-tubulin or β-tubulin in the cells, 8 endogenous SET8 co-immunoprecipitated with both, to roughly similar extents compared to the 9 level of expression of the tagged tubulin ( Fig 1D). As α-and β-tubulins stably heterodimerize in 10 cells, in vitro experiments were required in order to determine whether either of these 11 interactions was direct. To this end, purified recombinant proteins fusing maltose binding protein 12 (MBP) to either α-tubulin (TUBA1A) or β-tubulin (TUBB) were individually tested for 13 interactions with His-tagged SET8 purified from E. coli. SET8 directly interacted only with 14 α-tubulin, but not with β-tubulin ( Fig 1E). Conversely, recombinant proteins fusing glutathione 15 S-transferase (GST) to either full-length, or the N-or C-terminal overlapping portions of human 16 SET8 were tested for interactions in vitro with a purified tubulin preparation. The purified 17 heterodimeric tubulin interacted only with the full-length and N-terminal portion of SET8, even 18 though th C-terminal SET8 fusion protein was present at a higher level than the others (Fig 1F), 19 indicating specificity of this interaction. Taken together, these data demonstrate that α-tubulin 20 and SET8 directly interact with each other, whereas β-tubulin is only in a complex with SET8 in 21 the presence of α-tubulin. 22 23 SET8 methylates α-tubulin 24 SET8 was characterized historically as a nucleosomal H4K20 specific methyltransferase, and 25 subsequently as a regulator of the non-histone protein p53. However, since SET8 bound strongly 26 to α-tubulin, we tested whether tubulins could be a novel substrate of the enzyme. Purified 27 mammalian α/β-tubulin was incubated with the cofactor S-adenosyl-L-[methyl-3 H] methionine 28 (AdoMet) and purified, recombinant GST-SET8. In the presence of both SET8 and AdoMet, 29 radioactivity was incorporated into a protein band migrating at the position of α-and β-tubulins, 30 in addition to a less pronounced automethylation of GST-SET8 (Fig 2A, lane 3), but no 31 radioactive product was present at the position of α-and β-tubulins when either tubulin or SET8 1 was omitted from the reaction (Fig 2A, lanes 2 and 4). Interestingly, when histone H4 was also 2 included in the reaction, the amount of tubulin modification was reduced (Fig 2A, lane 1), 3 indicating that histone H4 strongly competed with tubulins for the methylation activity of SET8. 4 Furthermore, histone H4 also competed with SET8 itself as a substrate, as shown by the 5 significant reduction in SET8 automethylation in the presence of histone H4. 6 Since purified tubulin is composed of α-and β-tubulin heterodimers, we sought to 7 determine which species is methylated by SET8. Recombinant fusion proteins of either α-tubulin 8 or β-tubulin with MBP were purified and incubated with SET8 along with the radioactive methyl 9 donor. Upon incubation of SET8 with α/β tubulin and AdoMet, both SET8 and tubulin(s) were 10 labeled. However, only MBP-α-tubulin, but not MBP-β-tubulin was methylated, along with 11 SET8 itself, when the individual recombinant proteins were tested ( Fig 2B). These data indicate 12 that α-tubulin is the target for SET8. Mass spectrometry was used to determine which lysine 13 residue(s) of α-tubulin were methylated by SET8. In the control samples without exogenous 14 SET8, lysine methylation of α-tubulin on K304, and of β-tubulin on K19 and K297 were 15 detected (Appendix Fig S2A), none of which have previously been reported. As anticipated from 16 the previous data ( Fig 2B), incubation with exogenous SET8 did not induce detectable 17 methylation at any other sites on β-tubulin. However, SET8 did induce methylation of three 18 additional lysine residues of α-tubulin -K280, K311 and K352 -which were all monomethylated 19 ( Fig 2C, Appendix Fig S2A). Of these three lysines, only K311 is located on the outside surface 20 of MTs, whereas K352 is at the interface between α-tubulin and the β-tubulin in the adjacent 21 heterodimer, and K280 is on the inside surface of MTs ( Fig 2D, Appendix Fig S2B,C). In 22 addition, only the sequence surrounding K311 (RHGK311) resembles those of other known SET8 23 target sequences: histone H4 (RHRK20) and p53 (RHKK382) (Shi et al., 2007). In contrast, the 24 sequences of the other α-tubulin sites, SAEK280 and TGFK352, do not resemble other known 25 physiological SET8 tarets. To determine the relative efficiency of targeting these lysines in vitro, 26 each was independently mutated in the context of the full-length MBP-α-tubulin, and purified 27 proteins were tested for incorporation of radioactivity upon incubation with SET8. In order to 28 test the relevant degree of methylation by SET8 at the identified sites, each lysine was mutated 29 indivually to serine, maintaining a similar structure and hydrophilicity, but removing the charge. 30 Consistent with K311 being the best sequence match with other SET8 targets, only mutation of 31 K311 was no longer detectably modified by SET8, whereas mutation of K280, K304, or K352 1 did not appreciably affect the amount of methylation of the substrates (Fig 2E). 2 In order to test further the inherent targeting of the various α-tubulin sites by SET8, 3 peptides spanning these three sites (K280, K311, K352), as well as K40, reported to be 4 methylated by SETD2 (Park et al., 2016), and K304, modified in purified porcine tubulin 5 (Appendix Fig S2A), were incubated with purified wild type SET8 in vitro. Only the K311-6 containing peptide was robustly methylated (Appendix Fig S3A, Appendix Table S2). In 7 addition, radioactive incorporation into the K311-containing peptide was absent when incubated 8 with catalytically inactive SET8 (D338A) in vitro, and methylation abolished if the K311 residue 9 was either mutated (K311A, K311S) or already modified (K311Me, K311Ac) (Appendix Fig  10   S3B, Appendix Table S2). Although the in vitro targeting of the α-tubulin K311-containing 11 peptide by SET8 is robust (Appendix Fig S3C), SET8 methylates histone H4 much more 12 efficiently, consistent with the ability of Histone H4 to strongly compete against tubulin for 13 methylation by SET8 (Fig 2A). In contrast, but consistent with the previous report (Park et al., 14 2016), methylation by the only other reported tubulin methyltransferase, SETD2, of the α-tubulin 15 K40-containing peptide was not detectable over background, despite some methylation of 16 histone H3 by this enzyme in vitro (Appendix Fig S3D). 17 Taken together, these observations indicated that SET8 methyltransferase has the 18 capacity to directly methylate α-tubulin, particularly at K311. 19 20 Transcription factor LSF associates with both SET8 and tubulin 21 DNA-binding proteins recruit chromatin writers to modify histones (Brownell et al., 1996;Hassig 22 and Schreiber, 1997;Struhl, 1999;Taunton et al., 1996), suggesting the possibility that tubulin-23 binding proteins might similarly recruit SET8 to target sites on microtubules resulting in tubulin 24 modification. Our previous studies showed that the transcription factor LSF interacts with 25 DNMT1, and addition of an inhibitor of the LSF-DNMT1 interaction resulted in alterations in 26 the genomic DNA methylation profile (Chin et al., 2016); this is consistent with recruitment of 27 DNMT1 to DNA by LSF in order to facilitate DNA methylation at specific sites. Since DNMT1 28 complexes with SET8, and both SET8 and LSF (Rajasekaran et al., 2015) are required for 29 mitotic progression, we proposed the novel hypothesis that the transcription factor LSF might 30 also recruit SET8 to microtubules in order to facilitate α-tubulin methylation by SET8. In support 31 of this hypothesis, there is precedent for some DNA-binding transcription factors also binding 1 microtubules (Alexandrova et al., 1995;Dong et al., 2000;Giannakakou et al., 2000;Maxwell et 2 al., 1991;Niklinski et al., 2000;Ziegelbauer et al., 2001), although for the purpose of 3 sequestering the transcription factors in the cytoplasm and/or facilitating their transport into the 4 nucleus. 5 To test this hypothesis, multiple assays were initially performed to evaluate whether LSF 6 interacts with SET8 and tubulin(s). In vitro, direct interaction between recombinant, purified 7 SET8 and purified LSF was evaluated by a GST pull-down assay ( Fig 3A). Using fusion proteins 8 between GST and either full length SET8, or its N-or C-terminal fragments, the His-LSF bound 9 specifically to the N-terminal region of SET8 (Fig 3A), the same domain that bound purified 10 tubulin ( Fig 1F). The binding of purified α/β-tubulin to purified GST-LSF was also evaluated 11 and mapped to specific regions within LSF. Both α-and β-tubulins showed similar binding 12 profiles to the panel of LSF fusion proteins, as expected given their stable heterodimeric 13 structure ( Fig 3B). Binding of tubulins to the full-length GST-LSF was greater than to the control 14 GST, although weak compared to some of the other fusion proteins, due to the sensitivity of the 15 full-length GST-LSF fusion protein to cleavage in bacterial culture, resulting in significant 16 purification of the GST domain alone in the preparation. However, the tubulins interacted 17 strongly with two specific domains of LSF whose GST fusion proteins were stable: the DNA 18 binding domain (DBD), and to a lesser extent, the SAM domain ( Fig 3B). Further analysis 19 suggests that it is the C-terminal portion of the DBD that contains the tubulin interaction surface 20 in this domain, since the GST-LSF 2 protein also binds both tubulins to a high degree. Finally, 21 purified His-LSF also interacted in parallel assays with purified recombinant full-length GST-α-22 tubulin ( Fig 3C). These in vitro protein-protein interaction results indicate that all pairwise 23 interactions among SET8, LSF, and α-tubulin occur through direct binding with each other. 24 To examine whether interactions of LSF with both SET8 and tubulin also take place in 25 cells, multiple approaches were taken. First, upon co-expression of GFP-SET8 and 3xFlag-26 tagged LSF in transient transfection assays, the two proteins significantly co-localized, 27 predominantly in the cytoplasm ( Fig 3D). Second, the presence of complexes between 28 endogenous cellular proteins were tested by co-immunoprecipitation experiments using lysates 29 from the human HEK293 cell line. With antibodies against endogenous LSF, but not control 30 antibodies, both endogenous SET8 and endogenous α-tubulin co-immunoprecipitated with LSF 31 ( Fig 3E). Reciprocally, SET8 antibodies, but not control antibodies, also co-immunoprecipitated 1 endogenous LSF ( Fig 3F). Finally, the possibility of relevant LSF-tubulin interactions was 2 investigated by analyzing whether LSF was present in commercial, highly purified tubulin 3 preparations. These preparations are obtained in part by multiple rounds of 4 polymerization/depolymerization of the tubulin, and are more than 97-99% pure. They are well 5 known to contain additional proteins that are defined as microtubule-associated proteins (MAPs). 6 Immunoblots using a LSF monoclonal antibody did indeed detect a band comigrating with LSF, 7 albeit at a very low level ( Fig 3G). LSF was reproducibly detected in this manner in multiple 8 commercially purified preparations of tubulin. 9 Taken together, these results demonstrate that LSF interacts directly with both SET8 and 10 α-tubulin in vitro, and also associates with both in vivo. Furthermore, LSF, although a 11 transcription factor, appears to be a previously unidentified MAP. 12

LSF promotes tubulin methylation by SET8 14
The demonstration of pairwise, physical interactions between LSF, tubulin, and SET8 set the 15 stage for directly testing the hypothesis that LSF could mediate the methylation of α-tubulin by 16 SET8. Thus, recombinant GST-SET8 and the methyl donor were incubated with tubulin in the 17 presence of increasing concentrations of purified His-LSF ( Fig 4A). Tubulin methylation 18 increased upon increasing LSF from a 1:4 to 2:1 molar ratio of LSF:GST-SET8, suggesting that 19 LSF can mediate tubulin methylation by SET8. Note that in this experiment, there was more 20 SET8 relative to tubulin than in other experiments ( Fig 4A, bottom), resulting in a greater degree 21 of automethylation of SET8 compared to tubulin methylation ( Fig 4A, top), presumably due to 22 substrate competition between SET3 itself and α-tubulin. A similar experiment was performed 23 using recombinant MBP-α-tubulin as substrate for SET8, which also showed that increasing 24 levels of LSF enhanced methylation of the MBP-α-tubulin (Appendix Fig S4A). 25 The LSF small molecule inhibitor, FQI1, inhibits LSF binding to DNA (Grant et al., 26 2012), as well as binding of LSF to certain protein partners (Chin et al., 2016). To determine 27 whether FQI1 would also diminish the interaction in cells between LSF and tubulin, cell lysates 28 from vehicle-versus FQI1-treated cells were analyzed by co-immunoprecipitation assays. These 29 demonstrated a significant reduction in the LSF-tubulin interaction after FQI1 incubation (Fig  30   5B). Since FQI1 can inhibit the LSF-tubulin interaction in vivo, it was used to interrogate 31 whether the interaction of LSF and tubulin was important for stimulating the SET8-mediated 1 methylation of tubulin in vitro. Since LSF is already present in the tubulin preparations, initially, 2 FQI1 was added to reactions containing only SET8, methyl donor, and purified tubulin. Tubulin 3 methylation decreased with increasing concentrations of FQI1 (Fig 5C), consistent with the 4 presence of LSF and its ability to enhance SET8-dependent tubulin methylation. Whether FQI1 5 specifically inhibits LSF in these assays was tested in two ways. First, it was demonstrated that 6 the presence of FQI1 prevented any increase in tubulin methylation upon addition of purified 7 His-LSF (Appendix Fig S4B, compare lanes 7 and 8). Second, the possibility that FQI1 directly 8 inhibits SET8 catalytic activity was tested using Histone H4, instead of α-tubulin, as a substrate. 9 Limiting amounts of histone H4 were added in this experiment to enhance the sensitivity of the 10 assay. FQI1 did not inhibit methylation of histone H4 by SET8, in contrast to its effect on 11 α-tubulin methylation ( Fig 5D). In addition, when SET8 was incubated with whole cell extract in 12 the presence of the radioactive methyl donor, FQI1 did not appreciably diminish methylation of 13 any of the other proteins in the extract (Appendix Fig S4B, compare lanes 1 and 2). 14 These data indicate that LSF enhances tubulin methylation by SET8, and conversely that 15 FQI1, which abrogates the tubulin-LSF interaction (Fig 4B), impedes methylation of α-tubulin 16 by SET8. Overall, these data support a model that LSF recruits SET8 to tubulin, and/or that LSF 17 binding as a ternary complex to SET8 and tubulin activates the methylase activity of SET8 18 already associated with tubulin ( Fig 4E). To probe whether LSF recruits SET8 to tubulin, we 19 again used FQI1 to disrupt the LSF-tubulin interactions. FQI1 did diminish co-20 immunoprecipitation of endogenous tubulin with SET8 antibodies (Appendix Fig S4C), 21 supporting the recruitment model. However, a caveat to this straightforward interpretation was 22 that SET8 immunoprecipitation was also somewhat diminished, although less so, after 23 incubation of the cells with FQI1. α-and β-tubulins at multiple C-terminal glutamate residues by TTLL4, 5, and 7 (Ikegami et al., 3 2006;Janke et al., 2005). Although extensive research has been carried out on various modified 4 sites and has identified the relevant enzymes, only one study has previously identified lysine 5 methylation of tubulin (Park et al., 2016), which is the focus of this report. Walker's group 6 reported that SETD2, known as a histone methyltransferase for a chromatin activation mark, 7 H3K36me3, also methylates α-tubulin at K40. Furthermore, the loss of SETD2-mediated tubulin 8 methylation resulted in mitotic microtubule defects and genomic instability. Here, we describe a 9 distinct, novel lysine methylation of α-tubulin at K311 and identify the enzyme responsible for 10 its modification as SET8. In addition, we identify methylation of β-tubulin purified from 11 mammalian brain at K19 and K297, and are pursuing identification of the enzymes responsible. 12 Importantly, we also demonstrate the surprising finding that a transcription factor, LSF, 13 moonlights as a microtubule-associated protein, and that it can recruit SET8 to tubulin and/or 14 enhance its activity on tubulin to facilitate the modification. The recruitment mechanism mirrors 15 mechanisms of targeting of histone writers to chromatin, expanding the model of the parallel 16 nature between the generation of the histone and tubulin codes. Furthermore, these data indicate 17 that transcription factors more generally may be able to regulate tubulin modifications, and 18 thereby microtubule dynamics. Although several transcription factors have previously been 19 reported to bind microtubules, including c-myc (Alexandrova et al., 1995;Niklinski et al., 2000), 20 MIZ-1 (Ziegelbauer et al., 2001), p53 (Giannakakou et al., 2000Maxwell et al., 1991), and 21 Smads (Dong et al., 2000), in all these cases the biological relevance proposed or demonstrated 22 was to sequester the transcription factors in the cytoplasm, and/or to help transport the 23 transcription factor into the nucleus. Thus, all previous indications had to do with regulation of 24 transcription activity, not of microtubule function. This is the first case in which binding of a 25 transcription factor directly to microtubules would lead to altered microtubule modifications, and 26 to altered function. We propose that this presents a new paradigm that may well occur with other 27 transcription factors, as well. 28 Tubulin PTMs are generally thought to regulate protein-protein interactions within the 29 microtubule cytoskeleton, thereby regulating signaling events in the cell. To date, a large variety 30 of microtubule associated proteins (MAPs) have been characterized, many of which stabilize and 31 destabilize microtubules, are associated with the coupling of molecular motors and microtubules, 1 and play critical roles in spindle formation (Stanton et al., 2011). 2 Proper mammalian cell cycle progression requires precise modulation of SET8 levels, 3 which suggested that SET8 and H4K20me1 function as novel regulators of cell cycle 4 progression, although the previous focus has been on regulation of S phase (Milite et al., 2016). 5 With our demonstration that SET8 can also methylate α-tubulin, the roles of non-histone 6 substrates must also be considered as causes for SET8-mediated regulation of the cell cycle, and 7 in particular of mitosis when SET8 is most abundant (Rice et al., 2002). As in mammals, SET8 8 is essential for life in D. melanogaster, as SET8 mutants die during larval development. 9 However, upon replacement of all Drosophila histone H4 genes with multiple copies of a mutant 10 histone H4K20A, the flies survived to adulthood without apparent DNA replication defects 11 (McKay et al., 2015). Thus, contrary to the prevailing view, histone H4 was not the critical target 12 for SET8. Given the minimal biological effects in Drosophila of mutating histone H4K20, we 13 propose that α-tubulin methylation is a strong candidate for mediating critical SET8 14 consequences. Notably, D. melanogaster and human α-tubulins are 98% identical with all the 15 lysines throughout the sequence being conserved. 16 Mitosis is viewed as a vulnerable target for inhibition in cancer (Komlodi-Pasztor et al., 17 2012). In that light, it is notable that LSF promotes oncogenesis in hepatocellular carcinoma, the significantly inhibits tumor growth in multiple mouse hepatocellular carcinoma models, with no 24 observable toxicity to normal tissues (Grant et al., 2012;Rajasekaran et al., 2015). These new 25 findings that LSF interacts with tubulin and SET8, and that FQI1 disrupts the LSF-tubulin 26 interaction, may be related to the impact of the LSF inhibitors in hepatocellular carcinoma cells 27 and tumors. Expression both of particular tubulins (e.g. TUBA1B) and of SET8 are upregulated 28 in hepatocellular carcinoma tumor samples, as compared to normal liver (Guo et al., 2012;Lu et 29 al., 2013). Moreover, SET8 is required to maintain the malignant phenotype of various cancer 30 types (Hou et al., 2016). Given the current lack of effective treatments, further investigation into 31 the relevance of the LSF-tubulin-SET8 pathway to hepatocellular carcinoma and other cancer 1 types in which LSF is oncogenic may aid in targeted and effective treatment. 2

Material and Methods 3
Cell Culture, immunoprecipitation, and immunofluorescence 4 HEK293T and COS7 cells were cultured in DMEM media supplemented with 10% fetal bovine 5 serum. FQI1 (Millipore/Sigma #438210) treatment of HEK293T cells was for 24 hours at 37°C 6 with 2.5 μM FQI1. 7 Immunoprecipitation (IP) and immunofluorescence experiments were carried out as 8 described previously (Andrews and Faller, 1991;Estève et al., 2006). For the 9 immunoprecipitation, 1 mg of total HEK293T cellular extract was incubated with 5 μg of anti- For immunofluorescence to detect α-tubulin and SET8 co-localization, COS7 cells were 18 grown on coverslips and transfected with a GFP-SET8 expression plasmid. After cells were fixed 19 with paraformaldehyde, the cells were incubated with anti-α-tubulin and the microtubules 20 visualized with an anti-mouse IgG coupled with Alexa Fluor 488 (Molecular Probes) using a 21 confocal microscope (Zeiss LSM510). For the detection of SET8 and LSF co-localization, COS7 22 cells were co-transfected with GFP-SET8 and 3XFlag-LSF expression plasmids; the epitope 23 tagged LSF was detected by mouse anti-FLAG antibody (F3165, Sigma-Aldrich) and visualized 24 with an anti-mouse IgG coupled with Alexa Fluor 488 (Molecular Probes). DAPI was used to 25 stain nuclear DNA. For analysis of FQI1 effects on mitosis, paraformaldehyde-fixed cells were 26 analyzed for microtubules with anti-α-tubulin antibody (Abcam #AB7750) and stained with 27 DAPI (Invitrogen) for visualizing genomic DNA. Samples were imaged using a Zeiss 28 Axioimager M1 microscope utilizing 63x and 100x magnifications. 29 30 GST and MBP pull down assays 1 LSF, SET8 and α-tubulin cDNAs were cloned into the pGEX-5X-1 (GE Healthcare) or 2 pMalC4X (New England Biolabs) vector and GST-tagged or MBP-tagged proteins were 3 captured using Glutathione Sepharose beads (GE Healthcare) or amylose resin (New England 4 Biolabs), respectively. Sepharose beads containing approximately 10 μg of fusion protein were 5 incubated for 2 hours at 4°C with purified tubulin (MP-biomedical), recombinant His-tagged 6 LSF, or recombinant His-tagged SET8, the latter two being purified from E. coli. Proteins bound 7 to the beads were resolved by 10-20% SDS-PAGE. LSF, SET8 and α-tubulin were visualized by 8 immunoblotting by using anti-LSF (BD Biosciences), anti-SET8 (Active Motif) or anti-α-tubulin 9 (Sigma-Aldrich), respectively. gel was dried and exposed to autoradiography film for 1 week. For the peptide assays, the 21 specific peptides of α-tubulin were synthesized from AnaSpec Inc. Sequences are listed in 22 Appendix

Mass-spectrometric analysis 29
For identification of proteins in the tubulin preparations, >97% purified porcine tubulin 30 (rPeptide, # T-1201-1) and >99% purified bovine tubulin (MP-Bioscience, #0877112) were 31 separated by electrophoresis through a 10-20% Tris-Glycine polyacrylamide gel. The gel was 1 stained with Coomassie Blue; sections of the gels around 55 kDa, and below 55 kDa, 2 respectively, were excised (Appendix Fig S1A). Excised gel band were analyzed by the Taplin 3 Mass Spectrometry Facility, Harvard Medical School. For identification of tubulin 4 modifications, purified tubulin (MP-Bio, #0877115) was incubated with nonradioactive AdoMet, 5 with or without recombinant GST-SET8, overnight at room temperature and the samples were 6 separated by electrophoresis through a 10% Tris-glycine gel. Excised gel bands were digested 7 with either subtilisin or trypsin in 0.01% ProteaseMax in 50 mM NH4HCO3 for 1 hr at 50°C.   GST-pull down analysis of recombinant, purified His-tagged LSF to purified α-tubulin 1 fused to GST. Gels are as described in A, except that the protein gel was stained with 2 Coomassie. 3 D Plasmids expressing 3xFLAG-LSF and GFP-SET8 were transfected into COS7 cells. 4 Anti-FLAG antibody was visualized with a red fluorescing secondary antibody, and DNA 5 was visualized with DAPI. The merged image indicates colocalization of GFP-SET8 with 6 FLAG-LSF (yellow), concentrated largely near the nuclear membrane (Manders 7 correlation coefficient of LSF and SET8 colocalization is 0.9, as determined via the 8 Image J 3D analysis). The majority of overexpressed 3XFlagLSF was cytoplasmic with 9 only a minority detected in the nucleus. 10 E Specific co-immunoprecipitation of endogenous SET8 (top) and endogenous α-tubulin 11 (bottom) from HEK293 cellular extracts, using antibodies to LSF as compared to control 12 IgG. 13 F Specific co-immunoprecipitation of endogenous LSF from HEK293 cellular extracts, 14 using antibodies to SET8 as compared to control IgG. 15 G Immunoblotting of purified porcine brain tubulin (rPeptide, >97%) shows the presence of 16 LSF, using a LSF monoclonal antibody. Representative also of results obtained using a 17 GST-SET8. The higher relative levels of GST-SET8 to α/β-tubulin in this experiment led 26 to greater initial automethylation of SET8 relative to tubulin methylation. # indicates the 27 migration of [ 3 H]-labeled impurities. Bottom: Coomassie-staining of the same gel (shown 28 in grayscale) indicating relative levels of the components in the reaction. As in Fig 2B,  29 protein bands <50 kDa are from the purified GST-SET8 preparation, and are more 30 evident in this experiment. 31

B
Co-immunoprecipitation of endogenous α-tubulin with endogenous LSF from HEK293T 1 cell lysates was disrupted upon treatment of the cells with 2.5 μM FQI1 for 24 hr.