Functional divergence of TPX2 family members in Arabidopsis thaliana

TPX2 (Targeting protein for Xklp2) is an evolutionary conserved microtubule-associated protein important for microtubule nucleation and proper mitotic spindle assembly. The protein was also described as an activator of the mitotic kinase Aurora A in humans and the Arabidopsis AURORA 1 (AUR1) kinase. In contrast to animal genomes that encode only one TPX2 gene in their genome, higher plants genomes encode a family with several TPX2-LIKE gene members (TPXL). most closely We confirmed using in vitro assays functional kinase binding of

Eucalyptus grandis (19). However, the authors focused only on the presence of the TPX2_Xklp2 domain and did not take into consideration proteins with a TPX2_importin domain which is equally important for canonical TPX2 functions.
Here, we revise the phylogeny of the TPXL gene family. We identified a group of TPXL proteins with predicted Aurora kinase binding domain. In vitro kinase assays demonstrated that the Aurora binding domains of Arabidopsis TPXL homologs are able to activate recombinant AUR1. The TPXL members are characterized by different expression patterns and localization, suggesting diversification of TPXL genes for specific functions during plant development.

Results
The A. thaliana genome possesses 20 genes with similarity to TPX2 The EggNOG 4.5 database (20) was searched to discover proteins that contain typical TPX2-domains in Arabidopsis. In total, 20 proteins were identified, which can be classified into two groups based on their domain composition ( Figure 1A We performed a maximum likelihood phylogenetic analysis with the identified proteins. In agreement with the domain composition, the phylogeny ( Figure 1A; Additional File 3) revealed two major groups, A and B. Group A could be further divided into two clusters based on the absence (cluster I) or presence (cluster II) of a plant-specific KLEEK-motif.
The deepest clades (Additional File 2) comprised Arabidopsis orthologues of TPXL1, TPXL5, TPXL6 and TPXL7 (clades 1, 2 and 3 in (19)). It should be noted that TPXL7 formed a separate clade with only protein sequence from A. thaliana. Subsequently derived clade was composed of canonical TPX2 and TPXL2, TPXL3 and TPXL8. The canonical TPX2 clade (clade 3 after (19)) was subdivided into 5 groups representing metazoa, angiosperms, mosses, algae and fungi. Analyzed representatives of metazoa, algae and fungi contain a single copy of the canonical TPX2 (Additional File 3).
The group of plant-specific proteins with a KLEEK domain (WDLs) contains three major lineages. The first lineage consists of WDLs 5, 7, 9 and TPXL proteins. The second lineage contains WDLs 1, 2 and 3. Finally, the third lineage embraced WDLs 4, 5 and 6. Special attention should be paid to the moss Physcomitrella patens, which evolved a unique WDL paralog group (Additional File 3). Interestingly, the Aurora binding domain containing TPXL4 was grouped together with the WDL lineage (Additional File 3).
Each clade of plant TPX2 and TPXL proteins was divided into two sublades containing monocots and eudicots (Additional File 3), thus, there seems to be not a specific paralog for any of these groups.

Functional prediction of Arabidopsis Aurora binding domain
Activation of Aurora A kinases is a significant function of TPX2 proteins. Binding of TPX2 to human Aurora A activates the phosphorylation activity of the kinase and protects it from dephosphorylation by protein phosphatase 2A (6). Similarly, Aurora kinase binding domain of canonical Arabidopsis TPX2 is able to activate Aurora1 in vitro (15). Recently we showed that Aurora kinase binding domains of TPXL2 and TPLX3 also are able to activate Aurora1 in vitro (18). To characterize the Aurora binding domain of plant group A TPXL proteins, a multiple sequence alignment was performed with orthologues of Arabidopsis TPX2.
Additionally to TPX2 the Arabidopsis genome contains four genes, TPXL2, TPXL3, TPXL4 and TPXL8 with both TPX2_importin and putative Aurora kinase binding domains ( Figure 1).
The overall sequence conservation is very poor between human TPX2 and plant TPX2-like proteins, but the key residues important for binding of Aurora kinases are evolutionary conserved (Figure 2A, 2B). The Aurora binding domain of human TPX2 contains upstream and downstream helical stretches (6). Despite some amino acid substitutions in the upstream helical stretch of TPXL4/TPXL8 or TPXL2/TPXL3 compared to human TPX2, the hydrophobic side chain involved in stacking interactions is preserved (Figure 2A

All Arabidopsis TPX2 family members possessing an Aurora kinase binding domain activate AUR1 in vitro
The Aurora kinase binding domains of canonical Arabidopsis TPX2, TPXL2 and TPXL3 were shown to activate Aurora1 in vitro and increase its phosphorylation activity towards histone H3 as a physiological substrate (15,18). To address the functionality of the Aurora kinase binding domains of Arabidopsis TPXL proteins, we performed in vitro kinase assays.
The Aurora binding domains of TPXL2, TPXL3, TPXL4 and TPXL8 were expressed in E. coli, purified (Additional File 5A) and combined with recombinant AUR1 as enzyme and histone H3 as a substrate. An increase in histone H3 phosphorylation detected by incorporation of radioactive isotope P 32 into histone H3 was used as a means to measure the activity of AUR1 (26). The in vitro kinase assay showed that Aurora binding domains of all TPXL proteins are able to activate AUR1 kinase ( Figure 2C, Additional File 5B). TPXL3 has the highest activation potential with up to 5-fold increase compared to Aurora1 kinase alone, which was significantly higher than AUR1 activation by canonical TPX2 ( Figure 2C).
The eight closest homologues of canonical TPX2 are differentially expressed during

Arabidopsis development
In order to investigate the expression of TPXL genes, we analyzed publicly available RNA sequencing data from different developmental stages of Arabidopsis (25). To profile gene expression patterns of selected TPXL genes, we analyzed the expression across all developmental stages (Figure 3, Additional File 5). These heat maps illustrate distinct gene expression of TPXLs during development. TPXL2, TPXL3, TPXL5 and TPX2 seem to be the most widely expressed TPXLs. In general, expression of TPX2 was among the highest in all tissues. Strikingly, TPXL4 was considered as pseudogene (17); however, transcriptome analyses showed specific expression of TPXL4 in mature anthers. TPXL6 expression was restricted to siliques and TPXL7 was only detected in dry seeds. Taken together, our data confirmed the validity of our hypothesis that TPXL genes might have evolved different functions during plant development. Similar expression patterns for TPX2, TPXL2, TPXL3 and TPXL5 point to possible functional redundancy (18). Moreover, these four proteins are expressed in a similar pattern to Aurora1 and Aurora2 in agreement with the fact that they are physiological activators of the kinases.

TPXL proteins mostly localize on microtubules
The canonical TPX2 was shown to localize on microtubular arrays and the TPX2_Xklp2 domain is involved in microtubule binding (14,17). TPXL proteins of group A ( Figure 1) contain a TPX2_importin domain important to interact with alpha importin (26) and an Aurora binding domain. Group B proteins contain a TPX2_Xklp2 domain with a kinesintargeting signature. To uncover the localization patterns of the selected TPXL proteins, constructs to express fluorescently-labelled translational fusions were infiltrated into N. benthamiana leaves and visualized by confocal microscopy. Consistent with the presence of a TPX2_importin and an Aurora kinase binding domain, group A TPXL proteins showed strong microtubular labelling at the nuclei ( Figure 4A). Moreover, the proteins also faintly labelled cortical microtubules ( Figure 4A). It should be noted that two members group A TPXL2 and TPXL3 labelled microtubular fibers decorating the nuclear envelope like canonical TPX2 ( Figure 4A). Group B TPXL proteins mainly localized with cortical microtubules, TPXL1, TPXL5 and TPXL6 showed very strong labelling resembling cytoskeletal filaments ( Figure 4B). TPXL5 also labelled microtubules close to the nucleus ( Figure 4B). Interestingly, TPXL7 showed a different localization pattern compared to all other TPXL proteins. TPXL7 lacks microtubular localization and mainly localized in the vicinity of the nuclear membrane ( Figure 4B). The canonical TPX2 decorates cytoskeletal cables and bundles of microtubules around nuclei ( Figure 4C). These results indicate that TPXL gene family has probably evolved differential targeting and different functions in the regulation of microtubule cytoskeletal dynamics.

TPXL proteins re-localize Aurora1 kinase by loading it on microtubular arrays
During interphase, AUR1 kinase is localized in in very low amounts in the nucleus, while the microtubular localization of AUR1 is a hallmark of cell division (5,8). To check whether TPXL proteins co-localize with AUR1, DNA constructs of fluorescently-labelled fusion variants of AUR1 and TPXL were co-infiltrated into N. benthamiana leaves. Normally, AUR1 shows diffuse nuclear and weak cytoplasmic labelling in infiltrated N. benthamiana epidermal cells ( Figure 4D). Interestingly, after co-infiltration with TPX2 construct, AUR1-GFP is mostly localized on cortical microtubules ( Figure 4G). Co-expression of TPXL proteins also re-localized AUR1. TPX2, TPXL1, TPXL2, TPXL3, TPXL4, TPXL6 and TPXL8 relocalize AUR to the nucleus as well as to cortical microtubules, while TPXL4 and TPXL5 relocalize it to microtubules ( Figure 4E, 4F, 4G). Interestingly, co-expression of TPXL proteins with AUR1 not only changed the localization of the kinase, but also re-localized some TPXL proteins. The most striking redistribution was observed for TPXL7, which shared a strong nuclear localization with AUR1. These results indicate that co-localization of TPXL proteins with AUR1 is not only dependent on the presence of a functional Aurora binding domain, but other mechanisms must exist to regulate these proteins.

Diversity of TPX in plants
TPX2 is a widely conserved microtubule-associated protein required for mitotic spindle assembly and function (12,14) although recent findings also show its involvement in DNA damage response (27,28). Functions of TPX2 are well characterized in animals including human; however, the knowledge from plant systems is sparse. Moreover, unlike animals which contain a single TPX2 gene, plants contain a family of twenty TPX2-related proteins.
In this work, we performed phylogenetic analyses of the whole group of Arabidopsis TPXL proteins. The canonical TPX2 contains an Aurora binding domain, a TPX2_importin and TPX2_Xklp2 domains. Only the presence of these three domains defines the bona fide homologs of TPX2 (17). In plants, TPXL proteins have been also identified in Eucalyptus grandii (19); however, the authors did not take into account the importance of the In addition to the canonical TPX2, we have described four additional Aurora binding domain-containing proteins. Until now, the canonical TPX2 of Arabidopsiswas considered to consist of two adjacent Aurora kinase binding domains (14). However, crystallographic analyses of N-terminal part of human TPX2 with Aurora A catalytic domain showed the presence of two helical stretches separated with a short linker that are responsible for the interaction (6). Indeed, we were able to identify those two stretches in the Aurora binding domain of plant TPXLs. Although the overall plant sequences of the Aurora binding domain are highly divergent from the consensus sequence, residues important for binding and activation of the Aurora kinase in plant TPXL are highly conserved. Interestingly, the Aurora kinase binding domains of plants TPX2/TPXL, like in the case of animal TPX2 also localize at the N-terminus. Whether this is because of the sterical reasons or functional diversification remains to be determined.
Based on domain analysis of TPXL proteins, we observe a possible specialization of TPXL.

Activation of Aurora kinase by TPX2 family proteins seems to be evolutionary conserved
Aurora kinases are known to phosphorylate various targets. However, in plants only histone H3 (32), MAP65-1 (10), microtubule-associated proteins TPX2 (15), TPXL2 and TPXL3 (18) were confirmed as targets of Aurora1 kinase. Additionally, a number of transcription factors were shown to be phosphorylated by both AUR1 and AUR3 kinases (33). Several of these transcription factors are closely related to the regulation of developmental processes. It is tempting to speculate that phenotypical similarities of both AUR1 and TPX2 mutants (5, 8,16) are partially dependent on the activation of the AUR1 kinase by TPX2 family proteins. The in vitro kinase assay proved the activation potential of the Aurora binding domain of TPXL proteins on recombinant AUR1. TPXL3 was the strongest activator of Aurora1 with even higher activity than the canonical TPX2.
Importantly, activation of Aurora kinase by TPX2 seems to be evolutionary conserved, since the Aurora binding domain of TPX2 from a distantly related Brassicaceae species Eutrema salsugineum is also capable of Aurora1 activation ( Figure 2C, Additional File 5).

TPXL members with Aurora binding domain are strongly expressed
Although all the tested proteins were able to activate AUR1, not much was known about their expression during Arabidopsis development. Group A TPXL proteins showed generally stronger expression compared to group B. It is therefore possible, that plants need more TPXL proteins with Aurora binding domain. Despite this, there is always one of the proteins that seems to have generally higher expression and could probably fulfil housekeeping function. Consistently with the previous findings, TPX2 is expressed in highly dividing tissues such as flowers, flower buds and seedlings. The high expression of TPX2 is in agreement with the importance of AUR1 to phosphorylate various substrates, such as histone H3 during cell division (3) in Arabidopsis and during gametophyte development to phosphorylate CENH3 (Demidov et al., accepted to Frontiers in Plant Science). Moreover, the expression pattern of AUR1 and AUR2 is highly similar to that of TPX2, TPXL2 and TPXL3, suggesting a common regulation of these proteins. TPXL2 and TPXL3 were shown as interactors of AUR1 and AUR2 and TPXL3 is a primary activator of Aurora1(18) further supporting the importance of their common expression patterns.

The localization of TPXL is associated with their importance for spindle microtubules
We also speculate that activation of Aurora kinase by TPX2-related proteins TPXL2, TPXL3, TPXL4 and TPXL8 is related to its nuclear localization which is in agreement with the proposed function of TPX2 in chromatin-induced mitotic spindle assembly (16). On the other hand, we were not sure that TPXL without TPX2_Xklp2 domain can be localized on microtubules. Most likely similarly to canonical TPX2, other TPXL proteins contain a microtubule-binding domain in front of the TPX2_importin domain (13). Infiltration of N.
benthamiana confirmed the functionality of the tested TPXL proteins. Most of the proteins localized both in the nucleus and on microtubules. It has previously been shown that the overexpression of TPX2 in Arabidopsis results in nuclear envelope and in nuclear localization (16). Relatively similar localization was also observed for other members of TPXL gene family.
Co-localization analyses showed especially strong overlap with AUR1 kinase for those TPXL with an Aurora binding domain. In humans, the interaction between TPX2 and Aurora A is not only important for its activation but also for targeting of the kinase to the spindle microtubules (34) and assembly of spindle of the correct length (35). Importantly, most of the TPXL proteins changed the localization of AUR1. This was a big surprise for us, because it is unclear how TPXL isoforms without Aurora binding domain can interact with Aurora kinases. This phenomenon could possibly be explained by the presence of the coiled-coil motif, a domain involved in protein dimerization and protein-protein interactions. The coiled coil motif is present in all TPXL proteins. Importantly, human TPX2 is known to provide a scaffold for the chromosome passenger complex (36). Similarly, TPXL might have a critical role for the recruitment of the microtubule nucleation complex by Aurora kinase.

Conclusion
The Arabidopsis TPXL gene family consists of 21 TPXL genes. Phylogeny of the TPXL gene family from selected animal and plant genomes revealed two major groups of TPXLs, Awith Aurora binding and TPX2_importin and B-with TPX2_Xklp2 domains. The Aurora1 binding domains of TPXL2, TPXL3, TPXL4 and TPXL8 were able to activate AUR1 in vitro.
Analyses of publicly available gene expression data confirmed different expression profiles of TPXL family members. While TPX2, TPXL1, TPXL2, TPXL3 and TPXL5 showed expression throughout the plant development, TPXL4, TPXL6, TPXL7 and TPXL8 were expressed at specific developmental stages. Ectopic expression in N. benthamiana showed localization of all TPXLs except TPXL7 on cortical microtubules. Interestingly, coexpression with AUR1 changed its localization towards microtubules. Our findings suggest that for targeting of Aurora1, one or more TPXL proteins may respond as activation factors. We speculate that the expansion of this protein family arose to assist in the proper spatiotemporal regulation of α-Aurora kinases in plants.

Identification of TPXL proteins
The BLAST (Basic Local Alignment Search Tool, (37)) was used to identify protein homologues of canonical TPX2 protein (At1g03780) in Arabidopsis. To characterize the domain composition of TPXL proteins, in silico analyses of protein sequences using PFAM30 (38) and SMART (Simple Modular Architecture Research Tool, (39)) domain prediction programs was performed.

Phylogenetic analysis
The EggNOG4.5 database (http://eggnogdb.embl.de/; (20)) was used to identify orthologs of TPX2_Xklp2 and TPX2_importin domain-containing proteins. The database EggNOG4.5 contains an orthologous group of genes, which were retrieved from eukaryotic, prokaryotic and viral sequencing projects. Each identified homologue of TPX2 (At1g03780) in Arabidopsiswas submitted to search separately. Some identical sequences occurred in several orthologous groups and those were removed. The final dataset contained 458 protein sequences. Subsequently, multiple sequence alignment was performed in MUSCLE 3. 8.31 (40). The maximum likelihood phylogenetic tree was inferred in RAxML HPC 8.2.9 (41) using PROTGAMMALG model via Cipres Science Gateway (42). The tree topology was tested using rapid bootstrapping by 500 replicates. The tree was rooted to TPX2 protein of Naegleria gruberi.
The maximum likelihood phylogenetic analysis of TPX2-like proteins in A. thaliana was performed in MEGA 7 (GAMMA+LG model; (43)). The tree topology was tested using bootstrapping by 1000 replicates.

Gene differential expression analyses
Med normalized raw counts of A. thaliana gene expression data were downloaded from (25). Obtained raw counts were further within-sample normalized using transcript per

Production of recombinant proteins
GST-Aurora1 was expressed in E.coli C-43 strain (Lucigen, www.lucigen.com) and purified under native conditions as described in (15). Aurora binding domains of TPXL genes were expressed in E.coli BL-21 (GE Healthcare Life Sciences, https://www.gelifesciences.com) and purified under denaturing conditions as described (32).

In vitro kinase assay
Purified recombinant proteins were desalted in kinase buffer using 7K MWCO Zeba Spin Columns (Thermo Scientific) and processed as described in (15). Briefly, samples with Aurora1 were incubated at 30°C, 30 min with 0.1 mM ATP for activation of the kinases.

Infiltration of N. benthamiana and confocal microscopy
Transient infiltration of N. benthamiana leaf cells was performed as described in (47). For the infiltration of multiple constructs, bacterial cultures with an OD between 1 -1.3 were mixed in a 1:1 ratio. Distribution of fluorescence signals within the nucleus was recorded as Z-stacks. For a co-localization analysis, probes were excited with dual 488 nm and 561 nm laser lines in combination with a 488/561 nm beam splitter. eGFP emission was detected over a 490-540 nm range, RFP emission was detected over a 570-620 nm range.
Presence of both fluorophores was confirmed by photospectrometric analysis.

Accession numbers
Sequence data from this article can be found in the EMBL/GenBank data library under following accession numbers Q5XVC4 (AGI locus identifier At3g01015, TPXL1), Q4V3B0  family with Aurora binding domain are able to activate Aurora1. TPX2 of distantly related Eutrema salsugineum (EsTPX2) also activates Arabidopsis Aurora1. ***pvalue < 0.001 in hypergeometric test, **p-value < 0.01 in hypergeometric test.