Creating a new human-based gut-brain model to support personalized nutrition

The role of intestinal microbes in the regulation of the immune system, metabolism of nutrients, and production of neuroactive compounds has established them as key modulators of the gut-brain axis^1–4, the homeostatic, reciprocal communication between the human intestine and the brain^2,5–7. Disorders of gut-brain interactions, like irritable bowel syndrome (IBS), are characterized by low-grade inflammation and altered bowel habits, but they are also often associated with altered central nervous system (CNS) processing of gut sensory signals, as well as anxiety and depression^8–10. Although growing evidence has suggested that the gut microbiota impact brain function and behavior, mainly in the context of neurodegenerative and psychiatric disorders, the potential underlying molecular and cellular mechanisms remain vastly uncovered^2,4,11–14.

The translocation of certain gut microbiota-derived compounds from the luminal/mucosal side to the basolateral/serosal side of the intestine, and subsequently into the bloodstream, can affect gut-brain signaling. This translocation influences CNS function by modulating systemic levels of neuroactive compounds that can eventually cross or modulate the permeability of the blood-brain barrier (BBB)^2,12,15,16.

Gut microbiota may influence microbial and human metabolism, such as of tryptophan¹⁷, which is utilized for the biosynthesis of 5-hydroxytryptamine (5-HT), also known as serotonin. Although 95% of the total 5-HT is synthesized by enterochromaffin cells in the gut, gut microbiota can also moderate or contribute to that synthesis^18–22. Intestinal 5-HT is unable to pass through the BBB; however, it can bind to 5-HT₃ receptors on afferent vagal neurons locally and ultimately affect stress response and behavior in the CNS^17,23–25.

Interestingly, increasing evidence suggests that a high-fiber diet can have suppressive effects on microbial tryptophan degradation, allowing levels of circulating tryptophan and consequently central 5-HT to rise^18,23. One reason for this could be the higher luminal availability of short-chain fatty acids (SCFAs) deriving from the fermentation of indigestible dietary fiber by the microbiota in the colon. Butyrate is one of the mainly produced SCFAs that is rapidly absorbed by the gut mucosa and provides an excellent energy source for colonocytes²⁶. In addition, microbial synthesis of SCFAs, especially butyrate, has been associated with the suppression of the kynurenine pathway, which ameliorates levels of systemic tryptophan^27,28.

Moreover, endogenous butyrate enters the bloodstream and infiltrates the BBB in low concentrations, potentially exerting positive neurological effects^1,29, while exogenous butyrate at pharmacological doses has been shown to generate prominent effects on both the brain and BBB in various animal studies^1,30,31. For instance, increased expression of tight junctions and decreased permeability of the BBB were observed after mono-colonization of germ-free mice with butyrate-producing bacteria or by oral administration of sodium butyrate³⁰. Even though animal models provide a strong foundation to study the impact of gut microbiota alterations on the BBB, they inevitably fail to capture certain human physiological aspects (e.g., brain function, immune system, gut microbiota composition).

This weakness, along with sustainability and animal welfare concerns, has encouraged the rapid advancement of in vitro models which can significantly reduce or even deplete the need for animal studies³². In vitro modeling with dynamic cell culture conditions has been considered one of the most promising approaches to mimic the microbiota-gut-brain axis' complexity^32,33. In vitro BBB models are mainly based on endothelial cell cultures or co-cultures with other perivascular cells (i.e., pericytes, and astrocytes) in two- or three-dimensional platforms^34,35. Moreover, single- and multi-organ platforms, based respectively on the "organ-" or "body-on-a-chip" concept, have also been proposed^36–39, while more and more intestinal and BBB models strive to implement the use of induced pluripotent stem cells^{34,37,40–44}.

Following these rapid developments, our model uses a combination of ex vivo and in vitro experimental techniques to explore the potential interactions of gut-derived metabolites and cells representing the endothelium of the BBB.

In this study, we hypothesized that microbial metabolites, such as SCFAs, after passage through intestinal tissue, affect the uptake of serotonin-precursor tryptophan in fibroblasts.

Following the advancement of gut-brain axis models, this study aimed to establish an ex vivo-in vitro model for the passage of luminal metabolites through the gut barrier and their subsequent impact on the blood-brain barrier (BBB) through the humoral route, applicable for precision nutrition assessments. By collecting serosal fluids of distal colonic biopsies after Ussing chamber experiments, we attempted to imitate the passage of microbial metabolites, such as short-chain fatty acids (SCFAs) (i.e., acetate, propionate, butyrate), from the luminal to the serosal side of the gut lining and to partially capture the secretome of the biopsy itself.

To focus on the humoral route of gut-brain communication and to test how these fluids and their content could impact the BBB, we introduced them in the culture media of a human dermal fibroblast cell line. Fibroblasts present similarities in the expression of amino acid transport systems, such as tryptophan transporters LAT1 and 2, with brain microvascular endothelial cells that comprise the BBB^46,47 and are often used to model psychiatric^48,49 and neurodegenerative disorders^50–52.

Additionally, we applied our model to test whether serosal fluids could differentially impact the BBB by using proxies of fiber-rich diets (in vivo butyrate exposure, ex vivo exposure to supernatant of fecal fermentation enriched with wheat-bran fiber fractions) from a healthy or irritable bowel syndrome (IBS) intestinal environment. Lastly, we implemented the use of H₂O₂ as a stressor, thus partly assessing potential therapeutic or protective effects of the serosal fluids.

Serosal fluids had a close to physiological pH and did not present any microbial growth of aerobic bacteria, although the Ussing chamber is an open non-sterile system. As per protocol, fibroblasts were cultured with addition of antibiotics to the culture medium and at no point microbial contaminations were observed. After acute (1 h), as well as relatively extended (24 h) exposure to serosal fluids, fibroblasts showed typical spindle-shaped morphology and typical adhesive characteristics examined by microscopic examination. Two molecular biological tests showed that the cells were normally, metabolically active (cell viability assays) and had a normal cell membrane integrity (cytotoxicity assays). Hence, those findings support the feasibility of such experiments, and applied concentrations can be considered as adequate.

For both acute and relatively extended exposure solely to serosal fluids, neither gene expression, nor protein expression, nor protein activity (i.e., transmembrane tryptophan transport) were significantly affected.

To evoke oxidative stress, the usage of H₂O₂ in vitro is common practice, and we have validated that the stressor worked in our setting by compromising the tryptophan transporter system (see Supplementary Fig. 7). However, tryptophan transport remained unchanged and in one case (HC-FERFIB) even increased when the cells were exposed to serosal fluids prior to oxidative stress, which could indicate a protective effect of the serosal fluids against oxidative damage effects, further supported by the alterations we see in gene and protein expression levels.

Interestingly, as a preventive measure for stressed cells, all serosal fluids, especially HC-FERFIB, significantly decreased gene expression of relevant tryptophan transporters. Despite the pronounced effect on gene expression, tryptophan transport appeared unaffected. However, HC-FERFIB for the 24 h preventive measure increased tryptophan uptake, creating a pattern in disagreement with reduced gene expression of LAT1 and LAT2.

In contrast, as a treatment measure for stressed cells, these alterations were either absent or minimal. Serosal fluids evoked a few changes in gene and protein expression of only the minor tryptophan transporter LAT2, e.g. for HC-FERFIB ex vivo manipulations, but without altering tryptophan transport.

Generally, we observed few differences to be elicited by in vivo, intestinal manipulation (90-min butyrate infusion) serosal fluids (HC-BUT or IBS-BUT), with the IBS-derived samples being especially interesting since mucosal and fecal supernatants of such patients have been shown to cause nerve activations in the enteric nervous system^53–55. Instead, particularly serosal fluids obtained from the 90-min ex vivo exposure to fecal fiber or control fermentation supernatants (HC-FERFIB or HC-FERCON) seemed to produce prominent, biologically relevant effects on all levels of evidence (gene/protein expression of LAT1 and LAT2, and tryptophan uptake).

Furthermore, based on the metabolite profiles of the different serosal fluids, we observed that the total SCFA concentration is comparable to physiological levels in peripheral blood (80–180 μM)⁵⁶, yet at least 5-fold lower (~15 μM). However, the ratios are in agreement with the literature with acetate being the predominant SCFA. Although butyrate and propionate are usually untraceable in venous peripheral blood, they are found at concentrations of 1–15 μM in the systemic circulation, which are comparable to the measurements in serosal fluids of this study^57,58. In particular, the serosal fluid HC-FERFIB had a visually higher content of SCFAs, while also showing a pattern of decreased tricarboxylic acid cycle metabolites. These observations may indicate that the increased control over experimental conditions during the ex vivo manipulations of the intestinal biopsies, as well as the consecutive exposure of the biopsies to these manipulations could lead to more consistent effects throughout the different levels of evidence (as opposed to more varied effects detected after in vivo manipulations), which can be captured using the current gut-brain axis model. Noteworthy is also, that not only SCFAs but also structural motifs of fiber fractions have recently been detected in fecal fermentation supernatants, prepared in an identical manner as the ones utilized for the ex vivo manipulations, and interacted differentially with human toll-like receptors⁵⁹.

As for the proinflammatory cytokine profile, all ten cytokines were at detection levels in all serosal fluids (see Supplementary Table 1). Cytokines were detected in all samples; even in the serosal fluid carrier solution (G-Krebs), and in the serosal fluids of the healthy participants. Interestingly, IBS-derived serosal fluids did not seem to have higher levels of proinflammatory cytokines, at least when visually compared to the serosal fluids of healthy participants (Supplementary Fig. 1). The concentrations measured in our experiment are comparable to values detected in serum levels of both young and older adults, encouraging further exploration of their potential contribution to the results we observed in this study on gene and protein expression of LAT1 and LAT2, as well as their activity, in fibroblasts⁶⁰.

Several studies on human dermal fibroblasts have demonstrated the capacity of different compounds with antioxidant properties (e.g., carotenoids, dietary polyphenols, estradiol, curcumin) to protect against subsequent exposure to H₂O_2, while failing to produce a significant rescue effect as post-stress treatments^61–63. Contents in the serosal fluids with potential redox modulatory effects (i.e., SCFAs) could act in a similar way^58,64, although it should be noted that we previously showed that directly adding sodium butyrate to the fibroblasts' culture media successfully restored oxidative stress-induced deficits of tryptophan transport by e.g., increasing the gene expression of the two tryptophan transporters LAT1 and LAT2⁴⁵.

In our previous study, high concentrations (mM) of butyrate alone had effects, while low concentrations (µM) did not. Both had treatment effects, though preventive effects were not tested⁴⁵. Propionate (1 µM incubation for 12 h) showed preventive effects on efflux transporter activity and various expression patterns on human cerebro-microvascular endothelial cells before lipopolysaccharide (LPS) exposure, i.e., microbial stress, and protected from oxidative stress⁶⁵. Also, tight junction expression and cytoskeletal arrangements were positively affected in murine brain endothelial cells exposed to butyrate or propionate with and without microbial stress (1 µM for 24 h, co-incubation with LPS for the second half)⁶⁶. The effect of SCFAs has also been tested in a sequential system of four micro-physiological mesofluidic systems (the human 3xGLB system) including colonic epithelial cells from organoids (apical exposure to 20 mM in the ratio 6 acetate: 2 propionate: 2 butyrate for 4 days), peripheral blood mononuclear cells, a hepatocyte and Kupffer cell co-culture, and induced pluripotent stem cell (IPSC)-derived cerebral cells, the latter from a healthy and a Parkinson's disease donor—showing effects on gene expression patterns³⁷.

There have been few additional attempts to model the microbiota-gut-brain axis in vitro, all of which are unique in their setup and usually focus on the humoral signaling pathway. The MINERVA project adds fecal samples from a healthy individual and a patient with Alzheimer's disease to the microbiota compartment of a platform with five separate organ-on-a-chip devices including generic Caco-2 cells, macrophages and lymphocytes, endothelial cells and astrocytes, neurons and microglia cells³⁶. Another attempt is neuroHuMiX which co-cultures Limosilactobacillus reuteri F275, Caco-2 cells and IPSC-derived enteric neurons in proximity allowing diffusion of secreted factors and metabolites via semipermeable membranes without direct cell-cell contact⁶⁷. A dual-flow tissue perfusion device cultured an array of gastrointestinal and cerebral cell lines together with bacterial extracellular vesicles⁶⁸.

As in all models, translating results from the lab bench to human physiology is a serious challenge. In our case, membrane transporters of dermal fibroblasts are comparable to BBB endothelial cells; however, we do not account for the actual microenvironment surrounding BBB structural cells. Systemic tryptophan, either free or albumin-bound, can be transported through specialized brain microvascular endothelial cells (BMECs) of the BBB by two closely related plasma membrane transporters, large neutral amino acid transporter 1 (LAT1) and 2 (LAT2)^69,70. Commonly abundant in biological barriers, these transporters are also expressed in fibroblasts, making them a well-established in vitro model to study the effects of gut-derived substances on BBB physiology^{46,47,71–74}. Importantly, fibroblasts were cultured to visual determined confluency of 95%, not creating a tight barrier. The herein presented work focused on the uptake of tryptophan. Future work should consider the cross-barrier transport, and could utilize trans-well plate systems. For this reason, cell viability was assessed by morphological inspection (adhesive cells are alive, while floating cells or debris indicate cell death), and not measurements of transepithelial electrical resistance (TEER) prior to any experiment; and two molecular assays (AlamarBlue viability test and LDH cytotoxicity test) were performed for all conditions.

Similarly, colonic biopsies and Ussing chamber experiments are well-known approaches to study gut barrier function. Nevertheless, there are certain limits to maintaining the integrity of the biopsies and preserving the mucus layer on the luminal side of the biopsy, which is generally lost during processing, consequently eliminating its components from the current model. To ensure tissue viability, only biopsies with a potential difference <0.5 after reaching equilibrium in the Ussing chamber were used for experiments and subsequent collection of serosal fluids. The lowest, in any individual Ussing chamber, measured TEER was 9 Ω cm⁻². For further details, please refer to Scharf et al. 2025⁷⁵. As discussed above, the Ussing chamber creates an open, non-sterile environment. Several of the herein presented results indicate that samples collected from this system nevertheless seem to be suitable to be added to cell cultures.

Serosal fluids have been collected from recently conducted human trials that focused on other primary objectives. Available volumes of serosal fluids were limited, which allowed only a small number of biologically independent experiments, and limited statistical conclusions. While the colonic in vivo exposure to butyrate was conducted in both HC and IBS, the ex vivo exposure to supernatants from fecal fiber fermentations was solely conducted in HC. Although the assessment of serosal fluids from IBS biopsies ex vivo exposed to supernatants from fecal fiber fermentations would have been a strong addition to this work, such samples were not available, but should be considered in the future. As previously reported by Scharf et al. 2025⁷⁵, IBS symptoms were assessed using the Gastrointestinal Symptom Rating Scale (GSRS) for IBS for the week before sample collection, and averaged at 3.8 ± 0.8 which indicates moderate discomfort (on a scale from 1−7 for no to very severe discomfort) and is typical for IBS cohorts. Upon asking on the day of sample collection, none of the participants reported an alteration in their medical condition. However, this was a short-term observation and not compared to long-term IBS symptomatology.

While the effect of long-term diet on the results cannot be ruled out, the last meal before a fasting period prior to colonoscopy was standardized to low fiber intake for all participants.

Moreover, interactions between the involved systems (microbiota, gut and blood-brain barrier) are sequential in the presented model, more similar to functional coupling as defined and tested by Vernetti et al., which uses five topologically separate micro-physiological systems including primary intestinal cells and a BBB/neuron compartment⁷⁶. In comparison, other gut-brain axis models have a fluidic organ-on-a-chip structure, which offers different temporal and topological information^{34,36,38,67,68}.

Our approach is based on well-established methodologies, both on the intestinal and the brain side, focusing on the humoral route of gut-brain communication, and providing plenty of opportunities to be extended as well as to be applied bidirectionally. This highlights the humanized nature of this model, assembled by elements isolated directly from the human body and that are specific to the individual providing them, including the possibility to be linked to individual disease background. Namely, fibroblasts can easily be isolated from individuals for more personalized investigations in parallel with their colonic biopsies and serosal fluids. The well-documented use of fibroblasts allows for experimentation with more complex cultures, either three-dimensional or co-cultures^77,78. Also, dermal fibroblasts can be reprogrammed into IPSCs^34,37,40,79, giving our model the possibility to evolve and increase similarity with BBB characteristics or to expand by adding an ex vivo approach of the neuronal system. Taken together, our model opens possibilities for assessments on an individual level, hence supporting precision nutrition or medicine.

To our knowledge, there have only been a few first attempts to build a generalized gut-brain axis model ex vivo, allowing more detailed mechanistic assessments than sole in vivo examinations³³. Our humanized model can efficiently be used to detect effects on all levels—gene expression, protein expression, protein activity, and metabolite profiles. While serosal fluids alone did not exert effects, serosal fluids as a preventive measure (which is the physiologically most relevant one) exerted more effects than as a treatment measure. Generally, serosal fluids originating from ex vivo modification of the intestinal environment had more prominent effects on tryptophan uptake, likely due to methodological differences, than those of in vivo manipulation. Consequently, ex vivo approaches may be more relevant to study with the current gut-brain axis model.

Transparent Peer Review file Supplementary Information Description of Additional Supplementary Materials Supplementary Data 1 Reporting Summary