For my dissertation research I  investigated whether wood-eating catfishes could digest wood with the aid of an endosymbiotic community of microbes and truly subsist on a wood diet.  In 2006 I traveled to the Río Marañon in northern Peru and collected five species of xylivorous (wood-eating) catfishes from their native habitat: Panaque nocturnus, P. panope, P. albomaculatus, P. gnomus, and Hypostomus pyrineusi (formerly of genus Cochliodon).  I compared the gut structure and function of these xylivores to that of a closely related detritivore, Pterygoplichthys disjunctivus.  Below are the important findings, which suggest that wood-eating catfishes do not rely on an endosymbiotic community in their digestive tracts to digest wood.  See my publications page and  my CV for publications on this, including the new one with Nathan Lujan in Functional Ecology.

First of all, neither Panaque cf. nigrolineatus nor Pterygoplichthys disjunctivus could assimilate significant amounts (<30%) of cellulose and hemicellulose from wood in the laboratory.  In fact, over a four-six week span, the fish lost 5-15% of their body mass, indicating that they were not thriving on a wood diet.  
Second, Panaque cf. nigrolineatus and Pterygoplichthys disjunctivus passed wood through their digestive tracts in less than four hours (see below).  Because these fishes have very long digestive tracts (11-18X their body length), that is extremely fast transit, too fast for microbial digestion of cellulose.  This is further corroborated by low short chain fatty acid (SCFA) concentrations (<10 mM for the entire digestive tract) in all species examined.  Thus, even though I have observed extremely negative redox potentials (-600 mV) in their intestinal fluids, which indicates a highly anaerobic environment, these fish are not relying on endosymbiotic fermentation to meet their daily energetic needs.  Gut contractility is also strong in these animals and continues after death.  See this video of a P. nocturnus gut contracting.  
Third, through histology (light microscopy) and transmission electron microscopy (TEM) I have not observed dense populations of microbes in any region of the intestine in any of the wood-eating or detrivorous catfishes.  If these fish were relying on microbes, we should be able to see them with TEM.  This is corroborated by low microbial DNA concentrations (<8 ng/uL) in the fishes' guts.
Fourth, the patterns of cellulase and xylanase activities in the catfishes digestive tracts suggest that they cannot efficiently digest wood cellulose and hemicellulose (consistent with what we saw in the lab feeding trials).  For example, cellulase activities are low (<0.03 U per gram) and tend to decrease in activity towards the distal intestine.  This type of pattern is indicative that cellulases and xylanases are ingested with wood-detritus rather than being synthesized by an endosymbiotic community residing in their guts.  Remember that these fishes are eating dead wood in their environment that is already being degraded by microbes.  Thus, there are cellulases on the food they are eating.  One last piece of evidence here is that amylase activities (i.e., the ability of the fish to digest algal and microbial polysaccharides) are five orders of magnitude (100,000x) higher than cellulase and xylanase activities.  These fishes are clearly geared for the assimilation of soluble polysacchrides, more so than structural ones.
Fifth, it appears that the fish efficiently digest disaccharides (e.g., alpha-glucosides, beta-glucosides, beta-mannosides) from woody detrtius.  Evidence for this is found in the disaccharidases from the fishes' intestinal walls having Michaelis-Menten constants (Km) an order of magnitude lower than the Km values of microbial disaccharidases ingested with the food.  Thus, as the fish consume wood-detritus that is already being degraded by microbes, the fish are efficient at assimilating disaccharides released by microbial degradation.  This makes sense in light of their whole digestive strategy: eat as much as possible, pass it through the digestive tract quickly, assimilate the soluble components that are easy to digest, and let any recalcitrant material (e.g., cellulose) pass through the gut undigested.  In fact, given the gut morphology of the fish (i.e., long, thin-walled intestines), this strategy is predicted theoretically (Sibly and Callow 1986; Horn and Messer 1992), and has been observed empirically in other herbivorous/detritivorous fishes (Crossman et al. 2005; and, see the Campostoma story above).

Stable isotopes indicate that the fish can assimilate the carbon from wood-cellulose in nature.  This is likely from the disaccharides released from microbial degradation of wood in nature (see above).  However, the fishes nitrogen isotopic signatures suggest that they are not getting their protein from wood-detritus alone.  The delta-15-N signatures of the fish (>7 ‰)  are too high to be coming solely from wood detritus (1.3 ‰) or endosymbiotic nitrogen fixation (which produces very low delta-15-N near zero).  See Lujan et al. (2011).  Thus, the fish likely get their N from microbial decomposers (e.g., fungi), or supplement their wood diet with a different nitrogen source of either amorphous detritus (which we find in their guts) or even some animal material. Therefore, all of the data I have collected suggest that these fishes do not harbor an endosymbiotic community in their GI tracts.  Although bacteria and fungi can be cultured from the intestines of Panaque and Pterygoplichthys (Nelson et al. 1999), these microbes are likely free-living and ingested with food.  This is further supported by the observation that Nelson and colleagues were only able to culture aerobic and facultatively anaerobic microbes from Panaque.  This is despite Panaque having anaerobic intestines (negative redox potentials in intestinal fluid).  In my opionion, culturing microbes with cellulolytic capabilities from an animal's digestive tract does not show that these bacteria are symbionts.  For example, grass carp, which eat aquatic macrophytes rich in cellulose, have cellulase activities in their guts (Das and Tripathy 1991) and an active microbial population (Trust et al. 1979; Lesel et al. 1986), yet poorly digest the cellulose component of their plant diet (Van Dyke and Sutton 1977). This is likely due to rapid gut transit and low levels of microbial fermentation in the grass carp guts (Stevens and Hume 1998), just like the wood-eating catfishes.  However, I do concede that my findings do not show that there are NO symbiotic microbes residing the wood-eating catfishes' guts.  There may be some microbes involved in vitamin and/or aminoacid synthesis, but this needs to be investigated.  I just found no evidence for any in situ cellulose digestion. In conclusion, wood-eating catfishes of the genera Panaque and Hypostomus are detritivores, just like their close relatives in the genus Pterygoplichthys.  They display none of the adaptations observed in other wood-eating taxa (e.g., expanded hindgut in lower-termites and beavers; endogenous cellulases in higher-termites) for the digestion of wood in their digestive tracts.  However, these fishes are lilely important mediators of carbon cycling in their environment (i.e., the Amazonian basin), and, in my opinion, all of the really cool research lies on the level of ecosystem processes and nutrient cycling. Hence, I now study decomposition in the environment.        

Pterygoplichthys disjunctivus captured from the Wekiva Spings complex, near Orlando, FL

                                   

            Panaque cf. nigrolineatus in a tank with wood it has been  eating for two years.  This fish is about 200 mm SL.