Dissecting the roles of Haspin and VRK1 in Histone H3 threonine-3 phosphorylation during mitosis

Protein kinases that phosphorylate histones are ideally-placed to influence the behavior of chromosomes during cell division. Indeed, a number of conserved histone phosphorylation events occur prominently during mitosis and meiosis in most eukaryotes, including on histone H3 at threonine-3 (H3T3ph). At least two kinases, Haspin and VRK1 (NHK-1/ballchen in Drosophila), have been proposed to carry out this modification. Phosphorylation of H3 by Haspin has defined roles in mitosis, but the significance of VRK1 activity towards histones in dividing cells has been unclear. Here, using in vitro kinase assays, KiPIK screening, RNA interference, and CRISPR/Cas9 approaches, we were unable to substantiate a direct role for VRK1, or its homologue VRK2, in the phosphorylation of threonine-3 or serine-10 of Histone H3 in mitosis, although loss of VRK1 did slow cell proliferation. We conclude that the role of VRK1, and its more recently identified association with neuromuscular disease in humans, is unlikely to involve mitotic histone kinase activity. In contrast, Haspin is required to generate H3T3ph during mitosis.


Introduction
The structure and function of chromosomes undergo radical changes during cell division, including chromosome condensation, downregulation of gene expression, removal of cohesion between chromosomes, and the assembly of kinetochores at centromeres to drive chromosome segregation.
Protein kinases are instrumental in bringing about these alterations, and kinases that act directly on histones are well-placed to influence the behavior of chromatin. Indeed, a number of histone phosphorylation events occur prominently during mitosis and meiosis in most eukaryotes including yeasts, plants and animals (1). Among these are phosphorylation of histone H3 at threonine-3 (H3T3ph) and serine-10 (H3S10ph), and histone H2A at threonine-120 (H2AT120ph; or the equivalent H2AT119ph in Drosophila). For example, in mammalian cells, H3T3ph first becomes detectable at foci on chromosome arms in late G2/early prophase, concentrates most strongly at inner centromeres in prometaphase and metaphase, and declines during anaphase (2,3). A key function of H3T3ph is to position the Chromosomal Passenger Complex (CPC) correctly at inner centromeres (4)(5)(6). The CPC contains the kinase Aurora B that regulates chromosome segregation by modulating the phosphorylation status of a number of centromere and kinetochore substrates (7).
In addition, H3T3ph may participate in a phospho-methyl switch to prevent histone reading proteins such as the transcription factor TFIID from binding to the adjacent modification H3K4me3 during mitosis (8).
Central to understanding the role of these histone phosphorylation events is knowledge of the kinases responsible. Two kinases have been proposed to carry out phosphorylation of H3T3 during mitosis. One of these is Haspin, an atypical kinase (encoded by Germ Cell-Specific Gene-2, GSG2) that phosphorylates H3 peptides, recombinant H3, and nucleosomes, specifically at threonine-3 in vitro (9,10). Haspin RNAi, inhibitors, or CRISPR/Cas9 targeting all remove mitotic H3T3ph in cultured mammalian cells, supporting the view that this is a critical kinase for H3T3ph during cell division (9,(11)(12)(13)(14). Loss or inhibition of Haspin impairs chromosome alignment and segregation in cell culture (9,12,15), particularly when cells are transiently blocked in mitosis (14), or when another kinase that helps recruit Aurora B to centromeres, Bub1, is also inhibited (16,17). In contrast, overall chromosome condensation is not noticeably altered by lowered Haspin activity (9).
The relative roles of Haspin and VRK1 in H3T3 phosphorylation remain to be defined. Some authors find that both VRK1 and Haspin are required for full H3T3ph in mitotic cells (21,24). Others have proposed that VRK1 is essential to trigger generalized H3T3ph in early mitosis, and that Haspin acts only later in mitosis. In this model, VRK1 phosphorylates H3T3ph to condense chromosomes and recruit the CPC, and Aurora B then activates Haspin to generate localized H3T3ph at centromeres (23,36). This work contrasts with other models in which the timing and location of H3T3ph is controlled indirectly by Cdk1, Bub1, Plk1 and Aurora kinases acting through Haspin and the H3T3ph phosphatase Repo-Man/PP1 (37)(38)(39)(40)(41)(42)(43). VRK1 mutations in humans appear to cause neurodegenerative disorders (30, 44,45) and, consistent with this, VRK1 GT3/GT3 hypomorphic mice have reduced brain weight and mild motor dysfunction (46). These findings reinforce the need to understand the molecular functions of VRK1. In this study, we wished to clarify the contribution of Haspin and VRK1 to H3T3ph in mitosis.

Recombinant Haspin phosphorylates H3T3 more efficiently than VRK1 in vitro
To compare the activity of human Haspin and human VRK1 towards Histone H3T3, we conducted in vitro kinase assays using recombinant Haspin (residues 471-798), recombinant full-length VRK1, and peptides corresponding to the first 21 amino acids of H3. Three different preparations of recombinant VRK1, produced in both E. coli or in Sf9 insect cells, were similarly able to phosphorylate H3(1-21) at the high concentration of 10 nM. However, Haspin generated H3T3ph more strongly than VRK1, even at 0.1 nM ( Figure 1A). VRK1 has been reported to phosphorylate nucleosomes more effectively than free histones (47), so we also compared the activity of Haspin and VRK1 on purified recombinant nucleosomes. In this case, Haspin but not VRK1 was able to phosphorylate H3T3 ( Figure 1B). Therefore, in these in vitro assays, Haspin was at least 100-fold more active towards H3T3 than was VRK1.

A kinase activity screen implicates Haspin, not VRKs, as the mitotic H3T3 kinase
Recombinant kinases may lack cofactors (such as post-translational modifications or activating subunits) that are required for activity towards their substrates. To identify H3T3 kinases in a more physiological environment, and in an unbiased fashion, we made use of a new kinase screening methodology known as KiPIK (Kinase inhibitor Profiling to Identify Kinases) that uses cell extracts as a source of kinases (48). The method compares the pattern of inhibition of the unknown kinase activity in cell extracts with the inhibition profiles of known protein kinases determined in vitro using small molecule kinase inhibitors. We determined which kinase(s) in mitotic HeLa cell extract were able to phosphorylate H3(1-21) peptides, detected with anti-H3T3ph antibodies. To generate an inhibition fingerprint for H3T3 kinase activity in extract, we used a library of 140 kinase inhibitors that have been profiled on numerous purified kinases in vitro (Supplemental Table 1). This included 84 of the 178 inhibitors profiled at 0.5 µM on 300 kinases by Anastassiadis et al. (49), and 63 of the 158 inhibitors profiled at 1 µM or 10 µM on 234 kinases by Gao et al. (50), all using conventional in vitro kinase assays. In addition, it included all 72 inhibitors profiled on 388 different kinases by Davis et al. using the DiscoverX competition binding assay (51), and 50 of the 156 inhibitors tested at 10 µM for binding to 60 kinases by Federov et al. using a thermal denaturation assay (52). Together, this incorporates 444 kinases (405 unique human kinase genes; Supplemental Table 2).
The results showed that, in every case where Haspin was present in the profiling dataset, it was the top hit in the screen (Figure 2A-E, red circles). By contrast, no VRK kinase scored within the top 40% of kinases ( Table 3). These results are consistent with previous KiPIK screens for H3T3 kinases (48), where Haspin was the top overall hit, but VRK1 and VRK2 were never within the top 30% and were most often in the bottom 20% of kinases ( Figure 2F; Figure S1). Of note, another kinase that has been proposed to phosphorylate a number of sites on H3 including H3T3, PASK (53), was also not a notable hit in these screens ( Figure 2, Figure S1). Therefore, Haspin was reproducibly identified as the most likely mitotic H3T3 kinase in HeLa cell extracts.
A kinome-wide RNAi screen implicates Haspin, not VRKs, as the H3T3 kinase in mitosis KiPIK screening uses cell extracts and appears to be relatively insensitive to the effects of upstream kinases in signaling cascades, and therefore is useful for identifying direct kinases for specific substrates (48). As an alternative unbiased approach to screen for H3T3 kinases in a cellular environment, and in manner that would allow the possible indirect effect of kinases to be observed, we analysed the results of a kinome-wide RNAi screen for H3T3 kinases in mitotic HeLa cells (48).
Haspin was the top hit in this screen, and kinases known to regulate Haspin and the H3T3ph phosphatase Repo-Man-PP1 (i.e. Aurora B and Plk1 (37-39,41)) were also present in the top 5 kinase hits ( Figure 3A). By contrast, none of the VRK kinases (nor PASK) were in the top 100 of the approximately 500 kinases tested. Therefore, RNAi screening also favours the hypothesis that Haspin, but not VRKs, are involved in the regulation of H3T3 in mitotic HeLa cells.

Targeted RNAi for VRK1 does not reveal a role in H3T3 phosphorylation
Although the RNAi screen uses multiple pre-validated siRNAs against each kinase, we could not be certain that VRK1 was depleted in these experiments. We therefore conducted targeted experiments to deplete VRK1 from HeLa cells by RNAi. We were able to reduce the amount of VRK1 by approximately 80% +/-7% (n = 3). In immunoblotting experiments, this did not detectably reduce the extent of H3T3ph in asynchronous cells or in cells arrested in mitosis with the microtubule destabilizing inhibitor nocodazole ( Figure 3B). Immunofluorescence microscopy also confirmed that VRK1 depletion did not detectably alter H3T3ph in mitosis ( Figure 3C). By contrast, in HeLa cells lacking Haspin due to CRISPR/Cas9 targeting (14), H3T3ph could not be detected by immunoblotting or immunofluorescence microscopy ( Figure 3B, 3C). As expected (9,12,13), loss of Haspin did not alter the extent of H3S10ph in HeLa cells detected by immunoblotting ( Figure 3B) or immunofluorescence ( Figure 3C). Consistent with this, loss of Haspin, but not VRK1, prevented detection of H3T3ph in mitotic cells containing H3S10ph ( Figure 3D). VRK1 depletion caused a partial decrease in H3S10ph in immunoblotting experiments ( Figure 3B), but not in individual mitotic cells as determined by immunofluorescence microscopy ( Figure 3C). This is likely to be due to a small decrease in the proportion of VRK1 depleted cells blocked in mitosis by nocodazole compared to control cells (see below), although progression through mitosis itself was not obviously altered in asynchronously growing cells ( Figure S2). Therefore, using RNAi in HeLa cells, we could not uncover a role for VRK1 in phosphorylation of H3T3, or H3S10.

CRISPR/Cas9 targeting of VRK1 does not reveal a role in H3T3 phosphorylation
We were unable to remove all VRK1 from cells using RNAi, leaving open the possibility that low amounts of the kinase are able to maintain its putative role in H3 phosphorylation. To address this, we compared the effects of CRISPR/Cas9 targeting of Haspin and VRK1 in the haploid cell line HAP1 (see Figure S3). As expected, Haspin but not VRK1 could be detected in HAP1 VRK1 KO cells, and VRK1 but not Haspin could be seen in HAP1 Haspin KO cells ( Figure 4A). As in HeLa cells, loss of Haspin eliminated H3T3ph detectable by immunoblotting of nocodazole-arrested mitotic HAP1 cells, but did not markedly affect H3S10ph ( Figure 4B). By contrast, both H3T3ph and H3S10ph remained present in mitotic cells lacking VRK1 ( Figure 4B). The partial decline in both H3T3ph and H3S10ph  Figure S5A). These experiments also supported the view that progression through mitosis was not substantially altered by loss of Haspin or VRK1 ( Figure S5B). These results reveal that eliminating detectable VRK1 from HAP1 cells does not prevent normal H3T3 (or HS10ph) phosphorylation in mitosis.

RNAi does not reveal a role for VRK2 in H3T3 phosphorylation
Human cells contain a second VRK1-related protein with kinase activity, VRK2. To determine if H3T3ph remaining in HAP1 VRK1 KO cells could be due to the presence of VRK2, we depleted VRK2 from these cells by RNAi. We were able to reduce VRK2 protein expression levels by 84% +/-10% (n = 3). In immunoblotting experiments, this did not detectably reduce the extent of H3T3ph or H3S10ph in cells arrested in mitosis with the microtubule destabilizing inhibitor nocodazole ( Figure   6), suggesting that VRK2 is not responsible for the persistence of H3T3ph in HAP1 VRK1 KO cells. As discussed earlier, HAP1 VRK1 KO cells had slightly reduced levels of H3T3ph, H3S10ph, and MPM2 phosphoepitopes, most likely explained by the decline in the proportion of mitotic cells.

Discussion
Here, we find that Haspin is a much more potent kinase than VRK1 for H3T3 in vitro. In addition, using KiPIK assays in mitotic HeLa cell extracts, none of the VRK family kinases are found to be strong candidates for the H3T3 kinase in mitosis while, in contrast, Haspin is the highest ranked kinase.
Similarly, in a kinome-wide RNAi screen for mitotic H3T3ph kinases, Haspin was the top hit, while none of the VRKs were in the top 100. CRISPR/Cas9 targeting of Haspin, but not VRK1, causes loss of detectable H3T3ph during mitosis in HAP1 cells. Furthermore, depleting VRK2 in combination with loss of VRK1 does not lead to detectable reductions in H3T3ph, suggesting that redundancy with VRK2 does not underlie our failure to influence H3T3ph when VRK1 expression is ablated. VRK3 is unlikely to contribute as it is a pseudokinase with a degraded active site structure (31) that is unable to bind divalent cations or ATP (54), and is unable to phosphorylate casein or H3 (20,55). In addition, PASK did not score highly as a mitotic H3T3 kinase in either KiPIK or RNAi screens, concordant with the finding that CRISPR/Cas9 targeting of the PASK gene did not markedly alter H3T3ph in mouse C2C12 cells (53).
These results suggest that Haspin is the major H3T3 kinase during mitosis in human cells. This is consistent with reports that Haspin is required for normal mitotic H3T3ph in mouse cells (3), sea urchin embryos (56), fruit flies (57,58), Arabidopsis (59,60), and fission yeast (6), and for H3T3ph during meiosis in mouse oocytes (61,62) and frog egg extracts (5). Indeed, Haspin orthologues are present in all organisms that are reported to have H3T3ph in mitosis to date. By contrast, VRK family kinases are present in many metazoa (eg VRK-1 in C. elegans, and NHK-1/ballchen in Drosophila) but appear absent from plants and yeasts (28,55).
It remains uncertain why previous experiments have suggested a role for VRK1 in producing H3T3ph in mitosis. However, multiple reports indicate that VRK1 influences cell cycle progression. VRK1 regulates the expression of Cyclin D1, and VRK1 depletion impedes progression from G1 to S phase in cultured human cells (35, [63][64][65]. This could hinder cells reaching mitosis and so lead to a loss of modifications that are abundant during this time, such as H3T3ph (and H3S10ph). Indeed, although it is clear that HAP1 cells lacking detectable VRK1 can proliferate in culture, live imaging revealed that a lower proportion of VRK1 KO cells enter mitosis per unit time, consistent with a delay in the cell cycle outside mitosis, such as during G1. In addition, a previous report found that a high proportion of HeLa cells die in mitosis following VRK1 RNAi (66). The loss of H3 phosphorylation seen in previous immunoblotting experiments that used nocodazole to accumulate mitotic cells, therefore, may be an indirect effect of a reduced mitotic index. By contrast, in experiments where the intensity of H3T3ph (and H3S10ph) in individual mitotic cells was measured, we found no evidence for a decline in these phosphorylation marks.
Analysis of VRK1 function in cell division in model organisms is complicated because complete loss of function is lethal in mice, flies and worms (34,67,68), so only hypomorphic alleles or the results of partial depletions have been examined. Male and female VRK1 GT3/GT3 gene-trapped mice are sterile, indicating defects in gametogenesis. Indeed, there is a strong defect in proliferation of spermatogonial cells in male mice, commensurate with their reduced levels of cell cycle markers such as Proliferating Cell Nuclear Antigen (PCNA). The reduction in H3S10ph in these testes is most likely due to the reduced numbers of proliferating cells, not a specific loss of H3S10ph due to reduced VRK1 function (32). Consistent with this, sterility of female VRK1 GT3/GT3 mice appears to be due to meiotic defects in oocytes, but these oocytes have normal levels of H3S10ph (69). Putting together these findings, the reason that alterations in VRK1 activity give rise to different cell cycle defects in different systems may be due to the different mutants analysed and varying degrees of depletion obtained in different experiments. In addition, the extent to which other kinases such as VRK2 can act redundantly may vary depending on the organism, tissue or function in question.
Indeed, there is a second VRK-related kinase in Drosophila, CG8878, although it has a distinctive split kinase domain and its molecular function remains unclear (73). Nevertheless, there is strong evidence that VRK1 deficiency affects both mitosis and meiosis. However, our results suggest these effects are unlikely to be due to changes in H3T3ph phosphorylation. In addition, H3S10ph does not appear to be directly affected by loss of VRK1, in line with other studies discussed above (32,71,72).
Interestingly, fly NHK-1 and human VRK1 have also been reported to phosphorylate nucleosomal H2AT120 (H2AT119 in flies) (47,65). Phosphorylation of this site is clearly elevated in mitosis and meiosis in human cells, flies, budding and fission yeast (47,70,74,75). A female-sterile mutation of NHK-1 diminishes H2AT119ph in fly oocytes (70), but whether this is a direct or indirect effect of NHK-1 on H2A phosphorylation remains unknown. Ooctyes of VRK1 GT3/GT3 mice have normal levels of H2AT120ph (69). Moreover, neither mutation nor depletion of NHK-1 prevents mitotic H2AT119ph in flies or Drosophila S2 cells (70,74). Instead, the kinetochore-binding kinase Bub1 is clearly vital for H2AT120ph in budding and fission yeast, mice, and human cells (75)(76)(77). Together, there is little evidence to support a significant role for VRK1 in direct phosphorylation of H3T3ph or other histone residues in mitosis.
VRK1 phosphorylates and inhibits Barrier-to-Autointegration Factor (BAF) to regulate nuclear envelope dissolution during mitosis and meiosis in flies, worms and mammalian cells (66,68,71,72,78,79), and this seems likely be a major contributor to its cell division phenotypes. In addition, we have not ruled out a role for VRK1 in histone phosphorylation outside mitosis. For example, worm VRK-1 is expressed in post-mitotic cells and can influence lifespan by activating AMP-activated protein kinase (80), and Aihara et al. have suggested that VRK1 regulates gene expression by phosphorylating H2AT120 at gene promoters such as CCND1/Cyclin D1 in mice (65). A number of H3S10 kinases that act to regulate transcription in interphase cells have been characterized, including MSK1/2 (28,33). We are not aware of H3T3 kinases that act during interphase in vertebrates, but Haspin itself may have interphase roles in Drosophila (58), and the plant-specific MUT9 kinases generate H3T3ph in Chlamydomonas and Arabidopsis and appear to be involved in epigenetic silencing (81,82). Interestingly, like the VRK family, MUT9 kinases are part of the CK1 superfamily and it is conceivable that VRKs serve a similar function in metazoa. It also will be interesting in future to determine if the mitotic or non-mitotic functions of VRK1 underlie the neuromuscular disorders observed in mice and human patients (30,44-46).

Cell counting
HAP1 cell lines were counted using trypan blue (Sigma) and a haemocytometer, seeded at 100,000 cells/well in 6-well plates and recounted at 24 h intervals.

Quantification of immunofluorescence
To quantify immunofluorescence intensities in individual cells, and to determine mitotic indices, images were analysed in Fiji 2.0 (83). Cells were identified from DAPI staining by automatic thresholding of maximum intensity projections and binarization followed by use of the watershed tool to separate adjacent cells. Cells were then counted using the analyse particles tool with a minimum cut off of 20 µm 2 . The fluorescence intensities for H3T3ph and H3S10ph in each cell was quantified using sum intensity projections. Mitotic cells were identified as those whose fluorescence intensity for H3S10ph classified them as outliers using the iterative Grubbs' method ( = 0.0001) in Prism 9.0.2 (GraphPad). Visual inspection of DAPI stained cells confirmed that, for H3S10ph, this included essentially all mitotic cells. Automatically identified particles containing more than one mitotic cell were eliminated from the analysis. Background fluorescence was determined from the average fluorescence intensity of all non-mitotic cells. MPM2 positive cells were identified similarly with the additional use of the fill holes tool prior to watershedding.

Live imaging
Live imaging was performed in glass-bottomed FluoroDishes (WPI) in IMDM medium (12440053, ThermoFisher). DNA was stained with 25 nM SiR-DNA (Spirochrome). DIC and SiR-DNA images at 6 zsteps per field (2 µm step size) were acquired at 4 min intervals on a Nikon A1R confocal microscope equipped with a 20x 0.75 NA PlanApo VC DIC objective using Elements v5.22 software (Nikon, Japan). In a blinded fashion, the time of mitotic entry was determined as the earliest time point at which the initiation of cell rounding was observed by DIC and/or the time point prior to nuclear envelope breakdown observed by SiR-DNA staining, and the time of mitotic exit was defined as the first frame in which anaphase was apparent.
After washing, 40 µl/well of HRP-conjugated anti-rabbit IgG antibodies (Cell Signaling Technology, #7074) in TBS, 0.1% Tween20, 5% BSA were added for 1 h at room temperature. After washing, HRPconjugated antibody binding was detected using TMB substrate (New England Biolabs) according to the manufacturer's instructions. Pipetting was performed with a Biomek FX liquid handling robot (Beckman Coulter). Absorbance readings were made using a Polarstar Omega microplate reader (BMG Labtech).

KIPIK results analysis
Using the mean of duplicate determinations, a standard score reflecting kinase activity in the presence of each inhibitor was calculated where standard score = (mean absorbance in presence of inhibitor -mean absorbance of multiple DMSO controls)/standard deviation of DMSO controls. The inhibition by each compound was defined as: %inhibition = 100 × (standard score for inhibitor/lowest standard score on plate). The %inhibition scores for all inhibitors were then compiled to produce the inhibition fingerprint of H3T3 phosphorylation. Using Excel (Microsoft), Pearson's correlation () was then calculated for this fingerprint against the inhibition profiles of each of the kinases profiled in vitro against that inhibitor library, excluding mutant kinases. We calculated z-scores where z = (observed  -mean  of null distribution) / standard deviation of null distribution as described (48).

Statistical Analysis
Statistical analyses were carried out using Prism 9.0.2 (Graphpad).

Conflict of Interest
Jonathan Higgins is an inventor on international patent filing PCT/GB2020/051073 "Kinase screening assays" that covers the KiPIK procedure. The other authors declare that they have no conflict of interest.    Some results of this screen were previously published (48).         α-tubulin 50 -