Kinetochore protein depletion underlies cytokinesis failure and somatic polyploidization in the moss Physcomitrella patens

Lagging chromosome is a hallmark of aneuploidy arising from errors in the kinetochore–spindle attachment in animal cells. However, kinetochore components and cellular phenotypes associated with kinetochore dysfunction are much less explored in plants. Here, we carried out a comprehensive characterization of conserved kinetochore components in the moss Physcomitrella patens and uncovered a distinct scenario in plant cells regarding both the localization and cellular impact of the kinetochore proteins. Most surprisingly, knock-down of several kinetochore proteins led to polyploidy, not aneuploidy, through cytokinesis failure in >90% of the cells that exhibited lagging chromosomes for several minutes or longer. The resultant cells, containing two or more nuclei, proceeded to the next cell cycle and eventually developed into polyploid plants. As lagging chromosomes have been observed in various plant species in the wild, our observation raised a possibility that they could be one of the natural pathways to polyploidy in plants.


26
The kinetochore is a macromolecular complex that connects chromosomes to spindle 27 microtubules and plays a central role in chromosome segregation. Kinetochore malfunction 28 causes checkpoint-dependent mitotic arrest, apoptosis, and/or aneuploidy-inducing 29 chromosome missegregation (1). Most of our knowledge on kinetochore function and impact 30 on genome stability is derived from animal and yeast studies (2). Another major group of 31 eukaryotes, plants, also possesses conserved kinetochore proteins (3-5). Although the 32 localization and loss-of-function phenotype of some plant kinetochore proteins have been 33 reported before (6-15), the data are mostly obtained from fixed cells of specific tissues. No 34 comprehensive picture of plant kinetochore protein dynamics and functions can be drawn as 35 of yet. For example, 12 out of 16 components that form CCAN (constitutive centromere 36 associated network) in animal and yeast cells cannot be identified by homology searches (2, 37 5). How the residual four putative CCAN subunits act in plants is also unknown. 38 The moss Physcomitrella patens is an emerging model system for plant cell biology. The 39 majority of its tissues are in a haploid state, and, owing to an extremely high rate of 40 homologous recombination, gene disruption and fluorescent protein tagging of endogenous 1 genes are easy to obtain in the first generation (16). The homology search indicated that all 2 the P. patens proteins identified as the homologue of human kinetochore components are 3 conserved in the most popular model plant species A. thaliana (5): therefore, the knowledge 4 gained in P. patens would be largely applicable to flowering plants, including crop species. 5 Another remarkable feature of P. patens is its regeneration ability; for example, differentiated 6 gametophore leaf cells, when excised, are efficiently reprogrammed to become stem cells 7 (17, 18). Thus, genome alteration even in a somatic cell can potentially spread through the 8 population. 9 In this study, we aimed to comprehensively characterize conserved kinetochore proteins in a 10 single cell type, the P. patens caulonemal apical cell. We observed that many proteins 11 displayed localization patterns distinct from their animal counterparts. Furthermore, 12 kinetochore malfunction led to chromosome missegregation and microtubule disorganization 13 in the phragmoplast, eventually resulting in cytokinesis failure and polyploidy. 14 15

16
Endogenous localization analysis of conserved kinetochore proteins in P. patens 17 To observe the endogenous localization of putative kinetochore components, we inserted a 18 fluorescent tag in-frame at the N-and/or C-terminus of eighteen selected proteins, which 19 contain at least one subunit per sub-complex (Figure 1Figure supplement 1). Initially we 20 conducted C-terminal tagging since the success rate of homologous recombination is much 21 higher than N-terminal tagging (19). For ten proteins, function was unlikely perturbed by 22 tagging, as the transgenic moss grew indistinguishably from wild-type, despite the single-23 copy protein being replaced with the tagged protein. For other seven proteins, the 24 functionality of the tagged version could not be verified, since untagged paralogs are present 25 in the genome. The C-terminal tagging line for CENP-S could not be obtained after two 26 attempts, suggesting that tagging affected the protein's function and thereby moss viability. 27 The N-termini of CENP-S, CENP-O, and CENP-C were also tagged with Citrine. Among 28 them, no paralogous proteins could be identified for CENP-S or CENP-C; therefore, Citrine 29 signals would precisely represent the endogenous localization. Exceptionally, histone H3-like 30 CENP-A (CenH3) localization was determined by ectopic Citrine-CENP-A expression, as 31 tagging likely perturbs its function. 32 Consistent with their sequence homology, many of the proteins were localized to the 33 kinetochore at least transiently during the cell cycle. However, multiple proteins also showed 34 unexpected localization (or disappearance) at certain cell cycle stages (Figure 1Figure   Kinetochore malfunction causes chromosome missegregation and cytokinesis failure 41 We failed to obtain knockout lines and/or induce frameshift mutation using CRISPR/Cas9 for 42 the single-copy kinetochore proteins, except for the spindle checkpoint protein Mad2, 43 strongly suggesting that they are essential for moss viability. We therefore made conditional 44 RNAi lines, targeting different proteins from both inner and outer kinetochores (summarized 45 in Figure supplement 1). In this RNAi system, knockdown of target genes was induced by the 46 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; addition of β-estradiol to the culture medium 4-6 days prior to live-imaging (20). Since 1 RNAi sometimes exhibits an off-target effect, we prepared two independent RNAi constructs 2 for most target genes. Following the previously established protocol (20, 21), we screened for 3 cell growth/division phenotypes in ≥10 transgenic lines for each construct by using long-term 4 (>10 h) fluorescent imaging. We observed mitotic defects in multiple RNAi lines, such as 5 delay in mitotic progression, chromosome missegregation and/or multi-nuclei; these 6 phenotypes were never observed in the control line ( Figure 2A, BVideo 5). A full list of 7 targeted genes and brief descriptions of the observed phenotypes are provided in Figure   8 supplement 1. 9 We first selected CENP-A for detailed analysis, the only constitutive centromeric protein 10 identified in P. patens. As expected, we observed a significant mitotic delay and chromosome 11 alignment/segregation defects in the CENP-A RNAi lines (Figure 2Figure supplement 8; 12 Video 6). These phenotypes can be explained by a deficiency in proper kinetochore- 13 microtubule attachment. Consequently, micronuclei were occasionally observed in the 14 daughter cells, a hallmark of aneuploidy. We concluded that CENP-A, like in many 15 organisms, is essential for equal chromosome segregation during mitosis in moss. 16 Surprisingly, we also frequently observed cells with two large nuclei in both RNAi lines  Although we could not detect any kinetochore enrichment of the CCAN subunit CENP-X, we 31 analyzed its RNAi lines. Interestingly, we observed similar phenotypes to CENP-A and 32 SKA1, including cytokinesis failure (Figure 2BFigure supplement 8; Video 6). CENP-X 33 RNAi phenotypes were rescued by the ectopic expression of CENP-X-Cerulean that was 34 resistant to the RNAi construct (Figure 2Figure supplement 9). Thus, CENP-X has lost its 35 kinetochore localization in moss, but is still essential for chromosome segregation and cell 36 division. 37 By analyzing a total of 44 cells from SKA1 (9 cells), CENP-X (18 cells) and CENP-A RNAi 38 (9 cells for one construct and 8 cells for the other) lines that had lagging chromosomes, we 39 noticed a correlation between cytokinesis failure and lagging chromosomes lingering for a 40 relatively long time in the space between separated chromatids. We therefore quantified the 41 duration of lagging chromosomes' residence in the midzone between separating chromatids 42 following anaphase onset. Interestingly, a minor delay of chromosomes in the midzone (< 4 43 min) never perturbed cytokinesis (100%, n = 9 for CENP-A, n = 4 for CENP-X and n = 3 for 44 SKA1). By contrast, if we observed a longer delay of chromosome clearance from the 45 midzone, even when only a single chromosome was detectable, cytokinesis defects occurred 46 in 96% of the cells (n = 9, 14 and 5; Figure 2C, D).

47
. CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; During plant cytokinesis, a bipolar microtubule-based structure known as the phragmoplast is 1 assembled between segregating chromatids. The cell plate then forms in the phragmoplast 2 midzone (~4 min after anaphase onset in P. patens caulonemal cells) and gradually expands 3 towards the cell cortex, guided by the phragmoplast (22). We observed that microtubules 4 reorganized into phragmoplast-like structures upon chromosome segregation in every cell, 5 regardless of the severity of chromosome missegregation (e.g. 32 min in Figure 2B). 6 However, high-resolution imaging showed that microtubule interdigitates at the phragmoplast 7 midzone were abnormal in the kinetochore RNAi lines. In 5 out of 7 control cells, a sharp 8 microtubule overlap indicated by bright GFP-tubulin signals was observed during 9 cytokinesis, as expected from previous studies (22, 25) (yellow arrowhead in Figure 2E). In 10 contrast, CENP-A and SKA1 RNAi lines that had lagging chromosomes and eventually 11 failed cytokinesis never exhibited such focused overlaps (0 out of 12 cells); instead, the 12 overlap was broader and less distinguished ( Figure 2E). 13 Finally, we checked if the cell plate was formed at any point in the cells that had cytokinesis 14 defects, using the lipophilic FM4-64 dye. We could not observe vesicle fusion at the midzone 15 following anaphase onset; thus, the cell plate did not form in the cells that had lagging CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; we isolated and cultured several cells ( Figure 3C) that were seemingly multi-nuclear after 1 SKA1 RNAi via laser dissection microscopy (note that there is an unambiguity in identifying 2 multi-nucleate cells; see Methods for detailed explanation). After 6 weeks of culturing 3 without β-estradiol (i.e. RNAi was turned off), we obtained four moss colonies, two of which 4 consisted mainly of protonemal cells with a few gametophores ( Figure 3D, colony 3 and 4). Overall, the behavior of outer subunits was largely consistent with their animal counterparts, 19 suggesting that the mitotic function is also conserved. However, the timing of kinetochore CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; turnover and vesicle trafficking that is required for phragmoplast formation (35-38). By 1 contrast, in our study, both phenotypes were observed after RNAi treatment of CENP-A, a 2 constitutive centromeric histone protein that is unlikely to play a direct role in cytokinesis. 3 Furthermore, the cytokinesis phenotype frequently appeared in RNAi lines targeting other six 4 kinetochore proteins, and only when lagging chromosomes were present. Based on these 5 data, we propose that persistent lagging chromosomes cause cytokinesis failure. Lagging 6 chromosomes might act as physical obstacles to perturb phragmoplast microtubule 7 amplification and/or cell plate formation. Alternatively, persistent lagging chromosomes 8 might produce an unknown signal or induce a certain cell state that inhibits phragmoplast 9 expansion and/or cell plate formation in order to prevent chromosome damage, reminiscent 10 of the NoCut pathway in animal cytokinesis (39, 40). We favor the latter model, as abnormal 11 microtubule interdigitates were observed in the whole phragmoplast and not limited to the 12 region proximal to the lagging chromosome ( Figure 2E). Notably, in a recent study, 13 cytokinesis in moss protonema cells could be completed despite longer microtubule overlaps 14 (41). It suggests that abnormal microtubule interdigitates represent the consequence of 15 microtubule dynamics mis-regulation rather than the direct cause of cytokinesis failure. 16 Our data further suggest that, in P. patens, chromosome missegregation in a single cell could

32
Moss culture and transformation 33 We generally followed protocols described by Yamada et al (19). In brief, Physcomitrella 34 patens culture was maintained on BCDAT medium at 25°C under continuous light. 35 Transformation was performed with the polyethylene glycol-mediated method and successful 36 endogenous tagging of the selected genes was confirmed by PCR (19). We used P. patens 37 expressing mCherry-α-tubulin under the pEF1α promoter as a host line, except for Mis12-38 mCherry line where GFP-α-tubulin line was used as a host line. For knockout, CRISPR (47) 39 and RNAi transformations, we used the GH line, expressing GFP-tubulin and HistoneH2B-40 mRFP. P. patens lines developed for this study are described in Dataset S1.

42
Plasmid construction 43 Plasmids and primers used in this study are listed in Dataset S2. For the C-terminal tagging, 44 we constructed integration plasmids, in which ∼800 bp C-terminus and ∼800 bp 3'-UTR 45 sequences of the kinetochore gene were flanking the citrine gene, the nopaline synthase 46 polyadenylation signal (nos-ter), and the G418 resistance cassette. For the N-terminal tagging 47 we constructed integration plasmids, in which ∼800 bp 5'-UTR and ∼800 bp N-terminus 48 sequences of the kinetochore gene were flanking the citrine gene. CENP-A cDNA was 49 amplified by PCR and sub-cloned into a vector containing the rice actin promoter, citrine 50 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; gene, the rbcS terminator, the modified aph4 cassette, and flanked by the genomic fragment 1 of the hb7 locus to facilitate integration. All plasmids were assembled with the In-Fusion 2 enzyme according to manufacturer's protocol (Clontech). RNAi constructs were made by 3 using the Gateway system (Invitrogen) with pGG624 as the destination vector (21). 4 5 DNA staining 6 We followed the protocol described by Vidali et al (48) with the following modifications: 7 sonicated moss was cultured for 6-7 days on the BCDAT plate, containing 5 µM β-estradiol 8 for RNAi induction and 20 µg/ml G418 to prevent contamination. Collected cells were 9 preserved in a fixative solution (2% formaldehyde, 25 mM PIPES, pH 6.8, 5 mM MgCl2, 1 10 mM CaCl2) for 30 min and washed three times with PME buffer (25 mM PIPES, pH 6.8, 5 11 mM MgCl2, 5 mM EGTA). Following fixation, cells were mounted on 0.1% PEI 12 (polyethyleneimine)-coated glass slides and subsequently incubated with 0.1% Triton X-100 13 in PME for 30 min and 0.2% driselase (Sigma-Aldrich) in PME for 30 min. Next, cells were very close association with each other, so that nuclear boundaries often overlap. We interpret 47 that two of four regenerated protonemata had haploid DNA content due to our unintentional 48 isolation of a single cell with a deformed nucleus rather than multi-nuclei. Next, a piece of 49 cellophane with single isolated cell was transferred from the glass-bottom dish to estradiol-50 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
2 3 Sequence analysis. 4 Full-size amino acid sequences of the selected proteins were aligned using MAFFT ver. 5 7.043 and then revised manually with MacClade ver. 4.08 OSX. We used the Jones- Taylor  CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; Full localization data can be found in Supplemental data. Some kinetochore signals are marked with yellow 5 arrowheads, whereas autofluorescent chloroplasts are all marked with white asterisks. Images were acquired at a 6 single focal plane. Bars, 5 µm. See Figure  . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; combined data (D). Asterisks indicate significant differences between two groups (lagging chromosomes 7 observed for short time or for several minutes) for two outcomes: cytokinesis complete and cytokinesis failure, 8 calculated individually for CENP-A; CENP-X and SKA1 RNAi lines (*P = 0.0476, ***P = 0.0003, ****P < 9 0.0001; Fisher's test; see Table supplement 1). Each data point corresponds to a single cell. Mean ± SD are 10 presented. (E) Representative images of the microtubule overlap in the phragmoplast in the control line (GH) 11 and in RNAi lines (CENP-A and SKA1) with lagging chromosomes. Note that microtubule overlaps appear 12 more broad and fuzzy in RNAi cells. Yellow arrow indicates microtubule overlaps, whereas cyan arrows point 13 to lagging chromosomes. Images were acquired with z-stacks and a single focal plane that best shows 14 microtubule overlaps is presented. Bar, 5 µm. 15 16 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; Quantification of the nuclear DNA content in the interphase nucleus of regenerated moss colonies, 11 corresponding to (C) and (D). 12 13 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; Summary of knockout, CRISPR/Cas9 frameshift ("-" indicates that frameshift mutations could not be obtained) 8 and RNAi experiments pursued in this study. HR stands for homologous recombination. "+" indicates 9 successful transgenic line selection. 10 11 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; chloroplasts are marked with yellow asterisks. Images were obtained at a single focal plane. CENP-A was 5 localized at the centromeric region throughout the cell cycle, whereas KNL2-Citrine was visible only during 6 interphase (red arrowheads). Bars, 5 µm. 7 8 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; with yellow asterisks. Images were obtained at a single focal plane. CENP-C was localized at the centromere 6 from G2 to telophase, whereas none of the other CCAN proteins showed punctate signals throughout the cell 7 cycle. CENP-O showed weak midzone localization from prometaphase to anaphase (arrowheads). Bars, 5 µm. 8 9 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; intensity plot of Citrine signals at the centromeres and at the non-centromeric region in the nucleus (background 8 measurement). Each line represents average relative fluorescent intensity of ≥ 6 centromeres or ≥ 6 non-9 centromeric regions inside the nucleus in a single cell (four cells analyzed for both Citrine-CENP-A and Citrine-10 CENP-C lines), measured every 15 min from the maximum Z-projection. Note that we could not identify 11 centromeric Citrine-CENP-C signals during ~2 h after mitotic exit, and therefore, the data are missing from the 12 graph. 13 14 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; Autofluorescent chloroplasts are marked with yellow asterisks. Images were acquired at a single focal plane. 5 Bars, 5 µm. 6 7 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; acquired at a single focal plane. Punctate Citrine signals appeared after prometaphase. Bars, 5 µm. 6 7 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; that kinetochore localization of Mad2 was more clearly observed following addition of the microtubule-5 depolymerizing drug (500 nM oryzalin) (E). Autofluorescent chloroplasts were marked with yellow asterisks. 6 Images were acquired at a single focal plane. Bars, 5 µm. 7 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; CENP-X or SKA1. "GH" is the control line. Bar, 5 µm. (B) Duration of mitosis (from NEBD to anaphase onset) 5 was calculated from high-resolution live-cell imaging data for each RNAi line and the control line (GH). Bars 6 indicate mean and SEM, whereas asterisks indicate significant differences compared with the control (*P < 7 0.04, ***P < 0.0007, ****P < 0.0001; two-tailed t-test). More than 20 cells were analyzed for each line. (C) 8 Frequency of chromosome missegregation in different RNAi lines. Chromosome missegregation defects were 9 classified into three types: chromosomes detached from the metaphase plate (detached chromosomes), lagging 10 chromosomes in anaphase (lagging chromosomes), and their combination. More than 20 cells were analyzed for 11 each line. 12 13 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/438648 doi: bioRxiv preprint first posted online Oct. 9, 2018; Live imaging of P. patens protonemal apical cells expressing SKA1-Cerulean (A) or CENP-X-Cerulean (C) in 4 the SKA1 5'UTR RNAi or CENP-X 5'UTR RNAi lines, respectively. RNAi was induced by addition of β-5 estradiol to the culture medium at the final concentration of 5 µM, 5-6 days prior to observation. Bar, 5 µm. (B, 6 D) Mitotic duration (from NEBD to anaphase onset) for each RNAi line with or without the rescue construct 7 (two independent SKA1 rescue lines [#3, #16] were analyzed). "GH" is the mother line used for RNAi 8 transformation. Bars indicate mean and SEM, whereas asterisks indicate significant differences (*P < 0.03, 9 ***P < 0.001, ****P < 0.0001; one-way ANOVA). More than ten cells were analyzed for each line. 10 11 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

Cytokinesis complete
Lagging chromosomes observed in the midzone for ≤ 4 min 0 3 Lagging chromosomes observed in the midzone for ≥ 6 min 5 1 . CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.