Characterization of Cellulose Synthesis Complexes in Physcomitrella patens

Due to the enormous economic value and significance of cellulose in human consumption and plant cell walls, production of cellulose microfibrils is considered to be one of the most critical biochemical processes in plant biology. In the past decades, cellulose biosynthesis has been extensively studied in vascular plants. More and more fundamental questions related to this key process are being answered. One such question is: What are the protein components of the enzymatic complex for cellulose synthesis? In seed plants, membrane-embedded rosette Cellulose Synthesis Complexes (CSCs) producing cellulose microfibrils are obligate hetero-oligomeric, being assembled from three functionally distinct and non-interchangeable cellulose synthase (CESA) isoforms. For instance, Arabidopsis has two types of CSCs. One contains AtCESA1, AtCESA3, and AtCESA6, involved in cellulose synthesis in primary cell walls; the other consists of AtCESA4, AtCESA7, and AtCESA8, specialized for secondary cell wall deposition. Recently, the stoichiometry for the three Arabidopsis CESAs forming a CSC was determined to be a 1:1:1 molecular ratio. The constructive neutral evolution hypothesis has been proposed as a mechanism for evolution of these hetero-oligomeric complexes. Physcomitrella patens, a non-vascular plant, is one of the most popular models for genetics studies. A relatively small genome, dominant haploid phase, and high rate of homologous recombination make P. patens a simple and efficient system for genetic manipulation. Seven CESA genes (PpCESA3, PpCESA4, PpCESA5, PpCESA6, PpCESA7, PpCESA8, and PpCESA10) were identified in the P. patens genome, but proteins encoded by these genes are not orthologs of functionally distinct seed plant CESAs according to phylogenetic studies. The similar rosette-type of CSCs were observed in P. patens by freeze-fracture electron microscopy. It is not yet known whether the P. patens CSCs are homo-oligomeric complexes consisting of only a single type of CESA, or hetero-oligomeric complexes assembled by different CESAs like those in seed plants. Knowing this information would be helpful for understanding the roles of different CESAs that compose seed plant CSCs. Furthermore, answers to this question potentially will be useful for testing the constructive neutral evolution hypothesis, since moss CESAs diversified independently from seed plant CESAs. In this study, I generated PpCESA knock out (KO) mutants. Morphological analyses were carried out to identify mutant phenotypes of these KOs together with several previously made KO mutants. Cellulose defects in these mutants were also analyzed using quantitative methods. Reverse transcriptase PCR (RT-qPCR) was performed to examine the expression of all seven PpCESAs in KO lines to identify co-expressed PpCESAs that potentially reside within the same CSCs as the deleted PpCESA. Immunoblot analysis using specific monoclonal antibodies was used as an additional method to detect co-expression based on the accumulation of the protein products of these PpCESA genes. Finally, I carried out Co-immunoprecipitation (Co-IP) assays to identify potential physical interactions between different PpCESA isoforms. The results show that functionally distinct CESA isoforms have evolved in the moss P. patens independently from seed plants, and CSCs synthesizing cellulose microfibrils in secondary cell walls of P. patens gametophore leaves are obligate hetero-oligomeric complexes. Meanwhile, our research also suggests that PpCESA5 alone is able to form homo-oligomeric CSCs, making P. patens an intriguing model in which to study the evolution of cellulose synthase.


LIST OF TABLES
Cellulose is a biopolymer of β(1,4)-linked glucose that forms the microfibrils essential in most plant cell walls. It is extensively used for a variety of commercial and industrial purposes including lumber and textiles. The synthesis of cellulose in plants is catalyzed by enzymatic complexes called cellulose synthesis complexes (CSCs) located in plasma membrane . The membrane-bound CSCs were first observed to have a "rosette" structure and to be associated with the ends of microfibrils in freeze-fracture electron microscopy studies on maize . By searching a cotton fiber EST library for sequences similar to a bacterial cellulose synthase gene, the first putative plant gene encoding a cellulose synthase catalytic subunit (CESA) was identified (Pear et al., 1996). Antibodies against cotton CESAs were later produced to label the rosettes in freeze-fractured bean hypocotyls indicating CESAs are components of the multi-protein complexes inserted into the plasma membrane .

Cellulose synthase catalytic subunits (CESAs)
Among currently identified protein components in CSCs, CESAs are implicated by all sorts of evidence  to be the only functional subunits that produce individual glucan chains. Recently, a heterologously-expressed CESA isoform, PttCESA8, from Populus tremula x tremuloides (hybrid aspen), was reconstituted in liposomes and shown to be functional for cellulose microfibril formation in vitro  for the first time.
The CESA family is contained within the glycosyltransferase-2 (GT-2) superfamily characterized by an eight-transmembrane-helix topology and conserved cytosolic substrate binding and catalytic site . The site for substrate binding and catalysis consists of a D, DxD, D, QxxRW motif and is predicted to be in the loop bounded by transmembrane helix 2 and 3 (Pear et al., 1996). In this motif, the first two conserved aspartic acid residues are predicted to bind the substrate, UDPglucose. This was supported by the results of mutational analysis (Pear et al., 1996).
The functions of these residues have been confirmed by x-ray crystallography of bacterial cellulose synthase (Morgan et al., 2013). The third aspartic acid is thought to be involved in the addition of UDP-glucose to the existing glucan, and the QxxRW region is predicted to be a binding site for the growing glucan chain (Morgan et al., 2013). Compared with bacterial cellulose synthase, plant CESAs are larger. That is because the plant CESA also contains an extended N-terminal Zn-binding RING finger domain, a plant-conserved domain within the N-terminal cytoplasmic loop, and a class-specific domain within the central cytoplasmic loop in addition to the conserved catalytic region (Pear et al., 1996). These domains are specific to plants, hence they are thought to be important for the interactions between the CESA subunits and presumably involved in the formation of the rosette CSCs .

Interactions between CESAs
CESA genes are members of multigene families in plants. For example, Arabidopsis has 10 CESA genes from which distinct combinations are required for primary and secondary cell wall synthesis . The AtCESA4, AtCESA7, and AtCESA8 genes were first shown to be specifically involved in secondary cell wall deposition Taylor et al., 2000;. The mRNAs of the three genes are found to be coregulated in microarray analysis Persson et al., 2005a). Proteins encoded by the three genes physically interact and are exclusively required for assembly of CSCs in cells with thickened secondary walls .
Mutations in AtCESA1, AtCESA3, and AtCESA6 cause primary cell wall defects .
AtCESA3 and AtCESA6 interact with each other according to results of in vitro pulldown assays, and Bimolecular Fluorescence Complementation (BiFC) experiments show that AtCESA1, AtCESA3, and AtCESA6 can interact in vivo . AtCESA2 and AtCESA5 were shown to be closely related and partially functionally redundant with AtCESA6 . In Arabidopsis, therefore, a primary wall CSC might consist of AtCESA1, AtCESA3, and one or perhaps several AtCESA6 like AtCESAs .
Characterization of CSCs has also been carried out in another vascular plant model, PdxtCESA8B; the other one contains PdxtCESA1A and PdxtCESA3 . Altogether, current evidence suggests that vascular plant CSCs are obligately hetero-oligomeric. A theory known as constructive neutral evolution addresses how homo-oligomer complexes are driven towards hetero-oligomeric by neutral processes during evolution. According to this theory, in the initial complex assembled from multiple copies of the same subunit, additional obligate subunits could be evolved by gene duplication followed by relatively high frequency degenerative mutations causing specific interaction sites among them to be lost (Doolittle, 2012;Finnigan et al., 2012). A study showed that the extant Vo complex of the fungi V-ATPase proton pump which is composed of three obligate subunits, evolved from an ancient twosubunit complex by a gene duplication and subsequent complimentary loss of specific interfaces on each daughter isoforms on which they rely to interact with other subunits in the complex (Finnigan et al., 2012). So far, this is the only study that provided convincible experimental evidence. Hence, the generality of this hypothesis needs to be further tested. Plant CSCs are similar to the fungal Vo complex, which are also composed by paralogous CESA isoforms sharing a considerable amount of similarities.
Thus, characterizing the CESAs in plant CSCs will be helpful for continuing testing this theory.
Other components of the CSC Other than CESAs, several other protein components (Endo et al., 2009;Gu et al., 2010) of seed plant CSCs have been identified successively by Co-IP and BiFC. For instance, a putative endo-1,4-β-D-glucanase, KORRIGAN1 (KOR1), was identified to be a part of the primary cell wall CSCs in Arabidopsis (Vain et al., 2014). A microtubule-binding protein, Cellulose synthase interactive protein 1 (CSI1), was discovered to associate with CSCs and serve as a linker protein between CSCs and microtubule . Genetic evidence and the observed size of the cytosolic portion of the rosette demonstrated in electron micrographs (Bowling & Brown, 2008) imply that more other proteins related to cellulose synthesis might also participate in assembly of CSCs.

The moss Physcomitrella patens
Physcomitrella patens, a moss species, has also been shown to have rosette CSCs, but not members of the CESAs clades that contain the functionally distinct isoforms of the hetero-oligomeric CSCs in seed plants . The PpCESA family includes seven members that cluster in two clades (Roberts & Bushoven, 2007 Zimmer et al., 2013). More importantly, P. patens is capable of being genetically manipulated as a result of its high rate of homologous recombination (Reski & Frank, 2005;D. G. Schaefer & Zrÿ d, 1997) . Taking advantage of this unique property, functions of genes of interest can be identified by knockout (KO) mutations (Schaefer, 2002).

Thesis outline
ppcesa5KOs have cellulose defects in primary cell walls affecting gametophore bud development and resulting in a "no leafy gametophore" phenotype   . The mutant phenotype of ppcesa4/10KOs indicates the PpCESA4 and PpCESA10 play some roles in tipgrowing protonema cells, supporting the idea that cellulose is an essential cell wall component in cells undergoing tip growth .
Results of mutational analyses suggest that CSCs involved in cellulose deposition in We propose a mechanism for convergent evolution of secondary walls in which deposition of aggregated and helically oriented microfibrils is coupled to rapid and highly localized cellulose synthesis enabled by regulatory uncoupling from primary wall synthesis.

Introduction
In vascular plants, cellulose is a major component of both primary cell walls that are Cellulose microfibrils are synthesized by cellulose synthase (CESA) proteins that function together as cellulose synthesis complexes (CSCs) in the plasma membrane   ) and three required for synthesis of the lignified secondary cell walls of tracheary elements and fibers . Mutation of any of the secondary CESAs results in a distinctive irregular xylem phenotype characterized by collapsed xylem tracheary elements and weak stems (Taylor et al. 2004 (Hebant 1977). Although the stereid cell walls of P. patens are known to contain cellulose (Berry et al. 2016), the mesoscale structure has not been examined. Only one of the seven P. patens CESAs has been characterized functionally. When PpCESA5 was disrupted, gametophore buds failed to develop into leafy gametophores, instead forming irregular cell clumps.
The associated disruption of cell expansion and cell division are consistent with an underlying defect in primary cell wall deposition . Recently it was shown that PpCESA3 expression is regulated by the NAC transcription factor PpVNS7, along with thickening of stereid cell walls (Xu et al. 2014).
Here we show that PpCESA3 and PpCESA8 function in the deposition of stereid cell walls in the gametophore leaf midribs of P. patens and are sub-functionalized with respect to PpCESA5. We also used polarization microscopy and SFG to reveal similarities in the mesoscale organization of the microfibrils synthesized by PpCESA3 and PpCESA8 and those in the secondary cell walls of vascular plants. Finally, we propose a mechanism through which uncoupling of primary and secondary CESA regulation played a role in independent evolution of secondary cell walls with aggregated, helically arranged cellulose microfibrils in the moss and seed plant lineages.

PpCESA3 and PpCESA8 function in secondary cell wall deposition
Cellulose synthase genes PpCESA3 and PpCESA8 were independently knocked out by homologous recombination in an effort to examine their roles in development and cell wall biosynthesis in P. patens. Stable antibiotic resistant lines generated by transforming wild type P. patens with CESA3KO or CESA8KO vectors were tested for integration of the vector and deletion of the target gene by PCR (Fig. S1).
Integration was verified for five ppcesa8KO lines recovered from two different transformations, line 8KO5B from a transformation of the GD06 wild type line and lines 8KO4C, 8KO5C, 8KO7C and 8KO10C from a transformation of the GD11 wild type line (Fig. S1). Integration was verified for three ppcesa3KO lines recovered from a single transformation of GD11 and three double ppcesa3/8KO lines recovered from a single transformation of the ppcesa8KO5B line with the CESA3KO vector (Fig. S1).
The GD06 and GD11 lines are from independent selfings of the same haploid wild type line, as described in Materials and Methods.
The colonies that developed from wild type and KOs consisted of protonemal filaments and leafy gametophores (Fig. 1). Whereas wild type, ppcesa3KO, and ppcesa8KO gametophores grew vertically, the gametophores on ppcesa3/8KO colonies were unable to support themselves and adopted a horizontal orientation.
Superficially ppcesa3/8KO colonies appeared to produce fewer gametophores ( Fig. 1), but dissection revealed similar numbers of horizontal gametophores that had been overgrown by protonemal filaments. Thus, PpCESA3 and PpCESA8 are not required for gametophore initiation or morphogenesis, but they appear to contribute to structural support.
When examined with polarized light microscopy, the wild type gametophore leaves exhibited strong cell wall birefringence in the midribs and margins (Fig. 1). In contrast, the leaves produced by ppcesa3/8KOs lacked strong birefringence in these cells, consistent with reduced crystalline cellulose content. The ppcesa3KO leaves appeared similar to wild type leaves ( Cellulose Binding Module (CBM) 3a provides a third method for detecting cellulose and can be used to probe thin sections (Blake et al. 2006). In sections from fully expanded wild type leaves, the walls of the lamina cells were labeled relatively weakly with CBM3a, whereas the thickened cell walls of the central midrib and bundle sheath cells were strongly labeled (Fig. 1). The same was true for ppcesa3KO leaves.
However, midrib and bundle sheath cell labeling was nearly absent in ppcesa3/8KO and diminished in ppcesa8KO ( Fig. 1) compared to wild type and ppcesa3KO.
Differential interference contrast microscopy of the same sections showed enhanced contrast in wild type and ppcesa3KO midribs (Fig. 1). Partial cell collapse occurred during embedding in ppcesa3/8KO leaves (Fig. 1).
The cellulose content of the leaf midribs in wild type and single and double ppcesaKO mutants was quantified by measuring the intensity of S4B fluorescence. Statistical analysis confirmed that the S4B fluorescence was significantly reduced in double KOs, but not in ppcesa3KOs (Fig. 2). The intermediate phenotype of the ppcesa8KOs was confirmed and shown to be significantly different from both wild type and the double KOs (Fig. 2). Updegraff analysis showed that cellulose content of cell walls from whole ppcesa3/8KO gametophores (mean±S.E. of three genetic lines = 33.8±0.034%) was reduced significantly (p = 0.004) compared to wild type (GD06, mean±S.E. of three independent cultures = 60.1±0.030%).
To confirm that the observed ppcesa3/8KO phenotype was due to the absence of PpCESA3 and PpCESA8, the selection cassette was removed from ppcesa3/8KO-86 by Cre-mediated recombination of flanking lox-p sites (Vidali et al. 2010) to allow transformation with vectors that drive expression of PpCESA3 or PpCESA8 with their native promoters (Fig. S2). Stable antibiotic resistant lines selected for the presence of numerous erect gametophores were examined with polarization microscopy (Fig. S2).
For the transformation with proCESA8::CESA8, 13 lines were examined, 6 of these had strong midrib birefringence, and the first 3 were used for further analysis. For the transformation with proCESA3::CESA3, the first three lines examined had strong midrib birefringence and were used for further analysis. S4B staining confirmed that expression of PpCESA8 or PpCESA3 rescued the defects in cellulose deposition in the leaf midribs of the double ppcesa3/8KO (Fig. 2). Lines from the transformation with proCESA8::CESA8 were expected to be restored to the wild type phenotype because ppcesa3KO, which also expresses PpCESA8 under control of the PpCESA8 promoter, showed no defects in cellulose deposition in the leaf midrib. All three proCESA8::CESA8 lines had significantly stronger S4B fluorescence than ppcesa8KO. This demonstrates substantial restoration of the phenotype, although fluorescence was still significantly weaker than the wild type (Fig. 2). Two lines from a transformation with proCESA3::CESA3 (3R29 and 3R52) were not significantly different from ppcesa8KO-5B, which is expected since they both lack PpCESA8 and express PpCESA3 under control of the PpCESA3 promoter. In the third line (3R45) fluorescence was restored to wild type levels (Fig. 2). Y-axis scales differ between experiments due to the use of different exposure time settings.

Secondary cell wall microfibrils are helically oriented and laterally aggregated
A first order retardation plate was used with polarized light microscopy to determine the optical sign, and thus the cellulose microfibril orientation, of wild type and ppcesa3/8KO midrib cell walls (Fig. 3). In mature wild type leaves, the larger bundle sheath-like cells that surround the central stereids showed blue addition colors when oriented parallel to the major axis of the plate and yellow subtraction colors when oriented perpendicular to the major axis (Fig. 3), indicating that the net orientation of positively birefringent cellulose microfibrils is longitudinal. In contrast, the walls of the smaller central stereids were colorless when oriented parallel or perpendicular to the major axis (Fig. 3). However, when oriented at 45 o to the retardation plate, these cells showed alternating bands of blue and yellow (Fig. 3), indicating that the microfibrils in their walls are helical with an angle near 45 o . The central midrib cells of developing wild type leaves showed a transition from colorless to blue to yellow along the apical to basal developmental gradient when the midrib was oriented parallel to the major axis of the plate (Fig. 3). This indicates that the microfibril orientation changes from transverse to longitudinal and then to helical as the cells mature. In contrast, the central midrib stereids of mature ppcesa3/8KO leaves had blue addition colors when oriented parallel to the major axis, yellow subtraction colors when oriented perpendicular to the major axis, and no interference color when oriented at 45 o to the retardation plate indicating that microfibrils are longitudinal, rather than helical. Developing ppcesa3/8KO leaves had no longitudinal gradient in interference colors (Fig. 3).
The walls of midrib cells were examined by transmission electron microscopy in ultrathin sections of chemically fixed gametophore leaves. Despite the reduced cellulose content detected by other means, the walls of midrib cells were thickened compared to walls of adjacent lamina cells in all ppcesaKOs, as well as wild type leaves (Fig. 4). When we attempted to prepare specimens by high pressure freezing and freeze-substitution, the leaves fractured in a plane parallel to the midrib. This resulted in a loss of midrib cells and precluded examination of midrib cell walls in these specimens. We were able to examine the lamina and margin cells of freeze-substituted leaves in wild type and two lines of each mutant. The walls of these cells appeared similar between wild type, and single and double ppcesaKOs (Fig. S3).
However, measurements revealed that lamina cell external walls, i.e. those facing the external environment, were thinner in ppcesaKOs (Fig. S4).
The mesoscale organization of cellulose in the midribs of wild type, ppcesa3/8KO, and ppcesa8KO leaves was examined using a broadband SFG microscope (Lee et al. 2016  for cellulosic samples is 20-25 μm (Lee et al. 2016). Given that the thickness of turgid leaves is about 50-60 μm at the midrib and that they likely collapse to less than half their thickness when dried, we conclude that most of the leaf thickness contributes to the SFG signal. In spectra collected from the wild type, a strong peak at 2944 cm -1 , which is characteristic of secondary cell walls, was observed in the CH/CH 2 stretch region along with a 3320 cm -1 peak in the OH stretch region. In contrast, the spectra collected from ppcesa3/8KO and ppcesa8KO midribs had weaker peak intensity overall with a broad CH/CH 2 stretch peak centered around 2910 cm -1 . Compared to ppcesa3/8KO, the spectra from ppcesa8KO midribs had a weak signal at 2963 cm -1 that was absent in spectra collected from ppcesa3/8KO midribs. A scan across a wild type leaf shows that the 2944 cm -1 signal is associated with the midrib and was not observed in the cells of the lamina (Fig. 5). Equivalent scans of ppcesa3/8KO and ppcesa8KO leaves confirm the absence of a strong 2944 cm -1 peak from the midribs of these mutants (Fig. 5).

PpCESA proteins are functionally specialized
Based on the ppcesa3KO, ppcesa8KO, and ppcesa3/8KO phenotypes, PpCESA3 and PpCESA8 appear to be partially redundant. To determine whether the relative strengths of these phenotypes are related to gene expression levels, we used reverse transcription quantitative PCR to measure the expression of PpCESA3 and PpCESA8 in the wild type and mutants. In the ppcesa3KOs, PpCESA8 was significantly upregulated compared to wild type ( Fig. 6), providing a possible explanation for the lack of a mutant phenotype in these lines. In contrast, PpCESA3 was not significantly upregulated in the ppcesa8KOs compared to wild type, potentially explaining the intermediate phenotype in these mutants.
ppcesa3KOs, ppcesa8KOs and ppcesa3/8KOs were tested for changes in rhizoid and caulonema development to determine whether developmental defects were restricted to the gametophores. When cultured on medium containing auxin, all lines produced the expected leafless gametophores with numerous rhizoids ( and wild type (Table 1).
To determine whether other PpCESAs are functionally interchangeable with PpCESA3 and PpCESA8, we tested for rescue of ppcesa3/8KO-86lox by various PpCESAs driven by the PpCESA8 promoter. Polarization microscopy screening of at least 21 and up to 27 stably transformed lines for each vector revealed little or no midrib birefringence for the proCESA8::CESA4, proCESA8::CESA7 and proCESA8::CESA10 lines and moderate to strong midrib birefringence for 92% and 78% of the proCESA8::CESA3 and proCESA8::CESA5 lines, respectively.
Finally, we examined ppcesa4/10KOs and ppcesa6/7KOs produced for another study to determine whether they phenocopy the ppcesa3/8KO phenotype. Genotype verification for these lines is presented in Fig. S8 and Fig. S9. The ppcesa4/10KOs showed slight, but significant reduction in midrib S4B fluorescence. However, for ppcesa6/7KOs the reduction was substantial and significant ( Fig. 7), showing the PpCESA6/7 and PpCESA3/8 have non-redundant roles in secondary cell wall deposition in leaf midrib cells.

Discussion
PpCESA3 and PpCESA8 function redundantly in cellulose deposition in stereid secondary cell walls.
Targeted knockout of PpCESA3 and PpCESA8 blocked deposition of cellulose in the thick walls of stereid cells as indicated by 1) reduction of the strong birefringence associated with the midribs in ppcesa3/8KOs, 2) reduction in the midrib fluorescence of ppcesa3/8KO leaves stained with S4B, 3) lack of CBM3a labeling of sections from ppcesa3/8KO leaf midribs ( Fig. 1), and 4) reduction in ppcesa3/8KO gametophore cell wall cellulose content as measured by Updegraff assay. Evidence that knockout of PpCESA8 proteins are functionally interchangeable and that a dosage effect is responsible for the ppcesa8KO phenotype. The formation of morphologically normal gametophores in ppcesa3/8KOs ( Fig. 1) indicates that PpCESA3 and PpCESA8 serve a different role in development than PpCESA5, which supports normal cell division and cell expansion required for gametophore development . It is possible that PpCESA3 and PpCESA8 contribute to primary cell wall deposition since ppcesa3/8KO lamina cells had thinner external walls (Fig. S4) and tended to collapse during embedding (Fig. 1). Alternatively, PpCESA3 and PpCESA8 may contribute to secondary thickening of lamina cell walls after they stop expanding.

CESA evolution in both P. patens and Arabidopsis involve sub-functionalization and neo-functionalization.
There are many parallels in the evolution of the P. patens and Arabidopsis CESA families. In both species, different CESAs are responsible for primary and secondary cell wall deposition. In Arabidopsis, the secondary CESAs are AtCESA4, -7 and -8 ) and primary CESAs are AtCESA1,-3, and members of the 6-like group ). In P. patens, midrib secondary cell wall synthesis involves PpCESA3, -6, -7 and -8, whereas gametophore primary cell wall synthesis requires PpCESA5 . At least some primary CESAs can substitute for secondary CESAs and vice versa in both species. In Arabidopsis, AtCESA3pro::AtCESA7 partially rescues atcesa3, and AtCESA8pro::AtCESA1 partially rescues atcesa8 ). In P. patens, PpCESA8pro::PpCESA5 rescues ppcesa3/8KO. This indicates that the CESA division of labor for primary and secondary cell wall deposition in vascular plants and mosses is due at least in part to sub-functionalization. However, neo-functionalization has also occurred in both species, resulting in the requirement for two or more non-interchangeable CESA isoforms for secondary cell wall biosynthesis. In Arabidopsis, atcesa4, atcesa7, and atcesa8 null mutants share a phenotype    patens. A sharp SFG CH/CH 2 stretch peak at 2944 cm -1 is characteristic of angiosperm secondary cell walls (Park et al. 2013) and extensive empirical testing has shown that this spectral feature is attributable to lateral microfibril aggregation (Lee et al. 2014).
The 2944 cm -1 peak was also present in SFG spectra of wild type P. patens midribs. In contrast, the spectra of ppcesa3/8KO leaf midribs lacked the 2944 cm -1 peak and instead had a broad peak between 2800 and 3000 cm -1 , which is characteristic of primary cell walls and other samples lacking aggregated microfibrils (Lee et al. 2014;Park et al. 2013). This suggests that lateral aggregation of microfibrils is a common feature of the secondary cell walls of moss stereids and vascular plant tracheary elements and fibers. Polarization microscopy with a first order retardation plate revealed that the microfibrils in the stereid cell walls are deposited in a helical pattern, as observed in secondary cell walls of tracheary elements and fibers (Barnett and Bonham 2004). Although deficient in cellulose, the stereid cell walls of ppcesa3/8KOs were thickened, indicating that secondary cell wall synthesis involves deposition of non-cellulosic components, which proceeded in the absence of cellulose deposition.
This has also been observed in developing tracheary elements treated with cellulose synthesis inhibitors (Taylor et al. 1992). Thus, stereid cell walls share structural characteristics with the cell walls of tracheary elements and fibers.

Mosses and vascular plants have acquired similar secondary cell walls through convergent evolution.
Thick, cellulose-rich secondary cell walls provide added support for aerial organs of mosses and vascular plants alike. Within these cell walls, the lateral aggregation and helical orientation of the microfibrils contributes to their strength and resiliency.
Although cortical microtubules play an important role in cellulose microfibril orientation, oriented cellulose deposition can occur in the absence of cortical microtubules, and it has previously been suggested that aggregation and helical orientation of microfibrils in secondary walls is a consequence of high CSC density during rapid cellulose deposition (Emons and Mulder 2000;Lindeboom et al. 2008).
Regulation at the level of CSC secretion was emphasized in this model (Emons and Mulder 2000), but CSC density can potentially be regulated at the level of transcription.
Rapid cellulose synthesis during secondary cell wall deposition in specific cell types requires precise temporal and spatial regulation of CESA expression that is distinct from the regulatory requirements for primary cell wall synthesis. We suggest that these distinct regulatory needs were met through the evolution of independent regulatory control of primary and secondary CESAs by sub-functionalization in both mosses and seed plants. In seed plants, phylogenetic analysis shows that the first divergence of the CESA family separated the genes that encode the primary and secondary CESAs and was followed by independent diversification within each group . This, along with evidence that some primary CESAs are interchangeable with secondary CESAs , indicates that subfunctionalization was an early event in the evolution of the seed plant CESA family. In P. patens, the genes that encode secondary PpCESA3 and PpCESA8 and primary PpCESA5 are also sub-functionalized and therefore specialized, although they encode interchangeable proteins.
Several lines of evidence indicate that the capacity to deposit a secondary cell wall

Vector construction
All primer pairs are shown in which can be disrupted with no effect on phenotype (Schaefer and Zryd 1997). Rescue vectors were cut with SwaI for transformation into a P. patens ppcesa3/8KO line from which the hph resistance cassette had been removed (see below).

Culture and transformation of P. patens
Wild type P. patens lines (haploid) derived from the sequenced Gransden strain ) by selfing and propagation from a single spore in 2006 (GD06) or 2011 (GD11) were gifts of Pierre-Francois Perroud, Washington University. Wild type and transformed P. patens lines were cultured on basal medium supplemented with ammonium tartrate (BCDAT) as described previously ).
Protoplasts were prepared and transformed as described previously

Sum Frequency Generation spectroscopy
Leaves of wild type GD06, 8KO-5B, and 3/8KO-86 lines were mounted abaxial side down in water on glass slides and allowed to air-dry overnight. SFG spectra were collected 5 µm intervals along a 200 µm line scan perpendicular to the midrib at its thickest point using an SFG microscope system described previously (Lee et al. 2016).
The SFG spectra were collected with the following polarization combination: SFG signal = s-, 800 nm = s-, and broadband mid-IR = p-polarized with the laser incidence plane and the laser incidence plane aligned along the axis of midrib.
Reverse transcription quantitative PCR RNA was extracted from gametophores from two independent wild type and three independent lines each of ppcesa3KO and ppcesa8KO as described previously ). cDNA samples were tested in duplicate as described previously using primer pairs for amplification of PpCESA3 and PpCESA8. The primers have been previously tested for specificity and efficiency . Primers for actin and v-Type H + translocating pyrophosphatase reference genes were described previously (Le Bail et al. 2013). Target/average reference cross point ratios were calculated for each sample and standard errors were calculated for independent genetic lines.

Statistical analysis
For statistical analysis, one-way Analysis of Variance (ANOVA) with post-hoc Tukey Honest Significant Difference (HSD) test was performed at astatsa.com/OneWay_Anova_with_TukeyHSD/.

Supplemental Materials
Table S1. Primers used for vector construction and genotype analysis.

Acknowledgements
This work was supported primarily by National Science Foundation Award IOS-1257047. Analysis of mutants by SFG spectroscopy was supported as part of The     Full SFG spectra collected from leaf midribs (each is the average of nine spectra, from three different positions on each of three different leaves). A strong peak in the C-H stretch region (2944 cm -1 ) is present in spectra from wild type (WT), greatly diminished in spectra from ppcesa8KO (8KO), and absent in spectra from ppcesa3/8KO (3/8KO). (B) P. patens wild type, ppcesa8KO, and ppcesa3/8KO leaves with SFG scan trajectories traversing the midribs.
Step size was 5 μm/step. SFG spectra were collected from 2850 to 3150 cm -1 , covering the entire CH region. (C) 2D projection image of SFG spectra collected across the midribs of each leaf shown in B. Each column in each image is an entire spectrum collected from one point plotted against displacement along the scan trajectory. Colors indicate SFG intensity as shown in the legend. For each rescue genotype, three independent genetic lines were sampled in triplicate and measured with 6 samples of wild type (GD06) and 8 samples of ppcesa3/8KO-86lox. (A) For lines derived from transformation of ppcesa3/8KO-86lox with proCESA8::CESA3 (8pro:3R), proCESA8::CESA7 (pro8:7R), and proCESA8::CESA10 (pro8:10R) genotypes, the three independent lines did not differ significantly and were combined. proCESA8::CESA7 and proCESA8::CESA10 lines did not differ significantly from the parent double KO line (p > 0.05), whereas proCESA8::CESA3 lines had significantly higher fluorescence compared to the parent double KO line, but significantly less than WT (p < 0.05). Bars indicate the standard error of the mean for three independent lines. Genotypes with different letters are significantly different. (B) For lines derived from transformation of ppcesa3/8KO-86lox with proCESA8::CESA5 (pro8:5R) and proCESA8::CESA4 (pro8:4R), the three independent lines were significantly different and were analyzed separately. proCESA8::CESA5 (5R) lines were not significantly different from the wild type (p > 0.05), except for 5R7, which was not significantly different from ppcesa3/8KO-86lox (p > 0.05). proCESA8::CESA5 lines did not differ significantly from ppcesa3/8KO-86lox (p > 0.05). Bars indicate the standard error of the mean for three gametophores from the same line (n=3). Lines with different letters are significantly different (p < 0.05. (C) Mid rib fluorescence was slightly, but significantly reduced in cesa4/10KO compared to wild type (p = 0.037). Reduction in midrib fluorescence in cesa6/7KO was substantial and highly significant (p = 0.0011). Bars indicate the standard error of the mean for three independent mutant lines or 3 replicates of wild type (n=3).

References cited:
Albalasmeh AA, Berhe AA, Ghezzehei TA (2013)   integration tested by PCR with primer pair VectorF-hph/8KOFlankR produced the expected 637 bp fragment in the same 5 lines. Target deletion was verified in the 3 KO lines by the absence of a product from primer pair CESA8TargetF/CESA8TargetR, which anneal within the PpCESA8 coding sequence and amplify a 339 bp fragment in the wild type. (B) Genotyping strategy and results for ppcesa3 and ppcesa3/8 KO lines. 5' integration tested by PCR with primer pair 3KOFlankF/VectorR-npt produced the expected 1362 bp fragment in lines 3KO5, 3KO35, 3KO126, 3/8KO43, 3/8KO57, and 3/8KO86. 3' integration tested by PCR with primer pair VectorF-npt/3KOFlankR produced the expected 1259 bp fragment in the same 6 lines. Target deletion was verified in the 6 KO lines by the absence of a product from primer pair CESA3TargetF5/CESA3TargetR5, which anneal within the PpCESA3 coding sequence and amplify a 1266 bp fragment in the wild type.     Agreement EPS-1004057. We also thank Bowen Jiang for assistance with statistics.

Introduction
Cellulose is a key component in plant cell walls. In the primary cell wall (deposited during cell expansion), the oriented deposition of cellulose microfibrils serves the vital load-bearing role important in determining the orientation of cell expansion and thus overall plant morphology (Taylor, 2008). After cell expansion has stopped, certain cells, such as collenchyma cells, sclerenchyma cells, and xylem cells, can deposit thickened secondary cell walls (inside the primary wall) that mechanically support plants to stand upright and efficiently conduct water and minerals (Mauseth, 2012).
Cellulose is highly abundant in the secondary walls (Taylor, 2008). Cellulose microfibrils, in higher plants, are synthesized by rosette cellulose synthesis complexes (CSCs) embedded in the plasma membrane. The catalytic core of these complexes is assembled from cellulose synthase (CESA) subunits . In seed plants, the CSCs for cellulose deposition in both primary and secondary cell wall requires three types of functional distinct CESAs for function . In Arabidopsis, mutants for CESA1, CESA3, and CESA6 have cellulose defects in primary cells wall causing developmental retardation and phenotypic changes in hypocotyls and roots Williamson et al., 2001;. Mutations in any of the three secondary cell wall CESAs (CESA4, 7, and 8) result in severe defects in secondary cell wall cellulose deposition leading to collapsed xylem cells in Arabidopsis Taylor et al., 2000;. The moss Physcomitrella patens is an intriguing model bryophyte that is commonly used in genetics studies and mutational analysis because of its ability to be genetically manipulated due to the naturally occurring high rate of homologous recombination. Gene knockin and knockout transformations can be accomplished within one month and phenotyped in a few weeks in P. patens (Kamisugi, Cuming, & Cove, 2005). This is rapid compared to transformation and phenotypic analysis in Arabidopsis, which takes about three months (Clough & Bent, 1998). Rapidly elongating protonema cells in P. patens can be used as an alternative model to examine tip-growth related mechanisms (Rounds & Bezanilla, 2013). Leafy gametophores consist of several distinguishable cell types including support cells (stereids) and water-conducting cells (hydroids), but they develop from single-celled shoot apical meristems, making P. patens a less complicated model to study plant organ morphogenesis (Harrison et al., 2009). Physcomitrella patens has seven CESA genes which can be divided into two sub-clades (A-clade: PpCESA3, 5, and 8; Bclade: PpCESA4, 6, 7, and 10), but are not orthologs of seed plants CESAs (Goss et al., 2012;Roberts & Bushoven, 2007).
We carried out morphological analysis of CESA knockout (KO) mutants in order to investigate functions of CESAs in P. patens. So far, PpCESA5 is known to be required in gametophore development based on the "no leafy gametophore" phenotype of ppcesa5KO mutant . Both of double ppcesa3/8KO and ppcesa6/7KO mutants show significantly reduced cellulose deposition in secondary cell walls in midribs of gametophore leaves, indicating PpCESA3, 8, 6, and 7 are involved in secondary cell wall thickening of stereids . Here, we show that the quadruple ppcesa4/6/7/10KOs are able to produce morphologically normal leafy gametophores, indicating that the B-clade PpCESAs are not required in gametophore morphogenesis. Since ppcesa3/8KO also produces morphologically normal gametophores , together the current results suggest that PpCESA5 might be able to form homo-oligomeric CSCs, solely functioning in gametophore development. In addition, knocking out PpCESA4 and PpCESA10 causes morphological changes in protonemal colonies, suggesting the importance of cellulose in the tip-growing P. patens protonema cells.

Genotyping and morphological analysis (rhizoid, caulonema, and gametophore) of ppcesa4/6/7/10KO
Three verified quadruple ppcesa4/6/7/10KO lines were recovered from three different transformations of ppcesa4/10KO-4B with the CESA6/7KO vector  and tested for 5' and 3' integration of the vector and deletion of the target gene ( Figure 1). All of the quadruple KO lines were able to produce leafy gametophores that were morphologically similar to wild type (Figure 2 A-H) indicating that the Bclade PpCESAs are not required for gametophore morphogenesis. The quadruple KOs were also tested for developmental defects in rhizoid and caulonema development. All of the three quadruple KOs produced leafless gametophores with several rhizoids similar to wild-type after growing on medium supplied with auxin for two weeks ( Figure S1), indicating that PpCESA4, PpCESA6, PpCESA7, and PpCESA10 are not required for normal rhizoid development. When explants of quadruple KOs were cultured vertically in the dark, caulonemal filaments produced by the resulting colonies grew upright against gravity and were similar in appearance to wild-type controls (Figure S2 A-F). Caulonemal length was also not significantly different between the mutant lines and wild-type lines (Figure S2 G).

Cellulose deposition of the secondary cell wall in ppcesa4/6/7/10KO
By polarization microscopy and S4B staining,  showed a large and significant reduction in cellulose deposition in the midribs of ppcesa6/7KO gametophore leaves, whereas the gametophore leaves of ppcesa4/10KOs showed a small, but significant reduction compared to wild-type. To clarify the roles of the clade B PpCESAs in secondary cell wall deposition, we used polarization microscopy to examine midrib birefringence in ppcesa4/6/7/10KO compared to wild-type (Gd11).
We found that gametophore leaves of three ppcesa4/6/7/10KO lines all had substantially reduced midrib birefringence (Figure 2 J-L), similar to the phenotypes of previously described ppcesa6/7KOs and ppcesa3/8KOs, and more dramatic than ppcesa4/10KOs . To quantify the defect in secondary cell wall deposition relative to ppcesa4/10KOs and ppcesa6/7KOs, we stained mutant gametophore leaves with cellulose-specific fluorescent dye S4B and used fluorescence microscopy to measure the cellulose content in midribs of the mutant leaves. All mutants showed a significant reduction in brightness compared with the midribs of wild-type gametophore leaves (Figure 3), consistent with previous results. The quadruple KO midribs had significantly reduced brightness compared to ppcesa4/10KOs but were not significantly different from ppcesa6/7KOs (Figure 3).
The phenotype similarity of ppcesa4/6/7/10KO compared to ppcesa6/7KO, but not ppcesa4/10KO  indicates a major role for PpCESA6 and PpCESA7 and a minor role for PpCESA4 and PpCESA10 in secondary cell wall deposition in gametophore leaf midribs.

Morphological analysis of protonema colonies
Protonemal filaments of P. patens extend by apical cell division and tip growth, branching to form colonies (Cove, 2005). To test whether clade B PpCESAs are required for protonemal tip growth, Chlorophyll autofluorescence images of colonies were analyzed for area, solidity, and circularity. Figure 4 summarizes the results of this assay for ppcesa4/6/7/10KOs, ppcesa6/7KOs, and ppcesa4/10KOs. Circularity is the ratio of colony area to colony perimeter and indicates the degree of polarized extension. A score of 1 represents a perfect circle, while scores approaching 0 represent a more linear shape. Solidity quantifies the presence of concavities in the colony and reflects the degree of polarization and branching of the protonema filaments. The lowest solidity with the highest branching of the filaments was scored 0 and the highest solidity possible with less branching of filaments was scored 1 (Vidali et al., 2007). Graphs in Figure 5 show that when compared with wild-type control, colonies of ppcesa4/6/7/10KO showed increased solidity and circularity. pcesa6/7KOs showed no difference in area, solidity or circularity compared to wild-type (P>0.05, ANOVA). All ppcesa4/10KOs showed significantly increased solidity and circularity compared to wild-type similar to ppcesa4/6/7/10KO, consistent with defects in protonemal tip growth. We further analyzed single ppcesa4KOs and ppcesa10KOs.
Only two of the three ppcesa4KO lines, ppcesa4KO-13A and ppcesa4KO-14B, had significantly increased solidity and circularity compared to wild-type. There was no significant difference observed among the three ppcesa10KOs. We also tested ppcesa5KO and ppcesa3/8KO to test the roles of the clade A PpCESAs in protonema tip growth. One of the three ppcesa5KOs, ppcesa5KO-20, showed the significantly increased solidity and circularity compared to both wild-type and the other two ppcesa5KOs. None of ppcesa3/8KOs showed any significant difference compared with wild-type in the three parameters.

Discussion
Mutation analysis for the B-clade PpCESAs revealed that they are not required for gametophore morphogenesis. This is evident from the fact that quadruple ppcesa4/6/7/10KOs are still able to produce normal leafy gametophores (Figure 2), unlike ppcesa5KO. Gametophore buds of ppcesa5KOs are defective in cell expansion, cytokinesis, and leaf initiation, resulting in failure of leafy shoot formation (Goss et al., 2012). None of these phenomena were observed in ppcesa4/6/7/10KOs. The ppcesa3/8KOs also produce morphologically normal gametophores . Thus ppcesa5KOs are the only mutants that are defective in gametophore morphogenesis. It has also been shown that constitutively expressing PpCESA3 and PpCESA8 can rescue ppcesa5KO indicating A-clade PpCESAs are functionally interchangeable . Thus, the unique mutant phenotype of ppcesa5KOs might be attributable to PpCESA5 having non-overlapping expression with PpCESA3 and PpCESA8 and the non-interchangeable functions with the B-clade PpCESAs (Scavuzzo-Duggan et al., unpublished). According to this, PpCESA5 might be able to form homo-oligomeric CSCs in order to properly deposit cellulose microfibrils into the cell walls of newly emerged gametophore buds. The interchangeable functions of different CESA members are seen only in limited cases in seed plants. Promoter-swap assays in Arabidopsis showed that the defective phenotype of atcesa3 mutants can be partially rescued by driving expression of AtCESA7 using the AtCESA3 promoter and atcesa8 mutants can be partially rescued by driving expression of AtCESA1 using the AtCESA8 promoter .
Results of S4B staining (Figure 3) showed that: 1) there is no significant difference between ppcesa4/6/7/10KOs and ppcesa6/7KO in cellulose content in the midrib secondary cell walls of the mutant leaves; 2) there is a slight but significant reduction in ppcesa4/10KO compared to the wild-type. These results suggest that compared with PpCESA6 and PpCESA7, PpCESA4 and PpCESA10 only have a minor role in secondary cell wall deposition. This is consistent with previous gene expression data showing that PpCESA4 and PpCESA10 have lower expression in gametophores than in protonema . The fact that ppcesa3/8KOs and ppcesa6/7KOs are similar in phenotype showing cellulose defects in secondary cell walls provides a clue that CSCs involved in cellulose deposition in P. patens secondary cell walls might be hetero-oligomeric consisting of PpCESA3, PpCESA8, PpCESA6, and PpCESA7 .
Tip growth in certain types of cells, such as root hairs and pollen tubes, is regulated by highly coordinated mechanisms which guide deposition of new cell wall materials strictly proceeding in a limited area of the cell surface (Carol & Dolan, 2002;Cheung & Wu, 2008;Lee & Yang, 2008;Nielsen, 2009;Gu & Nielsen, 2013). Several studies pointed out that cellulose is an essential cell wall component in cells undergoing tip growth Galway et al., 2011;. Mutational analyses in Arabidopsis showed that some atcesa mutants are severely defective in germinating pollen and elongating pollen tube, indicating important roles of cellulose in the tip-growing cells . Elongating P. patens protonemal filaments are another ideal model to investigate the role of cell wall deposition in tipgrowth related mechanisms . Crystalline cellulose has been detected by affinity cytochemistry with Cellulose Binding Module 3A (CBM3A) in primary cell walls of subapical cells and the very tip region of the apical cells in expanding P. patens protonema filaments (Berry et al., 2016), indicating the potential roles of cellulose during tip growth of protonema. Here, our study shows that P.
patens CESAs (PpCESA4 and 10) have roles in tip growing protonema, supporting the point of view that cellulose is significant for cell tip growth. This is evident from the abnormal protonema colony morphology of ppcesa4/10KOs, which show significantly increased circularity and solidity ( Figure 5). Increased circularity and solidity are caused by slower elongation and less branches of the protonema filaments (Vidali et al., 2007). Quantitative affinity cytochemistry of cellulose content using S4B or CBM3A will be needed to prove that the mutant phenotypes were caused by the decreased cellulose in cell walls of tip-growing protonema cells. Based on available evidence, PpCESA6, PpCESA7, and the A-clade PpCESAs do not seem to contribute to protonemal tip growth, since no obvious phenotypic changes were observed in corresponding KO mutants. Although colony circularity and solidity of ppcesa5KO-20 was shown to be significantly increased in our analysis, this is likely due to other genetic effects since the other ppcesa5KO lines were not different from wild-type. It remains possible that PpCESA5, PpCESA3 and PpCESA8 function redundantly in tip growth. This can be tested by producing and analyzing a ppcesa3/5/8 triple KO mutant. Tip growth in our ppcesa4/10KO and ppcesa4/6/7/10KO was not abolished, suggesting the deposition of cellulose in cell walls of tip-growing protonema involves proteins other than the PpCESAs. Several members from one of the Cellulose Synthase-like (CSL) gene family, CSLD, were shown to be required for tip growth of root hairs and pollen tubes in Arabidopsis (Favery et al., 2001;Wang et al., 2001;Bernal et al., 2008;. P. patens also has the CSLD gene family, and expression of these genes have enhanced expression in cultures containing only protonema (Roberts & Bushoven, 2007). Thus, it will be interesting to carry out mutational analysis to investigate functions of CSLD genes in P. patens protonema.

Transformation and genotyping
Except ppcesa4/6/7/10KOs, the ppcesaKO lines used in this study were created previously and described in .
To create the quadruple ppcesa4/6/7/10KO lines, the hygromycin sensitive ppcesa4/10KO-4B line  was transformed with the CESA6/7KO vector conferring hygromycin resistance and stably transformed colonies were genotyped as described for primary ppcesa6/7KO lines in . Primers used for genotyping are listed in supplemental table 1.

Polarization microscopy of cell wall birefringence
Cell wall birefringence of leaf midribs was analyzed as described in . Three independent lines of each knockout mutant and three biological replicates of wild-type were cultured for 15 days on BCDAT medium. The first fully expanded leaf of each gametophore was cut off with a pair of micro-dissecting scissors (Electron Microscopy Sciences, Hatfield, PA, USA) and mounted in the water on a glass slide. An Olympus BHS compound microscope equipped with a polarizer and circular-polarizing analyzer (Olympus Corp., Shinjuku, Tokyo, Japan) was used to visualize the gametophore leaves. Images were captured with a Leica M165FC digital camera (Leica Microsystems Inc., Buffalo Grove, IL, USA) using identical settings for the knockouts and the wild-type control.
Pontamine fast scarlet 4B (S4B) fluorescence histochemistry S4B staining of leaf midribs was performed as describe . Three independent lines of each knockout along with three biological replicates of wild-type were cultivated on BCDAT medium for 15 days. For each genotype, three gametophores with 10-12 leaves were collected, permeabilized in acetone for 5 seconds, rinsed in PBS, and stained for 30 min in PBS containing 0.01% S4B. All leaves were rinsed in PBS after staining, cut off with a sharp razor blade, and mounted in PBS on a glass slide. Images were taken using the same microscope and conditions described previously in . For data analysis, the midrib of each leaf was outlined by hand and intensity was quantified using ImageJ as described previously .

Analysis of caulonema and rhizoid development
Caulonema and rhizoid assays were carried out as previously described  to test ppcesa4/10KO, ppcesa6/7KO, ppcesa4/6/7/10 KO lines for phenotypic changes. For the caulonema and rhizoid assays, samples were analyzed using a Leica M165FC stereomicroscope, and images were recorded using a Leica DFC310FX camera (Leica). The length of caulonema was measured as described in . Three independent experiments (n=3) were done. For each experiment, caulonema colonies were cultured on seven replicate plates containing solid BCDAT? medium. Four explants were placed along the equator of each plate, with each explant representing a unique genotype.
Protoplasts were isolated from three independent lines for each genotype along with three biological replicates of wild-type using the method described previously . However, it was necessary to add 21 units/mL of cellulase from Trichoderma reesei (Worthington Biochemical Corporation, Lakewood, NJ, USA) to the digestion mixture when using driselase lot # SLBP0654V (Sigma-Aldrich, St. Louis, MO, USA) for effective digestion. Five thousand protoplasts suspended in 1 mL of PRML were spread on each of three plates containing PRMB medium overlain with cellophane. The plates were incubated at 25 o C with constant illumination at 50-80 μmol/m 2 /s for 4 d and cellophane membranes were then transferred to BCDAT plates for an additional 2 d. Colony morphology was documented by capturing chlorophyll autofluorescence images of approximately 50 regenerated protoplasts per plate at 63X magnification using an M165FC stereo microscope with 10447407 GFP filter and DFC310FX camera (Leica). Images were analyzed for area, solidity, and circularity with ImageJ (National Institutes of Health, USA) using a macro developed by Vidali et al. (2007).

Statistical analysis
One-way Analysis of Variance (ANOVA) followed by post-hoc Tukey Honest Significant Difference (HSD) test was performed using "R" programming (Vienna, Austria; http://www.R-project.org/) to identify the potential significant difference in caulonema assay and tip growth assay.

Supplemental Materials
Table. S1. Primers designed for knockout vector construction and genotyping. Table. S2. Data of morphological analysis of protonema colonies.  We also thank Bowen Jiang for assistance with statistics. Figure 1: PCR-based genotyping of ppcesa4/6/7/10KO lines. Genomic DNA from wild-type P. patens (WT) was used as positive control. The expected band for the target gene of 1178 bp (PpCESA6) and 141 bp (PpCESA7) was observed in WT, but not in KO lines. The expected 5' integration band of 833 bp was present in the KO lines created with PpCESA6/7 KO vector but was not seen in WT. The expected 3' integration band of 1647 bp was observed in the same KO lines above but was also not present in WT. Primer sets used for 5' and 3' ends amplification are indicated as black arrows on the graph showing each gene's locus.

Wild type
Ppcesa4/6/7/10KO1 Ppcesa4/6/7/10KO2 Ppcesa4/6/7/10KO3  Ppcesa4/10KO leaf midribs have a moderate yet significant reduction in fluorescence intensity compared to wild-type. Fluorescence is significantly weaker in ppcesa6/7KO than it is wild-type and ppcesa4/10KO. Ppcesa4/6/7/10KO lines also have significantly decreased fluorescence intensity in leaf midribs compared to wild-type and ppcesa4/10KO, but there is no significant difference between ppcesa6/7KO and ppcesa4/6/7/10KO lines. Three independent genetic lines were tested in triplicate for each mutant genotype. The Gd11 line was used as the wild-type control and sampled in triplicate. Error bars indicate standard error of the mean (ppcesa4/10KO, n=3; ppcesa6/7KO, n=3; ppcesa4/6/7/10KO, n=3; wild-type, n=3).  Here, the bar graphs show changes of the three parameters in KO mutants. The height of the bar represents the ratio (KO mutant/wild-type). A ratio larger than 1 (indicated by a dotted line on each graph) indicates an increase of that parameter in KO lines compared to wild type and a ratio less than 1 indicates a decrease for KO lines. Error bars display standard error of the mean between each data set (n=3 for each data set). Statistical significant difference between KO mutant and wild-type is indicated by the "▲"sign. The statistically significant difference among KO mutants is indicated by the "•" sign. Raw measurements are reported in Table  S4.  Figure S1: B clade PpCESAs are not required for rhizoid development. P. patens wild-type (Gd11) and three independent ppcesa4/6/7/10KO lines were cultured on medium supplemented with1 μM naphthalene acetic acid (auxin) to stimulate rhizoid initiation, and inhibit leaf initiation. (A-C) Dark field images of wild-type leafless gametophores with numerous rhizoids. (D-F) Dark field images of ppcesa4/6/7/10KO1, ppcesa4/6/7/10KO2, and ppcesa4/6/7/10KO3 leafless gametophores with numerous rhizoids. None of the three ppcesa4/6/7/10KO lines showed defects in rhizoid initiation or growth.

Introduction
Cellulose is a biopolymer of β-(1,4)-glucose that forms microfibrils essential in land plant cell walls. It is synthesized by cellulose synthase complexes (CSCs) located on the plasma membrane . The CSCs of vascular plants were first observed to have a "rosette" structure and associate with the ends of microfibrils in freeze-fracture electron microscopy studies . Within the CSC, cellulose synthase catalytic subunits (CESAs) catalyze the synthesis of individual glucan chains and are currently the only verified functional subunits .
CESA genes are members of multigene families in vascular plants. Arabidopsis has 10 CESA genes . AtCESA4, AtCESA7, and AtCESA8 were first identified to be specifically involved in secondary cell wall deposition Scheible et al., 2001;Taylor et al., 2000;. Proteins encoded by these three genes physically interact and are exclusively required for assembly of CSCs in cells having thickened secondary walls . Mutations in AtCESA1, AtCESA3, and AtCESA6 cause primary cell wall defects . AtCESA3 and AtCESA6 interact with each other according to results of in vitro pull-down assays, and BiFC experiments show that AtCESA1, AtCESA3, and AtCESA6 can interact in vivo .
AtCESA2 and AtCESA5 are shown to be closely related and partially functionally redundant with AtCESA6 Timmers et al. 2009;Carroll et al. recently shown to have substantially reduced cellulose levels in midrib secondary cell walls of the gametophore leaf

Expression levels of PpCESAs in the knockout mutants by RT-qPCR
Correlated expression is expected for PpCESAs that reside in the same CSC. To test this, we performed RT-qPCR on RNA extracted from leafy gametophores collected from Gd11, ppcesa3KO, ppcesa8KO, ppcesa3/8KO, ppcesa4/10KO, ppcesa6/7KO, and ppcesa4/6/7/10KO plants. PpCESA8 was significantly upregulated in ppcesa3KO panel). The anti-PpCESA8 detected a 120 kD band in Act1pro::3xHA-PpCESA8 and Gd11. However, it also weakly detected band around 120KD in ppcesa8KO (Figure   2, right panel). When ppcesa3/8KO is used as a negative control, no band was detected. This indicates that anti-PpCESA8 has weak cross-reactivity with PpCESA3 in addition to detecting PpCESA8.

Protein expression profiling of the PpCESAs
We used western blotting to examine the protein expression patterns for PpCESA3, PpCESA8, and PpCESA6/7 at different developmental stages. Figure 3 shows that none of these proteins were detectable in the Day-6 wild-type cultures consisting of pure protonema. PpCESA3, PpCESA8, and PpCESA6/7 were detected in Day-10 cultures, which contain protonema, emerging gametophore buds, and young gametophores. Finally, much larger amounts of these three PpCESAs were detected in Day-21 cultures, which contain numerous fully developed leafy gametophores. This shows that PpCESA3, PpCESA8, and PpCESA6/7 exhibit similar expression profiles, with highest expression in the gametophores, consistent with roles in gametophore development.

Interactions between the PpCESAs
Based on the similarity of their mutant phenotypes, correlated gene expression, and protein expression profiles, we hypothesized that PpCESA3, PpCESA6, PpCESA7, and PpCESA8 physically interact with each other to form hetero-oligomeric complexes. To address this question, we carried out Co-IP experiments on detergentsolubilized protein extracts from the 15-day-old leafy gametophores of transgenic P.
patens lines that expressed HA-tagged PpCESAs under the control of their native knocked out, and vice versa). One exception is up-regulation of PpCESA8 to compensate for the loss of PpCESA3 . Taking results of phenotype analysis into consideration, it was suggested that PpCESA3 and PpCESA8 have interchangeable functions and may compete for the same positions in the CSCs . In addition, our Co-IP results show that PpCESA3 is coimmunoprecipitated with PpCESA8 and vice versa ( Figure 4A). PpCESA8 appears to be dominant over PpCESA3 (in the amount of protein) in wild-type P. patens. This is based on the observation that ppcesa8KO has an obvious reduction in the leaf midrib cellulose deposition, which is not shown in ppcesa3KO .
In Arabidopsis, loss of a single CESA usually is enough to cause either obvious morphological defects or even lethal developmental defects , which means Arabidopsis CESAs are functionally distinct. In contrast, the PpCESAs show less functional differentiation. With the exception of PpCESA5 , we have to knock out at least two PpCESAs from the same clade to observe a strong phenotype. In P. patens, CESAs within the same subclade (A-clade or B-clade) appear to be functionally interchangeable. The major functional differences might only exist between PpCESAs from different sub-clades.
For instance, the mutant phenotype of ppcesa3/8KOs can be rescued by expressing PpCESA8, PpCESA3 or PpCESA5 under control of the PpCESA8 promoter, but cannot be rescued by expressing any of the clade-B PpCESAs using the same PpCESA8 promoter . The "no gametophore" phenotype of In contrast, the AtCESAs have very limited interchangeability, with partial rescue only of atcesa3 by AtCESA3pro::AtCESA7 and atcesa8 with AtCESA8pro:: AtCESA1 .
Taken together, including this study, the current evidence indicates that secondary cell wall in the moss P. patens are synthesized by an obligate hetero-oligomeric CSC assembled from PpCESAs from both clade A and clade B. Our findings combined with phylogenetic analysis  suggest that heterooligomeric CSCs arose in both mosses and seed plants through independent evolution. Characterization of the PpCESAs shows some consistency with the theory of constructive neutral evolution which can be used to explain the evolution of the hetero-oilgomeric CSCs. According to the hypothesis, after ancestral gene duplication, simple and high-probability mutations are considered to be a sufficient cause leading to the increased complexity of a multi-protein complex (Doolittle, 2012;Finnigan et al., 2012). Most of these mutations are thought to be insufficient to cause changes in protein biochemical output. However, mutations occuring at the protein-protein binding interface can cause the mutant proteins to lose the ability to interact with the others members in the complex. In that case, a hetero-oligomeric complex eventually might evolve by complementary loss of asymmetric interactions of certain protein subunits in the original homo-oligomeric complex (Doolittle, 2012;Finnigan et al., 2012). We have shown that ppces6/7KOs phenocopy ppcesa3/8KOs; both show defects in secondary cell walls . This suggests that clade-A PpCESAs and clade-B PpCESAs carry out non-overlapping functions after neofunctionalization of the common ancestral PpCESA. The distinct functions of these PpCESAs might be caused by mutations at the binding sites, according to the theory above. This assumption is further supported by the Co-IP results here together with results of yeast two hybrid assay showing PpCESA8 cannot interact with itself (unpublished data). In addition, the promoter-swap assays mentioned above suggest there is no major functional difference among the paralogues from the same PpCESA clade, which is also consistent with constructive neutral evolution. To continue to test this theory, more precise characterization of PpCESAs need to be carried out to identify the binding sites of these PpCESAs.
Our study also indicate a possibility that PpCESA5 can form homo-oligomeric CSCs (Li et al., unpublished). We propose this hypothesis based on the distinct isoform function , unique gene expression pattern as well as the fact that mosses and seed plants derived from the common ancestor which had a single CESA and consequently homo-oligomeric CSCs (Roberts & Bushoven, 2007). If this hypothesis is proved to be true, it will be another evidence supporting the constructive neutral evolution. But, no matter what the answers will be, implications provided by these studies will be helpful for understanding the roles of the different CESAs composing seed plant CSCs.

Culture conditions
Wild-type and all transgenic P.patens lines used in this study were maintained on BCDAT plates and propagated by subculturing weekly as described . To produce growing leafy gametophores, explants of 7-day-old protonemal tissue was transferred onto BCD plates and cultured for 15 days before being harvested for experiments.

Vector construction and transformation
All the PpCESA KO lines used in this study were created previously. Construction of knockout vectors and transformations for making those lines were described in . Transformation for making quadruple ppcesa4/6/7/10KOs was described in an unpublished manuscript (Li et al., unpublished).
PpCESA overexpression lines used as positive controls for testing antibody specificity were selected from transformations of ppcesa5KO-2 with vectors driving expression of 3X-HA-tagged PpCESA3, PpCESA7 or PpCESA8 under control of the rice Actin1 promoter (Scavuzzo-Duggan et al., in revision).
Expression vectors for HA-tagged PpCESAs under control of their native promoters were created using Multi Site Gateway Pro (Invitrogen, Grand Island, NY, USA). The HA-PpCESA5, HA-PpCESA7, and HA-PpCESA8 coding sequences were amplified from cDNA clones pdp24095, pdp38142, and pdp39044 (RIKEN BRC), respectively, using forward primers containing a single HA tag coding sequence flanked by an attB5 site and a reverse primer flanked by an attB2 site (Supplemental Table 2). HA-PpCESA3 was amplified from a cDNA clone describe previously (Scavuzzo-Duggan et al, in revision) using appropriate primers (Supplemental Table 2 Total RNA extraction from gametophore leaves followed by cDNA conversion was carried out as described . RT-qPCR analysis was performed using the ΔΔCt method (Livak & Schmittgen, 2001) of relative quantification with a Roche Lightcycler 480, using SYBR Green I Master Mix (Roche) to monitor doubled strand DNA synthesis. Primers for PpCESA detection were as used in , and primers for reference genes, actin and v-Type h+ translocating pyrophosphatase, were as used in Bail et al. (2013).

Generation of monoclonal anti-PpCESAs
Peptide antigens, designed to regions of each PpCESA for the purpose of raising antibodies specific for each isoform (Table S2), were synthesized chemically and injected into New Zealand white rabbits (Covance Inc., Princeton NJ USA). For PpCESA6 and PpCESA7, it was not possible to generate two unique peptides in order to raise antibodies to differentiate these isoforms. The peptides were conjugated, via the cysteine residue, to Sulfolink Immobilization resin (Thermo Fisher Scientific) according to the manufacturer's instructions. The purification of PpCESA antibodies from total serum was carried out by affinity chromatography. Briefly, 10 mL of serum, buffered with WB (20 mM NaHPO 4 , pH7.2, 50 mM NaCl) was incubated with the resin-linked peptides for 18h at 4°C. The resin was loaded into a column and the flow through was passed over the resin twice. The resin was washed with 20 mL of WB followed by 10 mL of WB containing an additional 250 mM NaCl. Antibodies were eluted from the resin using 5 mL of EB (100 mM glycine, pH 2.5). Fractions of 250 μL containing NB (50 µL 1 M Tris-Cl, pH 8.0) were collected and mixed immediately to neutralize pH. Fractions containing PpCESA antibodies were identified by absorbance at 280 nm and combined. Glycerol was added to 30% and CESA antibodies were stored at -80°C. The specificity of each antibody was tested by western blotting against P. patens protein extracts.
stand (Thermo Fisher Scientific), and the unbound sample was removed. 400 L of TBS-T buffer  was added to the tube and gently mixed.
Beads were collected again by magnetic stand, and the supernatant was discarded.
This step was repeated twice. For the last wash, 400 L of ultrapure water was added to the tube and gently mixed. Beads were collected on a magnetic stand, and the supernatant was removed. For elution, 50 L of 2X SDS-PAGE sample buffer  and 50 L of ultrapure water were added to the tube, and gently mixed. The tube was incubated at 95°C-100°C for 10 min. Finally, beads were magnetically separated, and initial input (total protein), unbound fraction, wash, and IP eluate were stored at -20 o C for up to three months and used for western blot analysis. Gel electrophoresis and western blot using anti-PpCESA3, anti-PpCESA6/7, and anti-PpCESA8 antibodies were carried out as described in Scavuzzo-Duggan et al. .

Statistical analysis
One-way Analysis of Variance (ANOVA) followed by post-hoc Tukey Honest Significant Difference (HSD) test was performed using "R" programming (Vienna, Austria; http://www.R-project.org/) to identify the potential significant difference in each assay.

Supplemental Materials
Table. S1. Primers for amplification of HA-tagged PpCESAs. Table. S2. Peptide antigens, designed to regions of each PpCESA, used to raise specific antibodies for each PpCESA isoform.      Western blots of microsomal proteins isolated from wild-type P. patens cultures and probed with anti-PpCESA3, anti-PpCESA8, and anti-PpCESA6/7. Explants from protonema cultured on solid medium overlaid with cellophane for 6 days were cultured on solid medium without cellophane and harvested after 6 days (protonema only), 10 days (protonema and young gametophores) and 21 days (gametophores). Equal loading of protein Ponceau S Staining.   Figure