To identify representative protein sequences for plant kinesin 4s we searched the Phytozome v12 repository (phytozome.jgi.doe.gov). To search, we used the amino acid sequence of the motor domain of kinesin 4IIa from Physcomitrella patens (first 362 amino acids of protein Pp3c13_15600V3.1). Using the BLAST algorithm against the proteomes of 14 selected species we identified 947 protein hits. We selected the following species for analysis: all available Chlorophyte algae (Chlamydomonas reinhardtii v5.5, Dunaliella salina v1.0, Volvox carteri v2.1, Coccomyxa subellipsoidea C-169, v2.0, Micromonas pusilla CCMP1545 v3.0, Micromonas sp. RCC299 v3.0, Ostreococcus lucimarinus v2.0), all available Bryophytes (Marchantia polymorpha v3.1, Physcomitrella patens v3.3, Sphagnum fallax v0.5), a single available Lycopodiophyta species (Selaginella moellendorffii v1.0), three representative well-annotated angiosperms (Ananas comosus v3, Oryza sativa v7, Arabidopsis thaliana TAIR10).

To enrich for members of the kinesin 4 family, we identified the position the kinesin 2 member from P. patens in the BLAST results list and selected all sequences with a smaller E value than 1.5 × 10⁻⁸⁰. This resulted in a total of 87 sequences. From this list, we selected only the proteins derived from the primary transcripts, resulting in 53 sequences. Following a simple analysis with ClustalW alignment and neighbor-joining tree build method in Geneious R7, we removed all kinesin 5s and kinesin 2s with the exception of P. patens kinesin 2, which was used for rooting the final tree (see below). This resulted in 39 sequences. To identify the motor domain regions, we aligned the 39 sequences together with the head domain of human kinesin 1. The homologous regions were selected from all sequences, and the amino acid positions are listed in Supplementary Table 1. We found kinesin 4 sequences in all species analyzed except in the algae Dunaliella salina and Ostreococcus lucimarinus. This may be the result of incomplete annotations or a true absence in their genome; we did not pursue further analysis for these species.

The motor domains of all the selected kinesin 4s were aligned in Geneious R7 using the ClustalW algorithm with default settings. Because the homology is high between kinesin motor domains, it was not necessary to manually adjust the resulted alignment. To generate the final phylogenetic tree, we used the Maximum-Likelihood algorithm (PhyML) available as a plugin for Geneious R7 [44] with default settings, except a number of bootstraps set to 1000 and no optimization. A consensus tree was subsequently generated where any branch support with a bootstrap value under 50% was considered not significant. The tree was rooted using the kinesin 2 sequence from P. patens and the tree was displayed with Geneious R7 indicating the percent bootstrap support for each branch.

All cell lines used in this study were derived from the moss Physcomitrella patens (Hedw.) Bruch & Schimp., Gransden strain. Plants were subcultured by grinding weekly on PpNO3 or PpNH4 pates made from commercial packets (Caisson Laboratories Inc., Utah), 7 g of plant agar and 1 L of dH₂O. For the RNAi lines the culturing process was the same with only a couple differences. Initially, to make a plate where the kinesin would be knockdown for imaging, the RNAi line would be plated on normal PpNO₃ plates for four to seven days, then switched the cellophane holding the moss to a PpNO₃ plate with 1 µM estradiol for three days. This would allow for the RNAi system to be activated preventing new kinesins of the line in question to be made and any kinesin that were left over to degrade. The PpNO₃ with 1 µM estradiol plates were prepared from a 100 mM estradiol stock in water. For our experiments we used a P. patens cell line kindly provided by Dr. Goshima (Division of Biological Science, Graduate School of Science, Nagoya University, Japan) where an inducible construct that silences the kinesin 4II subfamily is present [43] . The line selected for most of the work was RNAi for kinesin 4IIb 359-1. For controls, we used the parental line originally used to make the kinesin 4II-RNAi line as well as the inducible line in the absence of estradiol.

Protonemal growth was evaluated from regenerating protoplasts using previously published conditions [45] [46] . In order to test the effects of knocking down kinesin 4II the following modifications were made: Protoplast were regenerated for 4 days with no estradiol, then moved to plates with 1 µM estradiol for 4 days while controls were not treated with estradiol. Cells were transferred to agar pads and stained with 10 µg/ml calcoflour and imaged using fluorescence microscopy. Area and solidity values per plant were measured using an ImageJ macro as previously described [46] .

Slides containing live moss were prepared for microscope viewing by mounting the cells on a thin layer of 200 µl of PpNO₃ agar containing 25 µM latrunculin B (diluted form a 5 mM stock in ethanol). This agar was placed on a slide with parallel strips of tape on either side of the agar to from a pad and subsequently removed. Small squares of cellophane from the desired plate of moss were cut and placed face down on the solidified agar, left to set for 1 - 2 minutes before the cellophane was removed carefully with tweezers. Liquid medium (20 µl) containing 25 µM latrunculin B, to depolymerize actin filaments, was added to the agar pad before covering with a coverslip. The coverslip was then sealed to the slide using melted vaseline, lanolin, and paraffin combined in equal parts.

Cells were analyzed using a laser scanning confocal system (Leica-SP5) using a 40X oil lens (NA 1.2). Moss cells that were being observed were always sub-apical, and the sub-apical cell also must have had enough room for the chloroplasts to move as necessary. Once a specific cell was chosen it was zoomed to a value of three and the laser rotated until the middle of the cell was horizontal on the screen, the confocal pinhole was set to 1 AU. The format was set to 512 × 185 pixels along with a 30 × 30 µm ROI is set up in the center of the cell for blue light illumination. Only this ROI portion of the cell was exposed to 488 nm blue light laser at 10% for 20 minutes, this initiates an avoidance response in the chloroplasts. The rest of the cell was illuminated with a 591 nm laser at 5%, which illuminates the rest of the cell without causing any chloroplast response. A double dichroic 488/561 nm mirror separates the emission for the excitation laser the emission is then captured by two different detectors. One detector (PMT4) was set at ~700 nm for visualizing the chloroplast. The second PMT (Trans detector) produces a bright-field version of the cell. The speed of the scanning was set to 200 Hz, images were taken two seconds apart. In order to accurately compare the results of each test, these time-lapse images were quantified using ImageJ software and underwent statistical analysis using R software. The 30 × 30 µm illuminated section from the time series was contrast-enhanced with ImageJ (enhance contrast function), setting saturated pixels to 0%. The mean fluorescence intensity was evaluated for all time points. For averaging across cells, the fluorescence values were normalized to the brightest time point in the series. For clearer data display, only 150 data points from the 600 available points were plotted for displayed. Plots were generated with R software.

We have previously reported that in P. patens exists a subfamily of kinesin 4 proteins (kinesin 4IIs) that were not present in A. thaliana [32] . With the recent availability of several more genome sequences, here we have expanded our phylogenetic analysis to explore additional species. For this analysis we selected all algal species with a fully sequenced genome, all fully sequenced bryophytes (two mosses and one liverwort), the club moss Selaginella moellendorffii, and three representative angiosperms with the best-annotated genomes, pine apple, rice and A. thaliana. (For details see Materials and Methods section). After amino acid sequence alignment of their motor domains, we constructed a phylogenetic tree using the Maximum-Likelihood algorithm PhyML [44] . The tree topology is consistent with our previous analysis: in the green lineage, there are two clearly distinct subfamilies of kinesin 4s (Figure 1). The kinesin 4I subfamily has representatives from all the species analyzed, except for two algal species, Dunaliella salina and Ostreococcus lucimarinus, which do not have any kinesin 4 sequences of any type. The absence of kinesin 4 sequences in these two species is puzzling, but it may be the result of incomplete genome annotation; future analysis may help clarify this point. In contrast, kinesin II representatives are only present in some of the green algae species, in the bryophytes, and in S. moellendorffii. Algae and the liverwort M. polymorpha each has only one representative of each subfamily, while both mosses have multiple representatives for each. The club moss, S. moellendorffii, has multiple kinesin 4I molecules, but only one kinesin 4II representative. Kinesin 4IIs from bryophytes and S. moellendorffii form a clear separate group, while the algae kinesin 4IIs form a less tight group. Kinesin 4Is have been considerably expanded in land plants, with bryophytes kinesin 4Is clustering separately from their angiosperm counterparts. Interestingly, the kinesin 4Is from S. moellendorffii are distributed between the two groups, with two molecules clustering with each group, and a fifth molecule showing more divergence.

We evaluated several RNAi lines designed to silence the kinesin 4II subfamily

and identified one line that showed a consistent reduced-growth phenotype and our preliminary analysis suggested also an effect on chloroplast motility. In addition to the reduced growth that resulted when the kinesin 4II-RNAi line was placed on estradiol we observed less branching. We also observed a reduction in caulonemata differentiation and an abundance of shorter chloronemata (Figure 2, compare Figure 2(a) and Figure 2(b)). Another interesting feature of these abnormal growing plants was the fact that tip cells were slightly swollen (Figure 2(b)). This phenotype was reminiscent of cells where the polarization machinery has been disrupted [47] [48] , suggesting a possible participation of kinesin 4IIs in cell polarization.

In order to quantify if the lack of kinesin 4II results in a significant growth phenotype, a growth assay was conducted. For this growth assay we used the RNAi-kinesin 4II line treated with estradiol and not treated, as well as the parental line originally used to generate the RNAi line, which was also treated with estradiol. The parental line acted as a control to confirm that the RNAi lines

were not being affected by the estradiol treatment. Representative images for each of the conditions illustrate the consistency of the observed phenotype (Fig- ure 2). Quantification of the morphological parameters, area and solidity, is shown in Figure 3. Both total area and solidity were significantly affected in the kinesin 4II-RNAi line (Figure 3(a) and Figure 3(b)). The significant reduction in area indicates that the plants are smaller, which is likely a result of a reduction in cellular growth, but cell numbers may also be affected. Solidity is the ratio between the plant’s area and the convex hull’s area, as this value increases the plants have a more “solid” morphology, which can be interpreted as a reduction

of cell polarization and branching. The observed significant increase in solidity in kinesin 4II-RNAi plants is consistent with the smaller plants having a reduction in polarization and branching, further supporting the participation of kinesin 4IIs in some aspect of cellular expansion. It is important to emphasize that we did not detect a difference in the overall area and solidity values between the control (untreated) line and the parental line treated with estradiol (Figure 2 and Figure 3). This indicates that any growth differences observed were a direct result of the induced RNAi lines and not the presence of estradiol.

Because of the observed growth phenotype in the kinesin 4II-RNAi line, which resulted in small cells, it was necessary to establish conditions to visualize chloroplast motility in well-developed cells. This was provided by the versatility of the inducible RNAi system. To allow for the cells to expand before performing the chloroplast motility assay, the kinesin 4II-RNAi plants were first cultured for a full week in regular medium and subsequently transferred to estradiol plates for three days; this resulted in a ten-day culture process before imaging. The control line was just plated on PpNO₃ plates for ten days. Following this procedure, fully elongated cells were available where the kinesin 4II RNAi system was induced.

Fully elongated sub-apical cells were selected that morphologically matched the untreated cells, and time series were collected to evaluate the chloroplast light avoidance response. Because in P. patens the chloroplast avoidance response can be mediated by either the actin or the microtubule cytoskeleton [8] , we depolymerized the actin cytoskeleton to focus on microtubule-dependent processes and avoid confusion while interpreting our data. To depolymerize the actin cytoskeleton, we used the actin monomer-sequestering small molecule, latrunculin B, at a concentration known to disrupt the actin cytoskeleton [45] . During the assay the cells were exposed to blue light at their central region and chloroplast motility followed by imaging chloroplast using chlorophyll fluorescence and brightfield imaging. Images were acquired at 2 sec. intervals, which due to the required confocal laser scanning acquisition time results in an almost continuous light exposure. This light exposure robustly stimulates the light avoidance response, which is evident just a few minutes after exposure (Figure 4(a)). In control cells, chloroplast moved almost completely out of the highly irradiated zone after 20 min. of constant illumination (Figure 4(a)). In contrast, in the induced kinesin 4II-RNAi line, the chloroplasts do not seem to move to the same extent after 20 min. of constant illumination (Figure 4(b)). However, looking at the time series (Supplementary Movies 1 and 2) and by quantifying the data (see below) some movement of the chloroplasts can still be detected.

To compare the magnitude and rate of the chloroplast light avoidance response between the control and induced kinesin 4II-RNAi lines, the fluorescence signal decay in the illuminated zone was measured and plotted as a function of time (Figure 5). As the chloroplast move out of the irradiated zone, the mean value for the chloroplast fluorescence signal is reduced. To correct for bleaching we normalized the analyzed images, using a linear histogram stretching, previous to quantifying the fluorescence. But, as is evident from the bright field images, the chloroplasts clearly move out of the irradiated region in control cells, and the light avoidance response is the main factor generating the reduction in fluorescent signal. Finally, to be able to compare across cells, the mean value of the fluorescence signal from each time point was divided by the mean value of the brightest time point of the series. The dynamics of chloroplast blue light

avoidance of control cells follow an initial constant rate, which is slightly reduced after 10 min., reaching a minimum value after 20 min. (Figure 5). Kinesin 4II-RNAi cells also follow a constant avoidance rate during the initial 10 min., but the rate is slower than that of control cells (Figure 5). After 10 min. the kinesin 4II-RNAi cells seem to reach a plateau with very little light avoidance (Figure 5). Hence, it is clear that the chloroplasts still move in kinesin 4II-RNAi cells, but they do it at a much slower rate in response to blue light stimulation. To compare the statistical significance of the difference observed between control and kinesin 4II-RNAi cells, we compared the two last time points using statistical analysis and confirmed that the observed difference is highly significant (Figure 5).