Assessment of the Gastrointestinal Fate of Bacterial Cellulose and Its Toxicological Effects After Repeated-dose Oral Administration

Background Bacterial cellulose (BC) is a nanobrillar polysaccharide produced by certain acetic acid bacteria. BC may be used in food, pharma and many other applications. However, detailed studies of the oral toxicology of BC are limited. Controversial data is published regarding this topic, specially when it comes to answering the question on whether cellulose is absorbed at the intestine. Following the European guidelines (EFSA), this work presents the results of a 21-day repeated dose oral toxicity of BC in male and female Wistar Han rats (Wistar rats). In parallel, microcrystalline cellulose Avicel LM310 (commercially available as a food additive) was used. Wistar rats were subjected to daily oral gavage of 0.75 mL of an aqueous suspension 1% (m/v) BC or of its counterpart of plant origin, Avicel LM310. Rats not submitted to gavage were included in the experiment as controls. Clinical observations, such as body weight measurements, food consumption and ophthalmologic evaluations were performed during the assay. After occision, serum chemistry, necropsic examination and histopathological analyses of the liver, kidneys, spleen and small and large intestines were performed. The presence of BC bers along the gastrointestinal tract was assessed histologically using a Green Fluorescence Protein coupled to a Cellulose Carbohydrate Binding Module (GFP-CBM) from Clostridium cellulolyticum. No adverse clinical observations related to BC administration were noticed, nor appreciable differences in the toxicological endpoints evaluated were detected. No evidence of BC persorption was found. Particularly, no BC was detected in the Peyer´s patches or in the mesenteric lymphatic nodules. Moreover, the histopathological analyses revealed that the global architecture and morphology of the organs and tissues was preserved, among the different experimental groups, with no signicant pathological changes among them. Regarding serum biochemistry, no signicant differences were recorded, for both sexes. These results 10% increase over controls in reactive oxygen species (ROS) with 1.5% w/w CNC, no signicant changes were observed. In vivo toxicity was evaluated in rats by “per os” administration, twice weekly, for ve weeks, with 1% w/w suspensions of NCF. No signicant differences in hematology, serum markers or histology were observed between controls and rats given CNF suspensions. These ndings suggest that the ingested NC has little acute toxicity and is likely non-hazardous when ingested in small quantities. In another study, Andrade et al. [19] assessed the effects of NCF from the Heartof-peach palm in a male rat model (Rattus norvegicus albinus). The addition of NCF to the diet resulted in a trend for an increased weight of the animals. The blood sugar, triglycerides and cholesterol concentrations did not show any signicant statistical difference over the whole duration of the experiment, indicating the maintenance of homeostasis of the animals. Moreover, Ong et al. [20] assessed the 90-day dietary toxicity, on Sprague Dawley rats, of brillated cellulose produced through mechanical homogenization of a wood pulp, as compared to Solka Floc, a conventional cellulose commonly used as a dietary ber in animal feed and human dietary studies. Survival rate, clinical observations, body weight, food consumption, ophthalmologic evaluations, hematology, serum chemistry, urinalysis, post-mortem necroscopic evaluation with histopathology were performed. No adverse reactions were noted in relation to the administration of brillated cellulose. The no-observed-adverse-effect level (NOAEL) for brillated cellulose was 2194.2 mg/kg/day (males)and 2666.6 mg/kg/day (females), corresponding to the highest dose tested (4 %). These results demonstrate that the brillated cellulose behaves similarly to conventional cellulose and raises no safety concerns when used as a food ingredient at these concentrations. Another study by Chen et al., [21] aimed at determining the effects of subchronic oral NCF consumption on various health aspects of Western diets-fed mice. This work demonstrated that subchronic impact (p>0.05) in blood biochemistry of Wistar rats. Observation of histological samples of the different regions of the rats’ intestinal tract and mesenteric lymph nodes did not show any local effects from the consumption of BC. Also, BC was not detected in the Peyer´s and colonica patches and the mesenteric lymphatic nodules. Data gathered in this study strongly indicates that the intestinal absorption of bacterial cellulose is particularly very improbable to occur and thus, its consumption is implausible to cause any toxicological effects.

Bacterial cellulose (BC) is an exopolysaccharide synthetized by Komagataebacter xylinus (formerly known as Gluconacetobacter xylinus), a strict aerobe Gram -negative bacteria. Although chemically identical to plant cellulose, BC presents some speci c features. It can be obtained with higher purity, exhibiting higher degree of polymerization (800-11000) and crystallinity index. Under static culture, BC is a gelatinous lm consisting of a 3D nano brillar network of pure cellulosic bers. These randomly assembled ribbon-shaped brils are less than 100 nm wide, being composed of elementary nano brils aggregated in bundles with lateral sizes of 7-8 nm. These brils have several micrometers in length [22][23][24][25][26][27][28][29]. Due to its high aspect ratio and surface area, BC has superior water retention properties. While vegetable cellulose presents water retention values around 60%, BC holds a water retention capacity up to 1000% of its dry mass. The unique structural, physicochemical, mechanical and biological features of BC have caught the attention as a versatile material for application in several areas, such as the biomedical, food, cosmetics, among others [30][31][32][33][34]. Particularly, it has been recognized as a gelling agent, thickener and suspension stabilizer with great potential for the food industry. Additionally, its potential role as an alternative source of dietary ber has also been suggested [34].
Currently, BC is produced and marketed mainly in Asian countries under the trade name 'nata de coco' [35,36]. The long history of BC consumption, without any reported cases of toxicity, suggests that BC is safe for consumption. Furthermore, some in vitro and in vivo studies assessing BC toxicity have demonstrated that it is not genotoxic, carcinogenic, pyrogenic or toxic for development or reproduction [37,16,5]. Moreover, BC does not induce acute or chronic oral toxicity or allergenicity [37,16,5]. In fact, the available data corroborate the general assumption that BC is non-toxic by ingestion or contact and does not trigger any in ammatory response or oxidative stress at the cellular level [37,16,5]. Some studies addressed the toxicology of BC. Hagiwara et al.
[16] evaluated the effect of BC sub-acute administration to F344 rats for 28 days, at dietary levels of 0, 1.25, 2.5, and 5.0%. The treatment had no adverse effects on mortality, body weight, food and water consumption, urinalysis, ophthalmology, hematology, blood biochemistry and histopathology. At necropsy, slight increased absolute and relative cecum weights, evident in females ingesting 2.5% and 5.0% dietary levels, were considered to be a physiological adaptation to the poorly absorbed fermentation-derived cellulose. The non-observed-adverse-effect level (NOAEL) was 5,331 and 5,230 mg/kg body weights/day, for males and females, respectively. Schmitt et al. [5] tested the acute oral toxicity of a commercial product named Cellulon ® bre (dried BC:sucrose at a mass ratio of 1:1) in Sprague-Dawley rats. Rats were fed a single oral dosage of 2000 mg Cellulon/Kg of body weight, via oral intubation, with no adverse treatment-related effects recorded. The sub-chronic toxicity of BC was investigated in Sprague-Dawley rats fed diets containing 0, 5, and 10% Cellulon ® bre or microcrystalline cellulose, for 13 weeks. There were no treatment-related deaths or clinical signs of toxicity. Body weight was unaffected.
Food consumption was generally increased in treated animals (both Cellulon ® bre and microcrystalline cellulose), assigned to the relatively high-test article concentrations used. Statistically signi cant changes in haematology and clinical chemistry parameters were not considered signi cant. There were no notable gross pathologic or histopathological ndings at necropsy and organ weights were unaffected. The outcomes were interpreted as proof that Cellulon ® bre and microcrystalline cellulose are toxicologically equivalent and that sub-chronic exposure to Cellulon TM bre did not adversely affect these animals. The NOAEL was 7,000 and 8,500 mg/kg/day for male and female rats, respectively.
These studies did not assess whether celluloses are able to cross the intestinal epithelium. Several publications have reported the uptake of a variety of particles, including nano-sized ones, across the gastrointestinal tract in animal models [38][39][40][41][42]. With regards to plant cellulose (micro)particles, contradictory results on the hypothetical persorption phenomena have been reported over the last years [1][2][3][43][44][45]. Reports have shown that 14 C labelled plant cellulose [43] and BC [4] are slowly and limitedly degraded by the micro ora of the large intestine of rats, resulting in metabolites that are partially absorbed by the colon. The absorbed cellulose degradation products were detected in the urine and on exhaled CO 2 . ADME (Absorption, Distribution, Metabolism, and Excretion) studies performed also with germ-free rats (without intestinal micro ora) have shown total excretion of either plant or BC in the feces [4,43], con rming that there is no absorption of non-degraded cellulose, but rather of its degradation products.
The present study was designed to assess the potential absorption and accumulation of BC at intestinal tract level, as well as any pathological changes in organs and tissues related with its potential toxicity, in male and female Wistar Han rats, following daily oral administration of BC for 21 days. The systemic toxicity of BC was also evaluated by measuring the serum standard parameters. After necropsy, to trace the fate of the administered BC, a Carbohydrate-Binding Module (CBM3A) from Clostridium cellulolyticum coupled to a green uorescent protein (GFP), was used for the histological visualization of cellulose, given its speci city and high a nity for cellulose, providing high detection sensitivity [6, 46-47].

Clinical observations and gross examination
All rats survived until the end of the experimental period, showing no abnormal clinical signs nor gross pathological changes at necropsy that could be ascribed to BC or plant cellulose administration. The feces of all animals had normal consistency throughout the study. Compared to the control group (without gavage), at any given day, rats subjected to either BC or plant cellulose oral gavages showed no signi cant differences (p>0.05) in their measured body weight (Table 1). Also, when compared to the control, the daily oral gavages of both cellulose types had no effect (p>0.05) on organ weight (expressed per animal bodyweight x 100) for either sex. Table 1 Mean (± SD) bodyweight evolution and organ weight at necropsy of Wistar rats. The animals were subjected to daily gavage of bacterial or plant cellulose suspension (0.75 mL, 1% m/v), during the 21 days of the experiment. The control group did not receive gavage. During the assay, food and water were provided to all animals ad libitum.

Assessment Of The Systemic Toxicity
To assess the potential in vivo systemic toxicity of BC, blood and serum markers were analysed ( Table 2). As compared to the control group, either plant or BC had no impact (p>0.05) in blood glucose levels. The results of serum markers of liver function Total bilirubin (TBIL), alanine aminotransferase (ALT), aspartate transaminase (AST), and alkaline phosphatase (ALP) indicated no signi cant differences (p>0.05) between bacterial and plant cellulose groups and the control group. The same was observed for the serum urea, creatine and electrolyte concentrations among all studied groups were noticed. The effect of BC on lipid metabolism was assessed by measuring the concentration of cholesterol (CHOL), with no signi cant differences (p>0.05) being observed among all experimental groups. Alanine aminotransferase (ALT), albumin (ALB), alkaline phosphatase (ALP), aspartate aminotransferase (AST), total bilirubin (TBIL), blood urea nitrogen (BU (GLU), osmolality, potassium (K), sodium (Na), total cholesterol (CHOL), total protein (TP)

Histopathological Analyses
Histological sections of the collected organs (liver, kidney and spleen) and the different regions of the intestinal tract of the animals were stained with H&E and compared by an experienced pathologist, who was blind for the experimental groups. The comparative examination evidenced a global preservation of the organ architecture in all cases and the absence of signi cant pathological changes ( Fig. 1). Regarding the intestinal sections, an apparent global maintenance of the organ normal architecture was observed with the absence of signi cant pathological changes, including in ammatory related alterations (Fig. 2). Also, no differences in histopathological analyses were detected between females and males from each experimental group.

Cellulose Tracking
Histological samples of the intestinal tract of all animals were stained with a CBM-GFP (green), to identify cellulose. Moreover, for actin-cytoskeleton and nucleus visualization the samples were also stained with Phalloidin-TRITC (red) and with DAPI (blue), respectively. In this experiment, two techniques were applied for the intestinal tissue preparation before freezing and cryosectioning: i) the swiss roll technique (SRT), which enables the observation of longitudinal sections of larger portions of the washed intestines, delivering more representative tissue sections for cellulose tracking as compared to the ii) gut bundling technique (GBT), which provides only transversal sections at speci c regions of the rats' intestinal tract.
We rst observed the distribution of cellulose in unwashed intestines using the GBT, also as a proof of concept of the cellulose staining method (representative images on Fig. 3). No cellulose was detected in intestinal associated lymphoid tissues -peyer patches and colonic patches (white arrows).
Cellulose was observed only on the luminal space of intestinal sections (Fig. 3).
To better assess whether cellulose is able to cross the intestinal epithelium, the intestines were washed to remove, as much as possible, the cellulose present in the lumen and processed with GBT and SRT (representative images on Fig. 4 and 5, respectively). Negligible amounts of cellulose were detected on GBT washed intestinal samples; when detected cellulose was generally found in the luminal space or trapped between the villi or in Lieberkhun crypts. No cellulose was detected in intestinal associated lymphoid tissues -peyer patches and colonic patches (white arrows) (Fig. 4). Negligible amounts of cellulose were detected on SRT washed intestinal samples without any distinct pattern. No cellulose was detected in intestinal associated lymphoid tissues -Peyer patches (highlighted by white arrows) and colonic patches (Fig. 5).
To demonstrate in more detail that cellulose did not penetrate the mucus layer of the intestinal tract, confocal microscopy images at higher magni cation and at different z-planes (depth) were collected (Fig. 6). Cellulose bers were detected on SRT washed ileum and colon sections trapped between the villi and in Lieberkhun crypts. No cellulose co-localizing with the intestinal absorptive epithelium or with lamina propria and submucosa was detected as can be observed in the different z-slices of the intestinal stained samples. In some cases, few discrete fragments of cellulose were found overlapping intestinal tissue.
Regarding the cellulose staining on serial histological samples of the mesenteric lymph nodes (MLN) from all animals, some discrete green dots were detected in all experimental groups (Fig. 7). This was observed with similar frequency in the positive and negative staining controls (omitting the incubation with CBM-GFP conjugate), both showing a similar green diffusive signal.

Discussion
General clinical observations of the animals over the experiment, together with gross examination and further histopathological analysis of the collected organs (liver, kidneys, spleen and the different regions intestinal tract), demonstrate that cellulose did not exert any local pathological effect on the examined tissues, which have preserved their global structure and morphology without any detectable in ammatory in ltrates or other pathologically relevant changes ( Table 1, Fig. 1 and 2).
The potential systemic toxicity of bacterial cellulose was assessed by serum markers analysis, taking plant cellulose and the control group (without gavage) as reference ( Table 2). No signi cant differences (p>0.05) were observed for either plant or BC. Total bilirubin, alanine aminotransferase, aspartate transaminase, and alkaline phosphatase are enzymes generally used to assess the liver function, with elevation in serum enzyme levels and TBIL being associated with liver damage [50]. Thus, apparently the different cellulose types have not compromised liver functions. The same conclusions stem from the analysis of the total protein (TP) values, which provide an estimation of the animals' nutritional status and assist on diagnosing several pathological conditions associated to liver and kidney disease [51]. A reduction in total protein (albumin and globulin), not observed in this work, is an indicator of an impaired hepatocellular function. The measurement of TBIL, serum urea and creatinine provides information on the likelihood of renal ailment or dysfunction [52]. As compared to the control groups no signi cant changes (p>0.05) in urea and creatinine levels for both cellulose types, were detected, suggesting that BC did not impact the normal renal function. The effect of BC on lipid metabolism and predisposition to atherosclerosis was assessed by measuring the concentration of cholesterol (CHOL). None of the cellulose types had an impact (p>0.05) in cholesterol levels [52].
It is important to note that the reference values established for biochemical tests, may not precisely represent those of a certain population or animal species and should, therefore, be carefully interpreted, once there is a wide range of physiological variation. These variations are in uenced by environmental conditions, such as gender, age, origin, breeding system, feeding and lineage, which might have interfered with the obtained results [49,53]. According to the Scienti c Committee for International Harmonization of Clinical Pathology Testing "the concurrent control data are more appropriate than historical reference ranges for comparison with test material groups" [49]. By analyzing the serum markers, we can note that depending on the reference source used, the obtained results may or not fall within the range of reference values. Nevertheless, no signi cant differences (p>0.05) being noticed among groups, the results suggest that BC apparently does not exert any systemic toxicity, at least amenable to be detected by serum biochemistry analysis.
The cellulose tracking using Carbohydrate-Binding Module (CBM3A) from Clostridium cellulolyticum coupled to a green uorescent protein (GFP) in different parts of the intestinal tract was performed. In unwashed intestines using the GBT (Fig. 3), the intestines of the control females exhibited higher cellulose content than the dosing groups, mainly in the jejunum, ileum and colon. For males, cellulose was detected in all experimental groups, mainly in the ileum and colon portions. The cellulose in the intestines of the control animals was ascribed to the type of diet supplied to the animals, which contains 4.1% crude ber.
In the males of the BC dose group, cellulose was detected in the lumen of the jejunum, ileum and in higher amounts in the colon, whereas for females, low amounts of cellulose were found. Since all animals grew normally and no relevant inter-group variations were found in serum biochemistry (section Assessment of the systemic toxicity), the observed differences in cellulose distribution may be due to inter-individual variability and to the (potentially variable) time gap between the last meal and euthanasia.
Unwashed intestinal histological sections processed by GBT have shown large amounts of cellulose, mainly in the ileum and colon regions. As expected, a signi cantly lower amount of cellulose was detected in the washed intestines, where cellulose was found only in a few cases, without any noticeable difference of frequency or location between the three experimental groups (Fig. 4 and 5). In such cases, it was located in the intestinal lumen or between the villi and in Lieberkuhn crypts (Fig. 6). Thus, intestine washing appears to have been incomplete, due to the intricate structure of the mucosa layer. From all of the collected images, no evidence of cellulose intestinal persorption (either bacterial or vegetal cellulose) was found in the lamina propria, the submucosa or in the gut associated lymphoid tissue. In particular, it never colocalized with the Peyer Patches, nor with colonic patches (Fig. 3-5; white arrows). The analysis of confocal microscopy images at higher magni cation and at different z-planes (depths) (Fig. 6) showed that cellulose particles/ bres (either bacterial or plant) were con ned to the Lieberkuhn cripts and between the intestinal villi. No internalized cellulose was detected in the intestinal absorptive epithelium nor in Peyer´s patches or colonic patches (Fig. 6), con rming that neither bacterial or plant cellulose can penetrate the intestinal tract mucus layer and be absorbed.
Regarding the cellulose staining on serial histological samples of the mesenteric lymph nodes (MLN) from all animals, some discrete green dots were detected in all experimental groups (Fig. 7). This was observed with similar frequency in the positive and negative staining controls (omitting the incubation with CBM-GFP conjugate), both showing a similar green diffusive signal. Therefore, this was assigned to auto uorescence of MLN tissues detected in green channel.
Some chromophore molecules, such as proteins, nicotinic coenzymes, avins and lipopigments, widely present in biological tissues, when excited by radiation of suitable wavelength, emit the so-called auto uorescence [54]. This has been previously observed in murine intestinal tissue sections [55], but to our knowledge there are no reports on auto uorescence of the rats' MLN. The rounded diffusive pattern of the observed auto uorescence signal is clearly distinguishable from the longer and sharper speci c staining patterns observed with cellulose, which also exhibited a brighter staining.
The intestinal absorption and translocation of microparticulate material, including cellulose, has been a longstanding study. Upon early publications in the 1960s on the persorption of ingested particulate cellulose and on the demonstration of its presence in the circulating blood stream, further research reported that microcrystalline particles ranging in size from 5-150 µm could be persorbed and detected in venous blood samples taken 1-2 hours after ingestion by rats, dogs, minipigs and in 1 human volunteer. Other studies suggested that, in single dose tests, persorbed particles are cleared from tissues within a few hours and they do not accumulate on repeated dosing even for several months [43].
Reaching the systemic dissemination would require a translocation route through the intestinal epithelium. Possible mechanisms of nanoparticle translocation are: (1) through the M-cells in the Peyer's patches; (2) through enterocytes by passive diffusion; (3) through enterocytes by transcytosis; (4) through the paracellular space [56]. As discussed by Koshani and Madadlou, 2018 [56] for cellulose nanocrystals, route 1 (through the M-cells in the Peyer's patches) is mostly employed for uncharged hydrophobic particles and includes an active transepithelial vesicular transport system from the lumen directly to lymphoid cells and tissues. The overall e ciency of this process is questionable since the population of M cell is < 1% of the intestinal epithelia. In this work, no cellulose particles were detected in the lymphoid tissues nor on the Peyer´s patches. The passive transcellular transport mechanism (route 2) requires a high level of hydrophobicity as it involves partitioning into and diffusion across the cell plasma membrane. This path is normally restricted to small highly hydrophobic molecules, thus excluding cellulose. Endocytosis (route 3) is classi ed into phagocytosis and pinocytosis. The phagocytic cells, such as macrophages, typically internalize foreign materials with sizes larger than 0.5 µm, while pinocytosis occurs in all types of cells for a wide range of particle sizes from approximately 50 nm to 5 µm. Due to the low endocytic activity of enterocytes, active transcellular transport mechanism will play a minor role in the clearance of particles from the intestinal lumen. However, phagocytic activity similar to that demonstrated by enterocyte TLR4, promoting bacterial translocation across the intestinal barrier, may result in absorption. Route 4, the paracellular pathway, is also not likely to allow the pass-through of cellulose particles, because it is restricted to particles (molecules) measuring less than 5 nm. BC presents randomly assembled ribbon-shaped brils with less than 100 nm in diameter. These are composed of elementary nano brils with lateral size of 7-8 nm, aggregated in bundles; these brils have several micrometres in length. Plant cellulose presents micrometer size [25][26][27][28][29]. Overall, the hypothetical absorption of cellulose particles bers is thus con ned to routes 1 and 3 and it may be expected to represent a very ine cient process. Indeed, the several studies addressing the fate of cellulose particles in the intestinal tract have found no evidences of translocation [1]. A recent study on the use cellulose nanocrystals as Pickering stabilizing agents in oil in water emulsions concluded that CNC were entrapped in the intestinal mucus layer and failed to reach the underlying epithelium. The authors concluded that CNC could be used safely as an effective emulsi er [57].
The intestinal absorption of the BC used in this work is particularly unlikely to occur, taking into account the size of the ber's bundles obtained by disintegrating BC membranes produced by static culture. Most of the fully swollen bers have at least few tenths of micrometers, spanning up to several hundred/thousands micrometers, as shown elsewhere [58]. Thus, while the persorption or translocation of microcrystalline cellulose is still a matter of debate, the occurrence of this process for such large BC ber bundles is rather doubtful. On the other hand, it must be considered that cellulose is partially hydrolyzed by the ora in the gut, although this is a process that occurs to larger extent in the rat than in humans [59]. In this case, a size reduction may occur along with degradation. Consequently, it may be speculated that smaller particles may be generated and translocated more e ciently. In that case, however, it may be questioned whether BC, microcrystalline cellulose and the cellulose naturally present in food will, once partially degraded, be much different from each other.
The cellulose naturally occurring in food, and specially the pure cellulose already used as food additive, through degradation, must undergo a comminution process and give rise to cellulose particles with a widespread range of sizes and shapes (for example in colloidal celluloses, from FMC Biopolymers, commercially used as food additives, 99% of the particles are <1 µm [60, 61]). While we did not nd evidence of cellulose translocation or accumulation in the gut lymphoid tissue, we speculate that the low e ciency translocation process -if it occurs to some extent -would occur as well with any other cellulose source present in food, as it undergoes degradation, in case this occurs to signi cant extent.  [7]. Although, the experimental design deviated in some points from the OECD Guideline 408, such as the duration, amount of test substance, timeframe and bloodwork analyses, its execution fully complied with the quality standards.

Methods
Eight weeks old Wistar Han IGS Rats [Crl:WI(Han)] (Wistar Rats), twelve males and twelve females, were used in this study. The rats were obtained from Charles River (Barcelona, Spain) and bread at I3S. On arrival the animals were examined for signs of health status, followed by a one-week adaptation period.
They were kept under controlled environmental conditions for one week before starting the experiment. Afterwards, each male and female rat was randomly allocated in groups of 2 to compose the different experimental groups. After allocation, each rat was uniquely numbered with a color marker in their tail and placed in polycarbonate type III H cages with a stainless-steel wire lid and a polysulphone ltertopcage (Tecniplast), with corncob and carboard tubes as bedding materials. An arti cial light/dark cycle with a sequence of 12 h was applied. The room was ventilated with about 15-20 air changes per hour. A temperature of 22 ± 2 °C and a relative humidity of 55 ± 15% was maintained.
Food and water were provided ad libitum to all animals during the experiment. They received a commercial 2014S Teklad rodent diet (Teklad diets, ENVIGO) based on 14.3 % crude protein, 4% fat, 48% carbohydrates and 4.1% crude ber in addition to vitamins and fatty acids.

Test substance and dosing concentration
Given the high viscosity of BC, the aqueous concentration for daily oral gavage was adjusted to 1% (m/v) -as measured by gravimetry after drying overnight at 105 ºC. The ground BC suspension was then sterilized by autoclaving for 20 minutes, at 120 °C and 1 bar.
In parallel, 1% (m/v) aqueous suspension of commercial plant cellulose (PC), Avicel LM310 (kindly provided by FMC Biopolymers) was also prepared and sterilized.
All the cellulose suspensions were prepared beforehand and stored at 4 ºC. Before oral administration, the suspensions were warmed to room temperature and homogenized by vigorously vortex agitation.

21-day repeated-dose toxicity study
The study comprised two dose groups and one control group, each having four rats/sex (Table 3). Each rat from the dosing groups received a single dose of 0.75 mL of 1 % cellulose suspension (plant or bacterial), daily, for a period of 21 days. The administration was performed in the morning at a xed time by oral gavage, using a polypropylene gavage needle of 1.3x1.3 mm without ball tip. The control group did not receive oral gavage. The animals' well-being was observed daily, and all the cages were inspected for possible deaths or moribund animals, twice daily. Detailed clinical observations were conducted prior to the rst exposure to the compounds (pre-dose) and once a week thereafter. All animals were assessed for the following clinical signs: changes in the eyes, skin, fur, mucous membranes, secretions and excretions, autonomic activity and changes in their gait. Their individual body weight was recorded at start of the treatment (day 0), in weekly intervals thereafter, and before the euthanasia. By the end of experiment, the animals were deeply anesthetized with gaseous iso urane in 0.2-0.3 L/min O 2 , and terminal blood was collected by intracardiac puncture into CAT serum Sep Clot Activator tubes (VACUETTE ® ) for analyses of the biochemical parameters. The blood samples collected for biochemistry were allowed to clot and the serum was obtained by centrifugation at 5000xg for 10 minutes. All animals were subjected to gross necropsy. The collected organs were xed with 4% paraformaldehyde (PFA) and processed for para n-embedding in an automated tissue processor (Leica EG140C

Clinical biochemistry tests
Pathology Embedding Center Para n Dispenser). Para n embedded specimens were cut in 5 μm-thick sections (MICROM HM325) and stained with hematoxylin & eosin (H&E), using an automated stainer (LEICA ST5010 Autostainer XL). Also, histological samples of the different regions of the intestinal tract from the different experimental groups were subjected to H&E staining. All H&E-stained samples were visualized using an optical microscope (BX61 Olympus) coupled to a DP74 digital camera (Olympus) and using bright eld. Images were acquired using CellSens Imaging Software Version 1.16 (Olympus).
The histopathological analysis was blindly conducted by an independent experienced pathologist.

Cellulose tracking -intestinal-tract histological processing
Gut bundling technique (GBT) The GBT was adopted for half of the females and males from each experimental group (Table 4). The small and large intestines were exposed and collected. Afterwards, they were cut into three equal segments to obtain the duodenum, jejunum, and ileum regions. The distal colon and the mesenteric lymph nodes

Swiss roll technique (SRT)
For the remaining animals of each group, two males and two females (Table 4), the small intestines were collected and cut into three equal segments to obtain the duodenum, jejunum, and ileum regions (as with the GBT). Similarly, the distal colon was also collected, as well as the mesenteric lymph nodes from each animal. The lumen of each small and large intestine-portion was washed with PBS, cut longitudinally, opened so that the lumen is facing upward and then rolled. All these tissues were embedded in OCT compound (Tissue-Tek®, SakuraTM, The Netherlands), frozen using isopentane in liquid nitrogen and cryo-sectioned in 20 µm-thick sections using a cryostat (LEICA CM 1900).

Statistical analysis
The obtained raw data were statistically analyzed using Graph Pad Prism software (version 5.01, windows). The values were expressed as the mean ± standard deviation of the mean (SD). Differences in data of body weight and body weight gain, organ weight, hematological and clinical biochemistry were statistically analyzed using Two-way repeated measure ANOVA, with Geisser-Greenhouse correction and Tukey post hoc test. All analyses and correlations were assessed at 95 % level of con dence (p < 0.05).

Consent for publication
Not applicable.

Availability of data and materials
The representative summary data of the analyzed datasets within this study were provided in the manuscript. However, all of the collected results can be made available by the corresponding author, upon reasonable request.

Competing interests
The authors declare that they have no competing interests Funding This study was supported by the Portuguese Foundation for Science and Technology (FCT) under the scope of the strategic funding of UIDB/04469/2020 unit.
Authors´ contributions ACR and LR these authors contributed equally to this work, contributed to experimental design, performed experiments, data analysis and co-wrote manuscript.
SD-S and AT-S these authors provided the cryostat to cryo-sectioned of intestine-portion using a and provide the equipment and conditions to microscopic observations. NL experienced pathologist performed the histopathological analyses of the animals' collected organs (liver, kidney and spleen) and the different regions of the intestinal tract.
FD and MG these authors contributed equally to this work, contributed to experimental design, data analysis and prepared the rst version of manuscript with input from all authors.
All authors contributed to the writing of the manuscript. All authors read and approved the nal manuscript.  Representative uorescence photomicrographs of histological sections of unwashed intestines processed by GBT, stained for cellulose (green), cytoskeleton (red) and nucleus (blue). Lymphoid tissues -peyer patches and colonic patches (highlighted by white arrows). Scale bar 200 µm.

Figure 4
Representative uorescence photomicrographs of histological sections of washed intestines processed by GBT, stained for cellulose (green), cytoskeleton (red) and for nucleus (blue) visualization. Lymphoid tissues -peyer patches and colonic patches (highlighted by white arrows). Scale bar 200 µm.

Figure 6
Different z-planes of confocal uorescence photomicrographs of histological sections from the washed ileum and colon, processed by SRT and stained for cellulose (green), cytoskeleton (red) and nucleus (blue). From bottom to top (in each row, from left to right). Scale bars-30 µm; 100 µm.