8.7.1 Work to date

The questions researchers have tried to answer over the last two decades using molecular mechanics and molecular dynamics have been, what are the structures of cellulose isotypes Ia, Ip, and II, and, can the uncertainties and irregularities in the experimental data be clarified? Many insights have been gained from simulations using multiple force fields, models, and programs. The structure of Ia has been robust in its behavior across several of the force fields as evidenced by the average structures after sufficient equilibration (37, 50-54, 56, 57, 71, 72). These simulations have been able to show the details of the tilting of sugar rings in alternating layers, the right-handed twist in the cellulose fiber, and the stable hydrogen bonding and changes in hydrogen bonding (53) over nanosecond time scales. Analysis of the water structure (57,73,74) and simulations of smaller bundles of cellodextrin chains suggest that the structure is more gel-like (72) in contrast to the simulations of larger bundles in which highly crystalline structure is found to be stable below the first layer. The study of the free energy surface of the alpha-(1-4)-glycosidic linkage including the effects of water (75) provides insight into the contribution of this linkage to more complex cellulosic structures using umbrella sampling to produce the surface.

More recently, MD tools have been used to study interactions between cellulose and other non-cellulosic molecules. The binding domains of cellulases have been docked onto the [100] surface of cellulose Ia resulting in observation of interesting and possibly important behavior in multi-nanosecond MD simulations (58). The entire CBH I enzyme was docked onto the cellulose surface and simulated for 1.5 ns. The complex was found to be stable and have significant water structuring around the interaction regions (76).

Synergism is usually only seen in the digestion of substrates that contain crystalline cellulose, probably because there are only a few regions in this substrate that are accessible to each cellulase. It seems likely that synergism occurs only when two cellulases attack different regions of the cellulose molecule and each one creates new sites of attack for the other enzymes in the mixture. There is no evidence that synergism requires interactions between the synergizing cellulases, as cellulases from unrelated organisms show similar synergism to those from the same organism. All good endocellulases are able to show synergism with any exocellulase, but most endocellulases do not show synergism with each other. Exocellulases show synergism with other exocellulases, but only if they attack different ends of the cellulose chains. Processive endocellulases can give synergy with all other types of cellulase (Wilson, D. B., unpublished). It is interesting that cellulose pretreated with an endocellulase is a better substrate for exocellulases than is untreated cellulose, but the reverse is not true (55). This is true, even though it has been shown in synergistic mixtures of an exocellulase and an endocellulase that endocellulase activity is increased as much in the mixture as is exocellulase activity (28). It seems likely that when an exocellulase degrades a cellulose chain, it disrupts some regions in adjacent cellulose molecules, making them more available to an endocellulase. However, over time the molecules in the disrupted regions may reform their interactions with adjacent chains, so that when an endocellulase is added, after exocellulase treatment, it is not able to utilize the transiently disrupted chains.

Pretreatment using lime has been studied as a low-cost process that primarily achieves acetyl and lignin solubilization (10,14,18,62,63). Lime pretreatment has been practiced at a wide range of temperatures, from 25°C to about 130°C, with lime loadings of about 10 wt% (on a dry feedstock basis) and solids loadings of 20% or less. At the higher temperatures, the pretreatment times are reasonably short (minutes to hours), but can extend to several weeks at lower temperatures. Because of the lengthy residence time at low temperatures, lime pretreatment can be conducted in a pile arrangement without expensive pressure reactors and can be performed as part of the feedstock storage system (18). Near-complete deacetylation generally occurs upon lime pretreatment of low-lignin herbaceous feedstocks and agricultural residues, with about 30% lignin removal. Much higher lignin removal (up to 80%) can be achieved by adding oxygen or air to the lime pretreatment system. In the lower temperature lime pretreatment pile arrangement, this can be accomplished by percolating air through the pretreatment pile. The additional lignin removal under oxidative conditions allows lime pretreatment to achieve reasonable enzymatic digestibility using more recalcitrant feedstocks, such as sugar cane bagasse and hardwoods (14). Lime pretreatment generally requires longer pretreatment times than other alkaline pretreatments, such as those using ammonia, but catalyst costs are lower. However, regeneration and reuse of the lime will probably still be necessary, and such recovery systems will add significant capital and operating costs to the lime pretreatment approach (21).

14.5.3 Solvent pretreatments

14.5.6.1 Organic solvents

Numerous organic or organic-aqueous solvent mixtures utilizing methanol, ethanol, ace­tone, ethylene glycol, triethylene glycol, and tetrahydrofurfuryl alcohol have been used as biomass pretreatment processes to solubilize lignin (10-13, 64, 65). Such processes are commonly referred to as organosolv processes. In some studies, inorganic acid catalysts, such as sulfuric or hydrochloric acid, are added to achieve significant levels ofhemicellulose hydrolysis and even cellulose hydrolysis (66) along with lignin solubilization. In some cases, the main components of biomass (cellulose, hemicellulose, and lignin) can be effectively fractionated, with each component potentially used for separate value-added products (67). Solvents must be effectively recovered and recycled using appropriate extraction and separa­tion techniques without leaving behind any inhibitory levels of residual solvents in process streams that undergo subsequent biological processing. While residual cellulose-rich pre­treated solids from such processes maybe highly digestible using cellulase enzymes, the cost of such processes and the potential value of the relatively pure fractions may make them better suited to higher-value applications.

The organization of walls can be described as consisting of cellulosic microfibrils and struc­tural proteins inserted into and reinforcing a multicomponent gel matrix, composed of stereo-irregular, non-cellulosic heteropolysaccharides (heteroglucans, heteromannans, het — eroxylans, (1^3,1^4)-p-D-glucans, and pectins). In primary walls these matrix polysac­charides may interact through physical associations with one another and the cellulosic microfibrils and also through covalent associations. In secondary walls the matrix polysac­charides also interact covalently with non-carbohydrate wall components such as lignins, suberin, cutin, cutan, and proteins. The following sections describe the nature of both non-covalent and covalent associations between wall polymers.

It is not known whether RG-II is synthesized on preexisting HG that is synthesized by GAUT1 or related GalATs, or whether a unique GalAT is responsible for synthesizing the HG backbone of RG-II. It is also not know if the side chains of RG-II are synthe­sized as individual oligosaccharides and transferred in bulk onto HG, or whether in­dividual glycosyltransferases transfer each distinct glycosyl residue individually onto the growing non-reducing end of the maturing RG-II molecule. The results of Egelund et al. (274) on the proposed RG-II:XylTs would be in agreement with the latter model (see below).

5.4.9.1 RG-II:apiosyltransferase (RG-II:ApiT)

As mentioned above it is possible that the apiosyltransferase activity identified during the studies of apiogalacturonan synthesis in Lemna is involved in RG-II synthesis. However, no gene for RG-II ApiT has been identified. Interestingly, when the gene reported to encode the UDP-apiose or UDP-D-apiose/UDP-D-xylose synthase was downregulated in Nicotiana benthamiana by virus-induced gene silencing of NbAXSl, the result was a reduction in the amount of RG-II in the walls (358). These results provide evidence that UDP-apiose is the substrate for incorporation of apiose into RG-II and that the UDP-D-apiose/UDP-D-xylose synthase gene encodes the enzyme that synthesizes the required substrate for RG-II synthesis.

Plant cell wall polysaccharides contain L-fucose (6-deoxy-L-Gal) derived from the sugar — donor, GDP-Fuc. GDP-Fuc synthesis occurs in two enzymatic steps (similar to the synthesis of UDP-Rha from UDP-Glc). These enzymes have been characterized and the corresponding functional genes have been identified in humans, plants, and bacteria. First, GDP-Man 4,6-dehydratase (GMD), converts GDP-Man to a GDP-4keto-6-deoxyMan intermediate. The latter is then converted by GDP-4-keto-6-deoxymannose-3,5-epimerase-4-reductase (GER1, FX) to GDP-Fuc. In Arabidopsis, there are two GMD isoforms (GMD1, At5g66280; GMD2 (mur1) At3g51160) which share 92% aa sequence identity to each other; and two GER isoforms (GER1, At1g73250; GER2, At1g17890) that share high (88%) sequence identity to each other. In some tissues it appeared that GMD isoforms are co-expressed, but in other tissues expression is restricted. For example, GMD2 is expressed in most cell types of the root, but not in the root tip where strong expression of GMD1 is observed (490). Within shoot organs, GMD2 appears to be expressed in most tissues while GMD1 expression is restricted to stipules and pollen grains. The lack of GMD2 above ground (murl mutant) corresponds to an almost complete reduction in Fuc in wall polysaccharides including XG whose Fuc can be substituted by L-Gal presumably as a result of increased GDP-L-Gal availability (491, 492). However, below ground the murl mutation leads to a 40% reduction of Fuc. Some isoforms may have redundant function in a specific cell, but in other tissues of the same plant the existence of isoforms may provide pools of the NDP-sugars to synthesis of different types of glycans.

In spite of their intractable nature, lignins can be fully removed, but only under quite drastic chemical conditions using either strong acid [e. g., sulphite pulping processes (147-150)] or alkaline [e. g., kraft (8)] conditions at elevated temperatures and pressures. They can also be partially removed using milder procedures, such as by extensive ball milling for 4-5 days followed by extraction in dioxane:H2O (9:1) (71). Indeed, the “milled wood lignin” or “Bjorkman” lignin (151,152) preparations are often considered as representing the mildest forms of treatment necessary to solubilize lignin-derived components with these frequently being described as structurally closest to native lignins. The overall yields resulting from such ball-milling manipulations are though generally low (<20%), but can be somewhat improved by pretreatment with cellulase(s) and other hemicellulose degrading enzymes. There is also one report of attempting to solubilize entire plant tissue using a mixture of DMSO/tetrabutylammonium fluoride (TBAF) on very finely ground plant cell wall residue (CWR) (153). In our hands, this approach was unsuccessful, in large part because total solubilization was not achieved, with the suspension also becoming very black, viscous, gelatinous, and unworkable.

In terms ofthe acidic sulphite treatment ofwoody plant tissue, the effects on delignification have been somewhat studied. For example, using a continuous flow (pulping) apparatus, it was possible to essentially delignify the tissue completely when treating black spruce (Picea mariana) stem wood. Characterization of the components in the resulting eluant established that initially various paucidisperse low molecular weight moieties, such as mono-, di-, and tri-sulphonated monomers, e. g., 38-41 (Figure 7.10), were released and these were facilely identified (147-150). These were then followed by solubilization of higher molecular weight sulphonated polydisperse lignin chains of increasing molecular weights ranging from 3.3 to 120 kDa (147); however, these polymeric entities have not yet been fully characterized in terms of their chemical structures thus far.

Alkaline (kraft) pulping delignification has also long been investigated, together with the nature of the accompanying chemical delignification reactions. Perhaps most interesting, Sarkanen etal. (8) demonstrated that solubilized “kraft” lignins, assumed to contain reflec­tions of the original lignin primary structures, displayed a capability to self-associate. This was interpreted then, and now, as due to extensive noncovalent dynamic electronic corre­lations between the associating (primary) chains resulting in aggregation of the polymeric lignin chains (e. g., up to several millions in molecular weight). This property is yet an­other complicating and potentially confounding feature in lignin analyses; such properties limit further the ability to facilely study lignin from a structural perspective (e. g., by NMR spectroscopy) as discussed below.

The fields of lignin chemistry, biochemistry, molecular biology, and plant cell wall biomechanics/anatomy are of greatly renewed interest as the twenty-first century unfolds. This contrasts with the preliminary studies of the mid-twentieth century on lignin consti­tution, which more or less relegated this important class ofmacromolecules to the scientific curiosity of being randomly assembled. This, in turn, was largely as a result of the severe technological barriers that existed during the 1950s and 1960s, and the speculations of the time. Yet, even today, there remains spirited discussion as regards both lignin structure and how they are formed.

From the standpoint of the authors, the recent advances made largely in the last decade or so now beckon a different viewpoint of the lignins — particularly as regards their potential to help facilitate either cellulose-to-ethanol or related biofuels/bioproducts technologies. This is thus a very exciting time for studying lignins, as there is still much to be learned about both lignins and the associated cell wall assembly processes. In addition, such problems need to be solved to meet humanity’s needs for future generations.

Several areas that seem to be of particular promise for future study are summarized below:

1 Monolignol Transport to the Cell Wall: Establishing the biochemical basis of how monolig — nols 1,3, and 5 (or derivatives/homologues thereof) are targeted to precise regions within the cell walls, and how this is regulated, are important questions to resolve; while (ABC) transporters (351) have been postulated as involved, this has not yet been proven.

2 Lignin Initiation Sites: Lignin deposition begins at so-called “initiation sites” in the cell corners and then progresses to eventually encapsulate the entire cell wall prior to apoptosis (see Figure 7.5). In this regard, establishing the nature and function of each of the proteins, enzymes and genes involved at these initiation sites appears to be a particularly attractive line of enquiry. Resolution to this should substantially clarify any remaining questions on how lignins are formed.

3 Lignin Primary Sequences/New Chemistries: A related need is to develop methodologies to establish the primary sequences of the lignins being formed initially, as well as developing new technological means and new chemistries to both identify and quantify all interunit linkages. This will also include establishing unambiguously whether lignins contain either no branches, or short or long branches, and how lignin-carbohydrate linkages are both formed and regulated.

4 Re-oxidation of the Growing Lignin Chains: Another urgently needed emphasis is in es­tablishing biochemically how monolignols 1, 3, and 5 and the growing lignin chains are re-oxidized in the cell wall following initial coupling, i. e., thereby enabling radical/radical generation to reoccur. This is a problem that has long been recognized, namely, how does one-electron re-oxidation of lignin continue at sites presumably distant from the one-electron oxidative enzymes. Whether this occurs, for example, via some form of elec­tron transfer through the lignin matrix (e. g., originating via oxidation by the presumed peroxidase/H2O2) or whether a diffusible oxidant, such as has been suggested for Mn3+ is involved (352), needs to be established.

5 Monolignol Radical and Lignin Primary Chain Interactions, Proposed Template Polymeriza­tion, and Lignin Association: Unambiguously determining whether the monolignol radi­cals being generated are never “free,” but instead are transiently immobilized as a result of strong л… л interactions between the substrates and a pre-existing lignin macromolec­ular template is also another important goal, i. e., following on the previously proposed template polymerization for replication of lignin primary chains (353). Recently, Sarkanen and Chen modeled non-covalent interactions between a monolignol (coniferyl alcohol, 3) radical and a representative monomer residue (= veratryl alcohol) for a lignin macro­molecule. Mo5-2x/6J1 + G (d, p) density functional theory calculations led to a gas phase stabilization energy of 13.4 kcal/mole for a cofacial complex with one strong and one weak intermolecular bond, versus that of 8.6 kcal/mole due to dynamic electron cor­relations in the interacting л-constituents alone. These researchers also suggested that head-to-tail orientations of the interacting species were preferred, and additionally pro­posed that an antiparallel double-stranded lignin template was responsible for obtaining macromolecular lignin domains lacking either crystallinity or optical activity. According to these researchers, the intermolecular interactions between monolignol radicals and the substructures in lignins may be considerably stronger than Watson-Crick A-T or G-C nucleobase pairs in DNA: in the latter, a weak hydrogen bond in an oligonucleotide has been estimated to only amount to an increment of ~0.4-2.0 kcal/mole to the stabilization energy. While the gas-phase calculations favor the concept of template polymerization, the potential ramifications of this clearly need to be more fully investigated. It will thus be most instructive to establish how many lignin-forming templates there possibly are, and to what extent each would be able to display limited substrate degeneracy. In addition to the question of proposed template replication, another important related emphasis is to unambiguously define the precise molecular basis for the strong associative forces observed between adjacent lignin chains/lignin preparations.

6 Transcriptional Control of Individual Cell Wall Formation Processes, Biomechanics, and Biodegradation of Plant Cell Walls: With the recent demonstration of specific transcrip­tion factors required for fiber formation, this now offers the opportunity to identify how the various cell wall types are differentially generated. This, in turn, may also permit investigation of lignification in specific cell types, and thus as to how they are individu — ally/differentially formed. In any event, this gives another new direction to the possibility of systematically modulating overall plant structure and plant properties for humanity’s use — one cell type at a time.

The areas of biomechanics, biodegradation, and factors affecting disease resistance are other most important emphases that need to be enhanced in terms of current levels of sci­entific investigation, i. e., in order to understand fully the potential of lignin modification. For example, to what extent can lignin contents and composition be manipulated without adversely affecting cell wall properties/vascular integrity for growth/development, harvest­ing, storage, and further processing? Furthermore, what effects do such manipulations have on biodegradation feasibility (e. g., with different lignin compositions/contents), as well as on plant defense.

In short, there is much still to do in determining how Nature’s second most abundant vascular plant biopolymers are formed and their potential (through manipulation) for

humanity’s varied needs. The next 5-10 years should be exciting ones for both lignin and plant cell wall research, and promise to be an exciting challenge for the twenty-first century.

Acknowledgments

The authors thank the National Science Foundation (MCB-0417291), the United States Department of Energy (DE FG03-97ER20259), the National Institute of General Medical Sciences (5 R01 GM066173-02), McIntire Stennis, and the G. Thomas and Anita Hargrove Center for Plant Genomic Research for generous financial support.

As a model system for hydrolysis and dehydration of xylo-oligomers and xylan, static quan­tum mechanical calculations were conducted on the decomposition of xylobiose. As with xylose, acid-catalyzed reactions were studied by computing the energy barriers for proto — nated xylobiose in vacuum. Because of the large size of this molecule, highly accurate CBS calculations could not be conducted. Instead DFT [B3LYP/6-31(d, p)] calculations were con­ducted. This level of theory could underestimate (33-37) reaction barriers by 5 kcal mol-1, but is sufficient for a semi-quantitative comparison of the barriers for different reaction pathways. Initial molecular structures of neutral xylobiose, the protonated reactants, and the transition states used the corresponding structures from the study of xylose (17, 18) discussed above. Likewise, analogous reaction mechanisms were studied for xylobiose. For example, Reactions (9.5) and (9.6) show the first steps in the decomposition of xylobiose protonated O2 and O3 on the non-reducing end to form a furanyl ring or the precursor to formic acid. Reaction (9.5) exhibits a xylobiose dehydration mechanism similar to a monomer xylose (Figure 9.1). The degradation of the xylobiose is initiated when O2 on the non-reducing end is protonated. The protonated hydroxyl group (i. e., H2O) leaves the sugar ring forming a carbocation, which reorganizes to form an uncharged five-member ring leaving the positive charge outside the ring structure. A water molecule will then likely hydrolyze the ether linkage to break the dimer into a furan ring and an intact xylose molecule. The further dehydration of the furan ring structure leads to the formation of furfural. Likewise, the reaction could also initiate by protonation of O3 on the non­reducing end of xylobiose in a similar mechanism to the xylose monomer. However, here the
product with the open structure will dissociate into a xylose ester and a four-carbon cation

Protonation at O2 on the reducing end of xylobiose results in the reaction shown in (9.7), while protonation at O3 on the reducing end produces an unstable structure, which without a barrier, transfers its proton to O5 on the non-reducing end. The product from Reaction

(9.7)

might then dehydrate through reactions similar to those shown in Figure 9.1 to form a furfural molecule and a xylose molecule.

(9.7)

Proton addition to O2

In addition to the dehydration reaction shown in Reactions (9.5)-(9.7), the hydrolysis reaction to form two xylose molecules was considered. For this reaction, a proton is added to the ether linkage, which decomposes to a xylose molecule and an oxonium ion as shown in Reaction (9.8). In aqueous solution, the oxonium cation willbe quicklyhydrolyzed to form xylose as was discussed earlier. A comparison of the barrier for this process to the barriers for dehydration reactions shown, (9.5)-(9.7), was used to determine which reaction pathways are most likely. Importantly, no reaction barrier could be found for the hydrolysis reaction

(9.7) . If a proton was added to the ether linkage, the molecule decomposed without a barrier into xylose and the oxonium ion shown in Reaction (9.8).These calculationswere conducted
several times and always led to the same result. On the other hand, the reaction barriers for the dehydration reactions were found to be significantly larger and were consistent with the barriers for the dehydration reactions of xylose monomer. Using B3LYP/6-311G(d, p), barriers of 17.8 and 19.8 kcal mol-1 were obtained for Reactions (9.5) and (9.6). These are close to the barriers calculated for similar reactions for neat xylose (Table 9.1). Figures 9.5 and

9.2 compare the calculated molecular geometries for the transition states for dehydration of xylose and xylobiose protonated at O2 and O3. Notice that the bond lengths at the reacting centers are similar for xylose and xylobiose. The barrier calculated for Reaction

97

(9.7) is lower (4.7 kcal mol 1), but is still significantly higher than the barrier for Reaction

(9.8)

.

The absence of a barrier for Reaction (9.8) suggests that this process is kinetically favored over the dehydration reactions (9.5)-(9.7). This is not surprising, because, as mentioned above, Reaction (9.8) results in the formation of a relatively stable oxonium. Figure 9.7 shows a comparison of calculated reaction barriers for xylobiose protonated at the O2 and O3 sites and protonated on the linker oxygen atom. These calculations were conducted in vacuum, and as we have shown in our calculations of xylose, the barrier in aqueous solution is likely to be higher due to the endothermicity of the proton transferring from the solvent water molecules to the sugar oxygen atoms. However, since ethers have higher proton affinities than alcohols (44), the increase in the barrier due to solvation should be less for hydrolysis reactions (9.8), than for dehydration reactions, (9.5)-(9.7). Our calcu­lations suggest that the barrier for hydrolysis of xylobiose should be significantly lower than the barriers for dehydration reactions and that the kinetics of hydrolysis, consequently, should dominate. Because the ether linkage is identical in other xylo-oligomers, this re­sult further suggests that for all xylo-oligomers, hydrolysis should dominate. Loss of xy­lose due to dehydration reactions should only result from the dehydration of xylose itself, not from dehydration reactions of the xylo-oligomers when there is only acid in the so­lution. The accelerate destruction of xylose and xylotriose molecules with the addition of

inorganic salts (9) is likely due to a catalytic function, which lowers the barrier of dehydration reactions.

Rumen protozoa are reported to account for up to 50% of the microbial biomass in the rumen, and it has long been suspected that certain species play a significant role in the breakdown of plant material. There has been uncertainty, however, whether protozoa pro­duce their own cellulases and xylanases, or whether the activities that can be measured for the protozoal fraction of rumen contents are due to ingested bacteria. Rumen protozoa cannot currently be maintained pure culture. To resolve this issue, cDNA libraries have been constructed from polyA+ RNA (presumed to be of predominantly eukaroytic origin) from a protozoal-enriched fraction of rumen contents. In order to focus on individual proto­zoal species, material for extraction was initially obtained from sheep that carry a complete bacterial flora, but that had been de-faunated and mono-associated with single strains of protozoa. Screening for activity allowed the recovery of multiple cDNA clones that express CMCase or xylanase activity from Polyplastron multivesiculatum (76-78). Consistent with their protozoal origin, these genes show extremely biased codon usage, and AT-rich flanking regions, and fail to cross-hybridize with DNA of bacterial origin. Representatives of family 11 and 10 xylanases from rumen protozoa are generally reported that have rather simple domain structures, although the family 10 domain in the enzyme XynB is adjacent to a family 22 substrate-binding module (78).

Polysaccharidases in rumen protozoa are assumed to be expressed mainly within the food vacuoles, rather than being extra-cellular as in rumen bacteria and fungi. What con­sequences this has for their organization remains unknown, but it perhaps makes it more likely that they exist as soluble enzymes rather than as a cell-bound complex. Expression of

Neocallimastix patriciarum XynA (1) Neocallimastix patriciarum XynA (2)

Aspergillus niger XynA _ T richoderma reesei Xyn1

Aspergillus niger XynB
— Bacillus subtilis XynA

Streptomyces lividans XynA Streptomyces lividans XynB T richoderma reesei Xyn2

Bacillus pumilus XynA

Ruminococcus flavefaciens XynA Ruminococcus flavefaciens XynB Ruminococcus flavefaciens XynD Fibrobacter succinogenes XynC (1)

Fibrobacter succinogenes XynC (2)

Figure 12.2 Phylogenetic relationship between a xylanase from the rumen protozoan Polyplastron mul­tivesiculatum and other representative GH family 11 catalytic domains (based on Devillard et al. (77)). Sequences from rumen bacteria (Fibrobacter, Ruminococcus) and fungi (Neocallimastix) are shown in bold;those from non-rumen microorganisms are shown in normal type.

recombinant protozoal genes in E. coli maybe reduced because of their highly biased codon usage.

The sequence relationships of protozoal glycoside hydrolases show that they are often quite closely related to bacterial enzymes (77, 79) (Figure 12.2). They are, therefore, strong candidates for acquisition by horizontal gene transfer. The dense, mixed communities of the gut create obvious potential for such transfer events, particularly in the case of protozoa that are constantly engulfing and digesting rumen bacteria.