A Specialized Outer Layer of the Primary Cell Wall Joins Elongating Cotton Fibers Into Tissue-like Bundles

Cotton provides the world’s dominant renewable textile fiber, and cotton fiber is valued as a research model because of its extensive elongation and secondary wall thickening. Previously, it was assumed that fibers elongated as individual cells. In contrast, observation by cryo-field-emission scanning electron microscopy of Gossypium hirsutum fibers developing in situ within the boll demonstrated that fibers elongate within tissue-like bundles. These bundles were entrained by twisting fiber tips and consolidated by adhesion of a cotton fiber middle lamella (CFML). The fiber bundles consolidated via the CFML ultimately formed a packet of fiber around each seed, which helps to explain how thousands of cotton fibers achieve their great length within a confined space. The cell wall nature of the CFML was characterized using transmission electron microscopy, including polymer epitope labeling. Toward the end of elongation, up-regulation occurred of gene expression and enzyme activities related to cell wall hydrolysis, and targeted breakdown of the CFML restored fiber individuality. At the same time, losses occurred in certain cell wall polymer epitopes (as revealed by Comprehensive Microarray Polymer Profiling) and sugars within non-cellulosic matrix components (as revealed by GC-MS analysis of derivitized neutral and acidic glycosyl residues). Broadly, the data show that adhesion modulated by an outer layer of the primary wall can coordinate the extensive growth of a large group of cells and illustrate dynamic changes in primary wall structure and composition occurring during the differentiation of one cell type that spends only part of its life as a tissue. characterization epitopes cell of a new group of chitinase-like genes are preferentially bulged these (G-I) JIM5 labeled CFML and the more 24 DPA during secondary wall deposition, labeling on the primary wall that remained after CFML degradation was more intense than DPA. Bars in: (A) = 5 µ m; (B) = 200 nm; (C) = 1 µ m; (F) = 300 nm for D, E, F, and I; and (G) = 150 nm for G and H.


INTRODUCTION
Cotton fiber from Gossypium hirsutum is the world's dominant natural textile fiber (Chen et al., 2007). This unique single cell is also a valuable research model because of its extreme length (> 2.25 cm) and thick (3-6 µm) cellulosic secondary wall (Kim and Triplett, 2001). Cotton fiber morphogenesis initiates with the bulging of selected seed epidermal cells near the day of flowering (anthesis) and is described temporally by days post anthesis (DPA). Seed and fiber development occurs within (typically) four individual carpels (locules) within the ovary (or cotton boll). A typical locule of G. hirsutum cv. Deltapine 90 contains 8 to 9 developing seeds each with ~14,500 fibers (Bowman et al., 2001). Therefore, ~123,000 fibers reach a final length of ~2.7 cm in the confined space of one locule, which expands to only ~5 cm 3 internal volume (inclusive of seed volume). Previously, there has been little consideration of how the extreme elongation of so many cells could be optimized within a confined space. Because cotton fibers initiate as individuals and form a fluffy mass after boll opening, it was previously assumed without direct evidence that they also progressed through the entire elongation stage as individuals.
Between elongation and secondary wall thickening in G. hirsutum cotton fiber, there is a "transition stage" of development when elongation slows and the wall thickening begins via deposition of the "winding" layer. Like the 'S1' layer in wood fiber, the winding layer has intermediary cellulose content and microfibril angle (relative to the longitudinal fiber axis) compared to the primary and secondary walls. At the end of the transition (typically at ~ 21 DPA), elongation and winding layer deposition cease while ~ 95% cellulose deposition in the secondary wall begins (Meinert and Delmer, 1977). Finally, the boll opens and the fiber and seed dries at ~45 to 60 DPA (Basra, 1999).
Since cotton fibers can be easily obtained in bulk as a single cell type, their staged cell wall synthesis offer unique opportunities for understanding the mechanisms of construction and deconstruction of cell walls. This is foundational knowledge to support optimization of plant biomass as feedstocks for biofuels production as well as improvement of fiber crops as renewable materials (Pauly and Keegstra, 2008). Previously, there was little specific knowledge about cotton fiber cell wall structure or understanding of how cell walls might control fiber differentiation beyond traditional concepts of thin, extensible, primary walls vs. thick, celluloserich, secondary walls (Basra, 1999). A notable exception was the demonstration of biphasic distribution of wall polymers in 1 to 2 DPA fibers (Vaughn and Turley, 1999), but functional consequences of this wall structure were unknown. Similarly, the transition stage involves upregulation of genes putatively encoding cell wall hydrolases as well as the loss of certain neutral cell wall sugars and reduction in pectin and XG mass (Meinert and Delmer, 1977;Shimizu et al., 1997;Tokumoto et al., 2002 and2003;Guo et al., 2007), but a cell biological explanation for these developmental events was lacking. This paper reports the surprising discovery that a specialized cell wall layer, which we called the cotton fiber middle lamella (CFML), fuses elongating cotton fibers into tissue-like bundles during elongation. The CFML was visualized in both cryo-field-emission scanning electron microscopy (cryo-FE-SEM) and transmission electron microscopy (TEM) and shown, through TEM immunolabeling, to contain cell wall polymers including HG and XG. The means of forming and consolidating cotton fiber bundles via the CFML, as well as the highly organized that: (a) a specialized outer layer of the primary wall can coordinate the extensive growth of a large group of cells; and (b) dynamic changes in primary wall structure and composition occur in one cell type to facilitate its oscillation between differentiation as an individual, a simple tissue, then an individual once more.

Time Course of Fiber Development
Fiber harvested from G. hirsutum cv. Deltapine 90 was used for the biochemical and gene expression experiments. Deltapine 90 is an advanced cultivar of upland cotton (the type most commonly grown world-wide), and it was the source of a unique set of ESTs biased toward genes expressed during early secondary wall deposition (Haigler et al., 2005). Because the temporal progression of cotton fiber development varies somewhat with genotype and is particularly sensitive to temperature differences, we determined the time-course of fiber length and weight gain, as well as cellulose content, during growth in an air-conditioned greenhouse with a 26°C/22°C day/night cycle ( Fig. S1). Under these highly controlled temperature conditions, late elongation and the first low-level rise in cellulose percentage within the winding cell wall layer occurred concurrently at 17 to 22 DPA, representing the transition stage. At 22 DPA, secondary wall deposition via ~ 95% cellulose deposition began, as shown by the subsequent rapid increases in fiber weight, fiber weight per unit length, and cellulose content, with elongation slowing greatly after that time (Fig. S1).

Microscopic Evidence for a Transient, Tissue-like Phase of Cotton Fiber Development
In contrast to previous conceptions of elongation of cotton fibers as individual cells, cross sections of elongating G. hirsutum fibers observed in the light microscope showed bundles of tightly packed cells (Fig. 1A) reminiscent of a simple plant tissue [a tissue composed of only one cell type (Evert, 2006)]. Higher magnification views in the TEM were also consistent with the existence of tissue-like aggregates of fiber (Fig. 1B). Hundreds of fiber cross-sections have been observed in one bundle, and Figure S2 proves that these bundles of cells do represent fiber crosssections.
Both chemical-and cryo-fixation prior to TEM observations revealed a thin (~300-400 nm), apparently coherent (double), wall between adjacent fibers in many locations ( Fig. 1, B and D). There were no discernible boundaries between adjacent fibers over large regions, proving that these bundles were not merely aggregates of co-aligned individual fibers. In addition, we sometimes saw material (Fig. 1E, arrow) that appeared distinct from the inner primary wall surrounding the protoplast (Fig. 1E, arrowhead). In this specimen, the outer material was probably attached to only one fiber because there was no lower fiber to keep it compressed within a fiber bundle. Occasionally 1 to 2 µm wide material-filled bulges occurred between adjacent fibers after both chemical and cryo-fixation (Fig. 1, F and G). In longitudinal sections, the bulges were ~4 to 12 µm long and randomly located along the fiber length between regions of coherent (double) primary walls (Fig. S3). The tissue-like bundles of fiber persisted during elongation, but the material between fibers disappeared and fiber individuality was restored as secondary wall deposition commenced ( Fig. 1H; the asterisk indicates the cleared interfiber space).

Cryo-FE-SEM Revealed How Fiber Bundles Formed and Their Packing Within the Boll
Using cryo-FE-SEM, we determined how the fiber bundles originated and their implications in situ within the cotton boll. After cutting a window in the boll wall, followed by rapid freezing and mounting on the ultra-cold cryo-FE-SEM stage, cotton fibers were observed in their life-like state within the boll. By this method, the outermost layer of fibers of each seed was observed. Shortly after initiation, the tips of elongating fibers began to intertwine, which appeared to nucleate and entrain fiber bundles ( Fig. 2A). Starting at 2 DPA (data not shown), a distinct outer layer of material was observed to fuse adjacent fibers (Fig. 2B, double arrow), and this material was present even at the fiber tips (Fig. 2C, arrow). During early elongation, the forming bundles also engaged in directional, curved, elongation (Fig. 2D) as a precursor to forming wave-like paths as elongation progressed (see below). Up until 3 DPA, many fiber tips were visible on the outer surface of the fiber mass, but afterwards only a few tips were visible, apparently because most of them turned inward toward the seed (Fig. 2E).
After specialized cryogenic processing for paraffin sectioning, sections of developing ovules with attached fiber showed the ovule (dark circle) surrounded by a packet of fiber with a distinct, relatively smooth surface because fiber edges, not tips, dominated the surface (left panel, Fig. 2F; see Fig. S4A for a higher magnification view). Throughout the depth of the fiber packet, fiber bundles were running at various angles relative to the plane of the section, but, where they traveled perpendicularly, bundled fiber cross-sections were once again observed (right panel, Fig. 2F, and also Fig. S4B). The upper surface of the fiber mass (appressed to the inner boll wall) showed tightly packed bundles of fiber in meandering, wave-like, patterns (Fig.   2G), and organized fibers filled the locule (Fig. S4C). Because of the constrained space within the locule, fibers did not change their overall packing pattern even after they once more became separate individuals during secondary wall deposition (data not shown).

TEM Immunolabeling Demonstrated the Cell Wall Nature of the CFML
Higher magnification views of the CFML and its degradation products were consistent with its existence as a specialized layer of adhesive cell wall polymers, distinct from the inner primary wall. In cryo-FE-SEM, the CFML appeared sheet-like as it stretched between two fibers at a separation point (Fig. 3A), which is suggestive of its adhesive, or glue-like, properties. After cryo-fixation for TEM, the fibrillar nature of the CFML was evident in a "fringed" outer layer of the primary wall, probably where fibers had separated during processing (Fig. 3B). The fibrillar degradation products of the CFML at the transition stage resembled cell wall components, and they cleanly separated from the inner primary wall (Fig. 3C, arrowhead) before they finally disappeared from inter-fiber spaces (Fig. 1H).
Twelve monoclonal antibodies (mAbs) recognizing epitopes in cell wall polymers (and one polyclonal antibody) gave positive results in TEM or fluorescence immunolabeling to show that the CFML contained cell polymers. Table S1 summarizes the probes used and all results. Figure S5 shows the larger cellular context of the TEM immunolabeling results shown in Figure   3 and demonstrates that immunolabeling was specific for cell wall regions of the samples.
Antibodies that labeled CFML material at 10 and 17 DPA included CCRC-M1 (Fig. 3, D and E), which recognizes the α-Fuc-(1,2)-ß-Gal side chain of Fuc-XG. In addition, nine of eleven other XG antibodies tested labeled positively, including CCRC-M106 with specific affinity for the XXFG epitope [containing α-Fuc-(1,2)-beta-Gal] and a polyclonal anti-XG that does not depend on a fucosylated side chain for binding and likely recognizes the XG backbone (Lynch and Staehelin, 1992) (Table S1). Positive labeling of the CFML was also obtained with JIM5, which recognizes short stretches of HG pectin, typically with a relatively low degree of methylesterification (Fig. 3, G and H). CCRC-M38, which is able to bind fully de-esterified HG, also positively labeled the CFML (Table S1; images not shown). Immunolabeling was negative at 10 and 17 DPA for four antibodies recognizing the RG-1 backbone and five antibodies recognizing possible RG-1 side chains (Table S1).
At 24 DPA, the CFML had been degraded so that the space between fibers was cleared ( Fig. 3, F and I). Interestingly, however, the density of JIM5 labeling in the residual primary wall was higher at 24 DPA during secondary wall deposition than it was during elongation (compare the numerous colloidal gold particles on the primary wall, labeled cw, in Fig 4I). After treatment of sections with Na 2 CO 3 as a de-esterifying agent, the JIM5 signal became markedly brighter as expected at all DPA tested (Fig. 4, D to F), whereas the JIM7 signal dimmed but did not disappear (data not shown). These results are consistent with an overall relatively high level of HG esterification in the cotton fiber primary wall during elongation, followed by de-esterification as elongation ends. However, there was evidence of zonation in the degree of HG esterification within the wall during elongation as shown by superimposing the images of 10 DPA fiber labeled with JIM5 before (Fig. 4A, signal changed to red) and after Na 2 CO 3 treatment (Fig. 4D, signal retained as green). In the composite image (Fig. 4J), the red JIM5 signal from the untreated sections occurred frequently in the middle of the double primary walls between paired fibers, as well as in corners between three fibers. These data are consistent with the TEM results for JIM5 labeling (Fig. 3, G and H), and cumulatively the data indicate that the CFML is enriched in HG with a relatively low degree of esterification.

Polysaccharides
Given that the CFML existed only during cotton fiber elongation, we sought evidence for biochemical changes in the cell wall correlated with its breakdown at the transition stage. One technique used was Comprehensive Microarray Polymer Profiling, or CoMPP, that involves sequentially extracting cell wall polymers. The cell wall extracts are then spotted as arrays and probed with mAbs or carbohydrate binding modules (CBMs) ( Table S1). Quantified spot signals provide semi-quantitative information about epitope or polymer occurrence (Moller et al., 2007;Fig. 4K). Three extractants [CDTA (a Ca ++ chelator), NaOH, and cadoxen] were used in succession to extract wall polymers that were less-then more-tightly bound within cell walls.
The consistency of CoMPP data with known aspect of cotton fiber differentiation is shown by: (a) strong signals from 6 to 30 DPA for CDTA-soluble HG (JIM5) and NaOH-soluble XG (CCRC-M1), which likely reflect the matrix components of the persistent inner primary wall; and (b) a high signal for NaOH-extractable ß-1,3-glucan (callose; detected by BS 400-2) beginning at 17 DPA. Other analyses have shown that callose is deposited just outside the cotton fiber at the transition stage (Maltby et al., 1979;Salnikov et al., 2003). In addition, HG deesterification at the end of elongation detected by fluorescence immunolabeling was paralleled by inverse changes for the signals from CDTA-soluble JIM5 (increased) and JIM7 (decreased) after 21 DPA.
We examined the CoMPP results for patterns of change that correlated temporally with the transition-stage degradation of the CFML. Such changes were evident in the cadoxen fraction: both the CCRC-M1 and JIM5 signals were approximately constant during elongation, approximately doubled between 17 and 19 DPA, and then declined with similar patterns thereafter. Since CCRC-M1 and JIM5 did not label the thickening cell wall (see lack of colloidal gold on areas labeled 'scw' in Fig. 3F and I), their maximum signals in the cadoxen fraction at 19 DPA in CoMPP profiling could reflect breakdown of the CFML (see discussion). Notably, this pattern was not shared by JIM7 in the cadoxen fraction, which is consistent with evidence that low esterified HG is enriched in the CFML (

Singh et al. 12
CoMPP profiling, LM8 recognizing an RG-1 side chain, xylogalacturonan, was tested because it labels walls of some cells that undergo complete detachment such as those in the root cap (Willats et al., 2004). No signals above 5 (the threshold for interpretation) were detected in cotton fiber extracts, and LM8 was also negative in immunolabeling experiments (data not shown). Similarly, the antibody 2F4 that recognizes a Ca ++ -dependent epitope in HG (Bush and McCann, 1999 and references therein) did not yield positive results when tested with optimum procedures in TEM or fluorescence analyses (Table S1; Supplemental Protocol S2).
Analysis of sugars in hot-water-soluble vs. insoluble cell wall matrix polymers (excluding cellulose) further documented changes consistent with CFML degradation at the transition stage. GC-MS analysis of trimethylsilyl (TMS)-methylglycoside derivatives allowed quantitation of both neutral and acidic glycosyl residues ( Fig. 5A; complete data are in Table S2 and Table S3). Expressing the data as nmol sugar/mg dried fiber revealed changes over developmental time within a unit of fiber cell wall mass. A 56% decrease in the total amount of non-cellulosic sugars in the hot-water insoluble fraction occurred between 17 and 19 DPA. The amounts of Ara, Fuc, Gal, GalA, and Rha showed significant decreases by 19 and/or 21 DPA compared to 17 DPA (p < 0.05). For Ara, Gal, and Rha, these results paralleled other cotton fiber neutral sugar analyses that showed decreases near the time of onset of secondary wall deposition, as well as low levels of Fuc throughout fiber development (Meinert and Delmer, 1977). Notably, these decreases were reflected in the lack of increase (trend downward) in fiber weight/length between 17 to 21 DPA even though cellulose content within the newly deposited winding layer was increasing simultaneously (Fig. S1). Such an effect could arise from CFML degradation. By 30 DPA, the amounts of Ara and Rha in the hot-water-soluble fraction cell wall fraction had increased significantly compared to 17 DPA (p < 0.05), and Gal and GalA showed the same trend (Table S3) In the microarray data, the group of genes with significant up-regulation at 20 DPA compared to 6 and 10 DPA (p < 0.05) included sequences apparently encoding cell wall modifying enzymes. Note that in the "Cotton_Microarray_Summary_Annotated" file posted at NCBI genes that are up-regulated at 20 DPA have negative 6/20 and 10/20 log 2 ratios. These included genes encoding two putative pectin modifying enzymes: (a) a pectin methylesterase (PME) (NCBI EST ID: CO495780, Ghpe-E) with 10/20 DPA log 2 ratio = -2.2 (4.6-fold upregulation at 20 DPA vs. 10 DPA); and (b) a polygalacturonase (PG) (NCBI EST ID: CO492632, Ghpg-A3) with 10/20 DPA log 2 ratio = -2.48 (5.7-fold up-regulation at 20 DPA vs. 10 DPA). These microarray results were confirmed by qPCR between 6 and 30 DPA of fiber development (Table S4 contains qPCR primers). Corresponding to its higher dynamic range (Busch and Lohmann, 2007), qPCR showed 4,229-fold and 12-fold up-regulation at 20 vs. 10 DPA for Ghpe-E and Ghpg-A3, respectively (Table S5)

Singh et al. 14
The qPCR analyses were expanded to include a total of 31 genes encoding representatives of several families of cell wall modifying proteins (Fig. 5, C-E; Table S5). Fiber developmental marker genes were also included: (a) GhRAC13 with peak expression at 21 to 22 DPA during the transition (Kim and Triplett, 2004); and (b) GhCTL1 with peak expression from 23 DPA onwards during secondary wall deposition (Zhang et al., 2004) (Fig. 5B). GhRAC13 encodes a small GTPase that can modulate H 2 O 2 production (Delmer et al., 1995), and GhCTL1 encodes a protein with an obligatory but unknown role in secondary wall cellulose synthesis (Brown et al., 2005). In order to emphasize developmental patterns of expression, the graphs show percentages of the maximum expression observed for each gene, and non-normalized data are in Table S5. The graphs show genes with transition-stage expression that was increased by a factor of at least four compared to 10 DPA (during rapid elongation) in four sequence families related to modification or hydrolysis of cell wall HG or pectin: (1)  and (4) polygalacturonase-like (pg) (Fig. 5E). In Figure 5, C to E, all but Ghpe-E reached their peak expression at 20 to 22 DPA toward the end of the transition when the CFML was degraded, similarly to GhRAC13 (Fig. 5B). The up-regulation of Ghpe-E was later and similar to GhCTL1, although Ghpe-E expression peaked at 23 DPA whereas GhCTL1 expression was sustained as secondary wall deposition proceeded (compare Fig. 5B to Fig. 5D).
The enzymes encoded by cell wall hydrolase genes could help to degrade the CFML, and relevant enzyme activities were demonstrated in fiber extracts. As before, percent maximum activity is graphed in each case (Fig. 5F), and non-normalized data are in Table S6. XTH enzymes with an H 2 O acceptor may act as XG endo-hydrolases (XEHs), degrading the XG backbone (Vicente et al., 2007). XEH activity peaked at 19 DPA, which correlates with the timing in CoMPP of the transitory increase in the cadoxen-soluble CCRC-M1 epitope within Fuc-XG (Fig. 4K). ß-galactosidases hydrolyze terminal, non-reducing, ß-D-galactosyl residues from ß-D-galactosides, including pectin side chains, and this enzyme activity was elevated between 19 to 23 DPA in fiber extracts. ß-galactosidase activity could account for the posttransition disappearance of the LM5 epitope, putatively in galactan, in the CoMPP NaOH fraction (Fig. 4K). NaOH-extractable LM5 epitopes may arise from a subset of pectin galactan domains that are covalently linked to other cell wall polymers (Moller et al., 2007)  Singh et al. 16

The Material Properties of the CFML
The existence of fiber bundles and apparently coherent "double" walls between adjacent fibers supports the adhesive properties of the CFML under physiological conditions. In contrast, as occurs for many synthetic adhesives (Bascom and Cottington, 1976), the CFML tended to crack upon exposure to ultra-cold temperature during cryo-fixation for cryo-FE-SEM or TEM freeze substitution. When this occurred, joined fibers separated to reveal distinct CFML material as the outer layer of the primary wall ( Fig. 1E; Fig. 3, A and B). Separated fibers were less frequent in cryo-FE-SEM views, probably because largely immobilized fibers still within the boll were observed immediately. In contrast, fibers on isolated seeds had many opportunities to separate during the prolonged freeze-substitution protocol with its multiple solvent exchanges prior to resin embedding for sectioning. Nevertheless, the observation of some joined fibers in sections after cryofixation and freeze substitution (Fig. 1, D and E) provided important confirmation of the reality of fiber bundling via an adhesive CFML in vivo because cryo-fixation instantly immobilizes cellular structure (Dahl and Staehelin, 1989). Because of frequent cracking of the CFML upon freezing, chemical fixation, which provided similar images of the cell wall ( Fig. 1, B, C, and F), was usually preferred for TEM immunolabeling experiments.
Other middle lamellae often contain HG and may include rhamnogalactoronan-I (RG-I) pectin and/or XG (Ishii, 1981;Moore et al., 1986;Moore and Staehelin, 1988;Bush and McCann, 1999), whereas the CFML clearly contains HG and XG. The middle lamella within tension wood of sweetgum trees, like the CFML, contains HG (labeling strongly with JIM5 and weakly with JIM7) and XG (labeling strongly with CCRC-M1 and CCRC-M38) (Bowling and Vaughn, 2008). Cumulatively, the positive immunolabeling results (Fig. 3, Fig. 4, and Table S1) support the possibility that the XG in the CFML includes regions with and without fucosylated side chains, while part of the HG has relatively low and/or no esterification. Low ester HG pectin has been commonly found in the middle lamella of dicotyledenous plants (Marry et al., 2006, and references therein). Possibly related to the earliest stages of CFML synthesis, initiating cotton fibers have bi-layered primary walls with a thin outer sheath enriched in low-ester, JIM5reactive, HG, whereas the inner layer was enriched in JIM7-reactive HG. The predominant occurrence of JIM5 epitopes in the outer sheath of these young fibers (Vaughn and Turley, 1999) could indicate initial synthesis of the CFML. Alternatively, the outer sheath could be an

Singh et al. 17
initiation-specific fiber wall layer, since, unlike the CFML, it was not enriched in XG (Vaughn and Turley, 1999).
The CMFL is also likely to contain special molecules that confer its adhesive properties.
Like the CMFL, the inner primary wall labeled with HG antibodies (JIM5, JIM7, and CCRC-M38) and some XG antibodies (anti-XG, CCRC-M1, and CCRC-M88). In contrast, several XG antibodies labeled the CFML, but not the inner primary wall (Table S1), suggesting that the CFML may contain specific forms of XG. In addition, although the list of cell wall probes is expanding, unusual variants of cell wall molecules are not likely to be comprehensively targeted.
As yet unknown constituents of the CFML, including cell wall proteins or phenolics, could help to confer its adhesive properties. Other supporting evidence for special wall polymers in the CFML will be discussed in the context of cell wall changes during its breakdown in the transition stage (see below).

Activities Related to Cell Wall Modification and Hydrolysis
Targeted cell wall hydrolysis occurs at the transition stage because the inner primary wall persists while the CFML disappears (along with its CCRC-M1-and JIM5/JIM7-reactive polymers), thereby restoring fiber individuality (Fig. 3, C, F, and I). As the CFML was degraded: (a) Fuc and Gal (plausibly in the XG side chain recognized by CCRC-M1) as well as GalA (plausibly in HG recognized by JIM5) within hot water-insoluble matrix polymers diminished; and (b) cadoxen-soluble CCRC-M1 and JIM5 epitopes (but not JIM7 epitopes) showed a transient increase in the CoMPP profile. Since CCRC-M1 and JIM5 did not label the thickening cell wall (Fig. 3, F and I), the latter CoMPP pattern could reflect breakdown of the interfacial CFML. As observed here in the CoMPP profile (Fig. 4K), substantial amounts of HG and XG in primary walls are typically CDTA-and NaOH-soluble, respectively, and the corresponding signals were strong and relatively consistent in polymers extracted from cotton fibers at 6 to 30 DPA. In contrast, transition-stage changes in the JIM5 and CCRC-M1 epitopes were observed only in the strong solvent, cadoxen. Therefore, we propose that a special polymer network that imparts the adhesive nature of the CFML also confers resistance to extraction in CDTA, NaOH, and even cadoxen until hydrolytic changes begin to occur during the transition. In such a case, the transition-stage increase in the CoMPP signals for CCRC-M1 and JIM5 epitopes in the cadoxen fraction could reflect increasing extractability of CFML polymers at the onset of CFML degradation. Similarly, JIM5-reactive antigens in the middle lamella of beet roots, but not the inner primary wall, were resistant to extraction by a Ca ++ -chelator (Marry et al., 2006). We could not immunolabel polymers that typically exist as RG-1 side-chains in situ, but LM5-and LM6reative antigens, typically galactan and arabinan, were detected in CoMPP profiling. Because of the sharp decline of polymer(s) containing the extractable LM5 epitope during the transition stage (Fig. 4K), it will be worthwhile to explore further a possible role of galactans in cotton fiber adhesion. Detailed analysis by qPCR over a finely resolved developmental time-course showed transition stage peaks in expression of genes in several classes required for XG or pectin modification or hydrolysis. Pectin esterases (e.g. PME that catalyzes pectin demethyesterification) and polygalacturonase-like (PG-like) enzymes degrade HG, with prior removal of methyl or other esters (by pectin esterases) making the backbone more susceptible to hydrolysis (by polygalacturonases) (Vicente et al., 2007). Therefore, PME activity up-regulated at the transition stage could facilitate CFML degradation and cotton fiber separation in a manner similar to the PME-mediated separation of pea root border cells (Wen et al., 1999) and Arabidopsis pollen tetrads (Francis et al., 2006). In contrast to positive results for PME activity, neither polygalacturonase nor pectin lyase activity were detected in vitro when commercial polygalacturonic acid from oranges was used as a substrate (data not shown). Perhaps the PGlike genes up-regulated in transition-stage fibers encode pectin degrading enzymes that are active against different pectin substrates, such as specialized CFML polymers. Table S5 also shows different transcriptional profiles for other genes within the same families, which shows that specific, developmentally regulated, hydrolase isoforms carry out fiber wall remodeling during the transition.
It is also possible that the cell-wall-related genes and enzymes activities up-regulated at the transition stage have additional or alternative roles in other processes. For example, after CFML degradation occurred, both immunolabeling and the CoMPP signals in the CDTA-soluble fraction showed that the degree of HG esterification in the primary wall was lower after the transition stage. Correspondingly, PME activity was up-regulated after 20 DPA. As normal elongation of Arabidopsis hypocotyls requires a degree of pectin esterification > ~60%

Relationship Between the Cotton Fiber Tissue and Cuticle
How can an elongation-stage cotton fiber tissue be reconciled with the cotton fiber cuticle given that the plant cuticle prevents cell and organ fusion (reviewed in Jarvis et al., 2003)? The existence of a hydrophobic "cuticle" has only been proven for mature cotton fiber. This strongly OsO 4 -positive, thin (~250 nm wide; ~1.3 % of fiber weight), morphologically homogeneous, outer layer contains an ~1:1 mixture of primary wall polysaccharides and wax/cutin molecules, and it is sharply distinct from the underlying cellulosic secondary wall (Yatsu et al., 1983;Cui et al., 2002;Degani et al., 2002;Wakelyn et al., 2007). This structure differs from the bilayered, cuticulated, surface of vegetative, epidermal, trichomes, which mixes gradually into an underlying cell wall containing several components (Marks et al., 2008). Within the cotton boll, elongating fibers would not require protection from a cuticle, and its absence during cotton fiber elongation would allow the fiber tissue to form. The absence of a cotton fiber cuticle during elongation, although contrary to common assumptions, is consistent with existing data. First, we did not observe any continuous cuticle-like demarcation between primary cell walls of adjacent fibers in bundles (Fig. 1, B, C, and D; Fig. 3G). Second, 2 DPA cotton fibers lacked the distinct cuticle layer that was easily seen in adjacent regular epidermal cells (Vaughn and Turley, 1999), and other images are consistent with an elongating cotton fiber leaving the ovule epidermal cuticle behind at its base ( Fig. 9 in Weis et al., 1999). Third, images from freeze-fracture transmission electron microscopy were interpreted in terms of a thinning of the cotton fiber cuticle during elongation. A multi-lamellate structure typical for cuticle was shown only at 2 DPA (Willison and Brown, 1977), and this could have existed at the fiber base. In contrast, at 10 DPA only a relatively smooth, thin, layer existed between primary walls of adjacent fibers (Willison and Brown, 1977); we propose that this was not cuticle, but instead the CFML described here. It is most likely that the wax and cutin components may be intercalated into (or attain their mature form within) the residual, inner, cotton fiber primary wall only after the CFML breaks down. Consistent with this possibility, the "hairy", OsO 4 -positive, surface of 24 DPA cotton fiber (Fig. 3, F and I; see also Salnikov et al., 2003) resembles the amorphous cuticle of Arabidopsis stem epidermis (Fig. 2.10e  provide further insight because cotton genes with best homology to three genes required for wax biosynthesis in Arabidopsis [CER3 (At5g57800), CER6 (At1g68530), and CER10 (At3g55630)] showed no significant change between 6 and 24 DPA in cotton fiber. How cuticle and wax are deposited remains largely unknown for all plant cell (Samuels, 2008;Pollard et al., 2008), and future research on single-celled cotton fiber will be worthwhile.

Other Implications of the CFML for Cotton Fiber Development
In addition to promoting fiber packing within the boll, the existence of the CFML is likely to have other implications for fiber development. For example, a fiber tissue may better withstand the high turgor that drives the rapid phase of elongation (Ruan, 2007), and the CFML may have mechanical properties that facilitate elongation (e.g. possibly allowing elongating fibers to slip past each other while still reinforced within a tissue-like bundle). Similar arguments based on consideration of biomechanical principles and the structure of parenchyma tissue have been made previously (Niklas, 1992).
The demonstration of cell wall mediated fiber adhesion also explains and unifies disparate prior observations about cotton fiber. We can now understand that twisting fiber tips on cultured ovules (Stewart, 1975) reflect a developmentally controlled mechanism that nucleates and entrains fiber bundles. Honeycomb-like aggregates of cross-sections of primary-wall-stage fibers observed after de-embedding prior to traditional SEM (Goynes et al., 1995) are now shown to reflect fiber bundles that existed in vivo. Fragmentary views of orderly fibers in replicas of the fiber surface inside the boll (Paiziev and Krakhamalev, 2004) are now demonstrated to exist on a large scale and to arise from the orderly packing of fiber bundles around each seed. Because the CFML breaks down before secondary wall deposition begins, a strong rationale now exists for changes in cell wall polymers and cell wall hydrolytic activity at the transition stage (Meinert and Delmer, 1977;Shimizu et al., 1997;Tokumoto et al., 2002Tokumoto et al., , 2003Guo et al., 2007, and new findings from current results). CFML hydrolysis may also explain the increasing extractability of glycan oligomers near the onset of secondary wall deposition in cotton fiber (Murray et al., 2001). sloughing of root border cells, pollen maturation, fruit ripening, dehiscence, and abscission (Roberts and Gonzalez-Carranza, 2007). However, none of these processes serves to remove a middle lamella that previously coordinated the extensive elongation of a large group of cells, implying that the degradation of the CFML is developmentally unique. Interestingly, several cotton genes with peak expression at the transition-stage, including Ghbgal12, Ghxth30,and Ghxth6,have (Vicente et al., 2007;Martel and Giovannoni, 2007). Since cotton fiber differentiation occurs within the developing boll (fruit), similarities between CFML degradation and edible fruit ripening are reasonable.

Future Research
The highly defined cotton fiber system will facilitate future research to determine: (a) linkage structure of CFML polymers; (b) how this defined cell wall polymer network develops adhesive properties; (c) activities of purified enzymes that hydrolyze particular CFML polymers; (d) mechanisms of synthesis and spatial localization of the CFML as the outer primary wall layer; (e) possible signaling effects of small molecules released by degradation of CFML polymers; (f) commonalities and differences of CFML-related mechanisms in various Gossypium species; and (g) effects on cotton fiber quality of genetic experiments to modify CFML-related mechanisms. Currently, we are analyzing G. barbadense, which has longer fiber, a longer period of fiber elongation, and greater overlap of elongation with secondary wall deposition, in order to gain additional insight into how the CFML modulates fiber development.

MATERIALS AND METHODS: Cotton Plant Growth
Flowers were tagged on the day of anthesis to allow harvesting of bolls of known age. All qPCR gene expression assays, enzyme assays, chemical fixation for TEM, growth curves,

Singh et al. 22
CoMPP profiling, and sugar analyses were performed on fiber of Gossypium hirsutum cv.
Deltapine 90 grown in an air-conditioned NCSU phytotron greenhouse. Plants (1 plant/6 L pot) were grown in a 1:2 (v/v) mixture of Redi-earth Plug and Seedling Mix ® (product number F1153, Sun Gro Horticulture Canada CM Ltd.) and gravel (#16, construction grade) and watered twice a day with a solution containing macro-and micro-nutrients (Table 7, http://www.ncsu.edu/phytotron/manual.pdf). The greenhouse had a 26°C/22°C (12h/12 h) day/night temperature cycle and natural lighting, expect the night was interrupted for 3 h to regulate flowering (from 11 pm -2 am, light was provided from incandescent bulbs at 11-12 µm s -1 m -2 ). Figure S1 shows the time-course of fiber development during July under these conditions, which was typical for April to October when bolls were harvested for these experiments. The slightly cool day temperature for cotton plant growth avoided any high temperature stress effects and slowed fiber development somewhat, which allowed relatively transient changes in transition-stage gene expression to be captured reliably.
Microarray analysis was performed with RNA extracted from greenhouse-grown G.

Determination of Fiber Growth Parameters
For fiber differentiating in the 26°C/22°C Phytotron greenhouse, length and weight were measured from 5 bolls of different plants collected on each DPA. For weight determination, fiber from 2 locules of each boll was separated from the counted seeds (except 30 DPA, for which only one locule was used), dried (48 h, 80°C), and weighed. For length measurements, fibers of at least 5 seeds per boll were relaxed by boiling in 0.025N HCl. Fiber still attached to the seed was straightened by a flow of water on the convex side of a watch glass, and the length of the bulk of the fiber mass was measured with a mm ruler. SE values were determined for each mean, but some are not visible when plotted in Figure S1. The sharp increase in the ratio of fiber weight/length beginning at 22 DPA (Fig. S1) showed this to be the time of onset of secondary wall thickening.

Microscopy
For cryo-FE-SEM, fibers that had been snap frozen in slushed liquid nitrogen were observed through a window cut in the boll wall using a cryo field emission SEM coupled with a cryogenic preparation unit. Additional details for cryo-FE-SEM are published on-line in Supplemental Protocol S1.
Prior to thin sectioning, cryo-fixation and freeze substitution were accomplished as previously described (Salnikov et al., 2003). For ultrastructural analysis of cell walls and CFML, a lab microwave was used to process samples from the steps of secondary fixation onwards.
TEM immunolabeling was performed similarly to published methods (Ruan et al., 2004).
Additional details for TEM structural analysis and immunolabeling prior to TEM or fluorescence analysis are published on-line in Supplemental Protocol S2.
For histology, developing ovules with attached fiber were snap-frozen, freeze-substituted, then embedded in paraffin prior to sectioning and staining by combination and modification of published methods (Grimson and Blanton, 2006;Ruzin, 1999). Additional details are published on-line in Supplemental Protocol S3.

Analysis of Cell Wall Cellulose and Sugars
Frozen fiber at 10, 17, 19, 21, and 30 DPA was pulverized under liquid nitrogen, lyophilized, and determined to be starch-free. Crystalline cellulose content was determined by a modification of the Updegraff assay (Updegraff, 1969). Sugars were analyzed as TMSmethylglycoside derivatives, which quantifies both neutral and acidic glycosyl residues in noncellulosic cell wall matrix polymers (York et al., 1985). Additional details are published on-line in Supplemental Protocol S4.
Relative composition for each sugar was expressed as nmol/mg starting material. One sample for each DPA contained lyophilized pooled fiber from 3 to 5 individual bolls from several plants. For insoluble sugars, means and standard deviations were based on two separate samples of pooled fiber, each assayed in triplicate so that 5 to 6 independent TMS procedures contributed to each mean (in only a few cases, outlier values due to experimental error were www.plantphysiol.org on August 26, 2017 -Published by Downloaded from Copyright © 2009 American Society of Plant Biologists. All rights reserved. discarded). For soluble sugars, means and standard deviations were based on 2 independent TMS procedures (one assay each of two separate pooled fiber samples). Significant differences (p < 0.05) between means from the two pooled fiber samples at 17 DPA vs. other DPA were assessed by t-test (one tailed test with unequal variance).
For each of three replicates of fiber at each DPA (100 mg fresh weight fiber, previously liquid nitrogen pulverized, then freeze-dried), the fiber was further ground in a mixer mill and sequentially incubated in the three solvents (250 µL solvent, shaking, 2 h) with retention of the supernatant and pellet at each extraction step. The supernatants from the three sample replicates were pooled and each pooled sample was printed in six replicates and three dilutions (undiluted and 1:4 and 1:24 v/v dilutions), giving a total of 18 spots representing each DPA for each solvent. Arrays were probed with mAbs or CBMs and developed similarly to a dot blot. Arrays were scanned and the spot intensities (signal mean minus local background median) were quantified using microarray analysis software (ImaGene 6.0, BioDiscovery, El Segundo, CA, USA). A heat map was generated to display the relative intensity of each signal to the maximum signal observed in the experiment, which occurred for JIM5 in the CDTA fraction at 30 DPA (brightest yellow; value of 100). Although CoMPP signal intensities reflect the quantity of a particular epitope within a sample, the technique is not fully quantitative because mAb and CBM probes bind with different avidities and epitopes are not necessary extracted from samples with equal efficiency.

Microarray Analysis
Based on the G.h.fbr-sw ESTs that were biased toward the early secondary wall stage of cotton fiber development (Haigler et al., 2005), a spotted cDNA microarray of 3,185 cDNAs was hybridized with Cy3/Cy5 probes prepared from RNA isolated from fiber at 6, 10, 20, and 24 DPA. In the microarray analysis, all time points were compared to the 20 DPA sample as a common reference (Smyth, 2005) because: (a) this time point was expected to show measurable expression given that the spotted cDNAs derived from 20 DPA RNA (Haigler et al. 2005); and (b) our major interest was to explore genes that were up-regulated for secondary wall deposition.
Two copies of each amplified cDNA were spotted on each slide and three biological replications and two technical replications (dye swaps) were performed for each data point, followed by quality control and statistical analysis of the results (R Development Core Team, 2005;Smyth, 2004Smyth, , 2005Smyth and Speed, 2003;Smyth et al., 2005). Briefly, RNA was labeled using the amino-allyl procedure (http://www.tigr.org/tdb/microarray/protocolsTIGR.shtml). Hybridization was performed using SlideHyb #1 buffer (Ambion, Austin, TX). Slides were scanned using an Affymetrix 428 Array Scanner (Affymetrix, Santa Clara, CA) and analyzed using the GenePix Pro 3.0 software (MDS Analytical Technologies, Toronto, Canada). Array normalization and statistical analysis were performed using the "limma: Linear Models for Microarray Data" library module (version 2.2.0) of the R statistical package (version 2.2.0). Slide intensity data were normalized using the global LOESS method. The least squares method was used for the linear model fit utilizing the Benjamini and Hochberg method to control the false discovery rate.
A summary file of the microarray data was created, "Cotton_Microarray_Summary_Annotated", including expression ratios (log2) and p-values (for averaged duplicate spots of each cDNA on each slide) for the following DPA comparisons: 6/20, 10/20, and 24/20. Note that following the convention of placing the value for the common reference in the denominator resulted in negative 6/20 and 10/20 ratios for genes that were upregulated at 20 DPA during the transition stage. To provide annotations, the cotton cDNAs on the array were compared by BLASTx (August 5, 2008)  information automatically retrieved by BLAST, the file was parsed to generate a brief, informative, annotation for inclusion in the summary file.

Quantitative Real-time Reverse-transcription PCR (qPCR)
Details of gene selection, primer design, RNA isolation, and qPCR procedures are published on-line in Supplemental Protocol S5. In qPCR, Gheif5 (eukaryotic translation initiation factor-5) was used as an endogenous normalizing gene; it had nearly equal expression in 6 to 30 DPA fiber (Haigler et al., 2009). The qPCR procedure was repeated three times, beginning with RNA isolation from independent samples at each DPA. Average gene expression patterns were calculated relative to a time of high rate elongation at 10 DPA, which showed clearly any potential up-regulation during the shift to secondary wall deposition at the transition stage (Table S5). Selected gene expression profiles were graphed as percent maximum increase for each gene in order to allow comparison of developmental patterns of expression regardless of possibly different expression amplitudes (non-normalized data are in Table S5).

Enzyme Activity Assays
Additional details of assay methods are published on-line in Supplemental Protocol S6.
Liquid-nitrogen-ground fiber samples from 10 to 30 DPA were assayed, with results normalized to total soluble protein (Bradford Assay; Thermo Fisher Scientific Inc.). The reported incubation time was determined to be in the linear range of each assay. Fiber extract was replaced with equal volumes of the extraction buffer in blanks. For each data point, an average for 4 biological replications was determined ± SE. β-galactosidase acivity was measured by hydrolysis of β-D-galactopyranoside (Pressey, 1983). Activity of XG endo-hydrolase (XEH) was determined as described with minor modification (Kallas et al., 2005). Pectin methylesterase (PME) activity in cotton fibers was quantified as described (Jiang et al., 2005).
Using published methods and polygalacturonic acid (P-3889, Sigma-Aldrich) as substrate, negative results for cotton fiber extracts were obtained for polygalacturonase activity (Zhang et al., 1999) and pectate lyase activity (Payasi et al., 2004). Similarly, negative results for cotton fiber extracts were obtained in gel diffusion assays (Taylor and Secor, 1988)

in which
Rhizopus polygalacturonase (P-2401, Sigma-Aldrich) yielded positive results.   Figure S4. Images of fiber packing around an ovule at high magnification, and fiber packing within a locule Supplemental Figure S5. Overviews of fibers immunolableled in Figure 3 Supplemental  During elongation, thin coherent walls (B-E) join adjacent fibers, including filling corners between three fibers (asterisk in B). (E) shows the distinction between the inner primary wall (arrowhead) and material between adjacent fibers (arrow) in a case where a missing fiber (from space at bottom center) likely separated from the forming bundle during specimen preparation.
Occasionally, the thin joined walls are interrupted by material-filled bulges (F,G arrows). In (G), the lower fiber is missing, and the bulge material is clearly distinct from the traditional primary cell wall (cw) of the upper fiber. (H) shows a cleared space (asterisk) between three fibers at the onset of secondary wall thickening after the interfacial material had broken down (see below).    Fluorescence micrographs of semi-thin sections of 10, 17, and 24 DPA cotton fiber crosssections immumolabeled with JIM5 or JIM7. Sections were optionally pre-treated with 100 mM Na 2 CO 3 as a de-esterifying agent prior to labeling. Successive sections of the same sample are shown for each DPA. The exposure time was equivalent and no image processing was performed for A to I, which allowed comparison of labeling intensity between samples. The boxed areas in A and D were superimposed to produce J (see below). At 10 and 17 DPA (D, E, G, and H), strong signals arose from HG distributed throughout the primary wall following labeling with JIM5 (after Na 2 CO 3 ) and JIM7. These images showed that the fibers were bundled into a simple tissue during elongation. In contrast, by 24 DPA CFML degradation had occurred to release individual fibers (C and F). Comparison of C to A and B shows that HG in the main primary wall was less esterified after elongation ended, a conclusion that was reinforced by the strong decrease in the JIM7 signal at 24 DPA in I compared to G and H. (J) The dim positive signal for JIM5 untreated (boxed area in A, signal changed to red) was superimposed onto the bright signal after Na 2 CO 3 pretreatment (boxed area in D, signal retained as green). The red signal aligned with the middle of joined primary walls between adjacent fibers in many locations and filled larger spaces between three fibers (arrows), indicating that fiber junctions were enriched in HG with a lower degree of esterification. The bar in I = 20 µM and applies to A to I. The bar in J = 10 µM. (K) Heat Map arising from CoMPP analysis of 6 to 30 DPA fiber extracted with CDTA, NaOH, and cadoxen. The 24 extracts (listed on the y-axis) were printed and exposed to 13 cell wall polymer probes (listed on the upper x-axis). Table S1 shows binding properties of the probes, and all intensity values are relative to 100. In the CDTA fraction, the inverse changes in the JIM5 (increasing) and JIM7 (decreasing) signals after 20 DPA correlates with fluorescence immunolabeling (compare C and I). In the cadoxen fraction, the CCRC-M1 and JIM5 epitopes (but not JIM7 epitopes) had elevated signals at 19 to 21 DPA, which correlated with the timing of release of these epitopes from the CFML as observed in TEM.   Table S2 and Table S3 for complete results for sugar analysis in the insoluble and soluble fractions, respectively. In (B-F), each graph shows percent of the maximum observed for that measurement type; non-normalized data are in Tables   S5 and S6. Transcriptional profiles in 6 to 30 DPA fiber are shown for genes related to: (B) developmental stages--GhRAC13 with peak expression in the transition and GhCTL1 with peak expression during secondary wall deposition; (C) XG modification (Ghxth30 and Ghxth6; XG endotransglycosylase/hydrolase-type); (D) pectin modification (Ghpe-E and Ghpae-c; pectin esterase-type) and galactosidase (Ghbgal12); and (E) pectin hydrolysis (Ghpg-A1 and Ghpg-A3; polygalacturonase-type). (F) Transition-stage peaks and/or up-regulation of enzymes activitiespectin methylesterase (PME), ß-galactosidase (ß-Gal), and XG hydrolysis (XEH).