RORγt-Expressing Pathogenic CD4⁺ T Cells Cause Brain Inflammation during Chronic Colitis

Growing evidence suggests that disturbance in the gut/brain axis in inflammatory bowel disease (IBD) is associated with neuropathological conditions. Both symptomatic and asymptomatic CNS abnormalities as well as neurobehavioral disorders such as anxiety and depression have been reported extensively in both Crohn’s disease and ulcerative colitis (UC) patients (1–10). In both types of IBD, a high degree of correlation with several neurodegenerative disorders such as multiple sclerosis (MS)–like conditions, dementia, and Parkinson’s disease have been reported (11–14). Demyelinating disorder is observed to occur more commonly among patients with IBD compared with non-IBD patients (11, 15). IBD patients can present various manifestations, including vascular injuries, inflammation of blood vessels in the brain, and increased neural infections (10). In particular, research shows that 4 in 10 people with IBD experience feelings of anxiety and/or depression, and IBD patients were significantly more likely to have a lifetime diagnosis of major anxiety and depression (8, 9). Psychological disorders in patients with IBD may lead to high risk of relapse and poor treatment compliance (16). Moreover, IBD patients with depression/anxiety have higher rates of hospitalization and increased disease severity than do those without IBD (17). The limbic system, including the hippocampus and the hypothalamus, constitutes one of the main areas of the brain that have been implicated in depression and anxiety disorders related to colitis. Decreased hippocampal neurogenesis has been demonstrated in colitic mice (18). The hypothalamus is implicated in emotion processing and plays a major role in depression, anxiety, and pathology of mood disorders (19–21). It has been demonstrated that psychological stress acting through the hypothalamic/pituitary/adrenal (HPA) axis increases intestinal motility in IBD patients, which causes local inflammation in the gastrointestinal tract and worsens IBD (7, 22, 23). In turn, gut inflammation in IBD patients is believed to activate peripheral inflammatory responses leading to the development of mood disorders associated with the hypothalamus (24).

The cause-and-effect relationship between IBD and neurologic comorbidity is controversial (10, 22). Researchers have previously demonstrated that intestinal inflammation can influence brain activity and psychological behavior in preclinical models (22). Studies of animal models with colonic inflammation have shown inflammatory changes in the brain (25–27). Despite these observations, little is known about the main mechanistic factors behind these phenomena (1, 2, 5, 6). It has yet to be established whether the hyperactive immune response associated with inflammation of the gut triggers brain inflammation that results in neuropathological disorders during IBD. Possible explanations, including vasculitis, thromboembolism, and malnutrition, have been proposed (10). Although it is known that proinflammatory CD4⁺ T cells contribute to the pathogenesis of both IBD and MS (28, 29), whether IBD can trigger infiltration of gut-derived proinflammatory CD4⁺ T cells into the brain has not been investigated. New studies have demonstrated roles of a broad range of immune cells, including innate lymphoid cells and T cells, in regulating inflammation and neurologic diseases in the brain (30). Moreover, IL-17 and IFN-γ–producing autoreactive CD4⁺ T cells are the main pathogenic populations that infiltrate the CNS and determine the clinical course of autoimmune disease of the brain (31). Among the different subsets of CD4⁺ T cells, Th17 cells are more effective inducers of neuroinflammation compared with Th1 cells, as they cross blood–brain barrier (BBB) more effectively and preferentially target the astrocytes (32–34). However, their role in modulating the gut/brain axis during IBD is not yet known.

Due to the unusual association of IBD with neuropathological comorbidity, we asked two central questions. 1) Does chronic colitis cause inflammation in any part of the brain leading to neuropathological disorders? 2) How does chronic colitis cause brain inflammation? To understand whether neuropathological conditions during IBD represent a simple association or a causation of IBD, we investigated the development of brain inflammation in a mouse model of chronic colitis. We found a significant correlation between gastrointestinal inflammation and proinflammatory response in the brain contingent on the presence of colitogenic retinoic acid–related orphan receptor γt (RORγt)-expressing CD4⁺ T cells. Moreover, this link between peripheral and CNS inflammation appeared to be causally linked to deleterious effects on brain neurobehavioral function. Therefore, these findings suggest that gut-derived CD4⁺ T cells trigger inflammation in the CNS that contributes to the pathogenesis of anxiety and depression in IBD patients.

The following mouse strains used were purchased from The Jackson Laboratory: C57BL/6J (B6), Rag1^−/−, and B6(Cg)-Rorc^tm3Litt/J. For all of the experiments, 6- to 10-wk-old mice were used. Abs were purchased from either eBioscience, BD Biosciences, or Fisher Scientific, including CD3, CD4, CD25, CD45RB, T-bet (eBio4B10), Foxp3 (FJK-16S), RORγt (AFKJS-9), IL-17A (TC11-18H10 or eBio17B7), and IFN-γ (XMG1.2).

For the adoptive T cell transfer model, CD25⁻CD45RB^hi CD4⁺ T cells (4 × 10⁵/mouse) from either B6 or Rorc^fl/fl Cd4-Cre mice were injected through the i.p. route in age- and sex-matched Rag1^−/− mice (8–10 wk old, males or females) for colitis induction, and recipient mice were monitored for 12 wk. For the dextran sulfate sodium (DSS) model of chronic colitis induction, mice were treated with three cycles of 2.5% DSS (colitis grade, m.w. 36,000–50,0000; MP Biomedicals) dissolved in drinking water for 7 d, and in between DSS treatments, mice were kept in normal drinking water for 10–12 d (35).

As per the University of Alabama at Birmingham Institutional Animal Care and Use Committee approval, mice were anesthetized using isoflurane. For isolation of the brain, mice were perfused with saline through the left ventricle until clear perfusion fluid flooded from the right atrium to reduce the presence of blood mononuclear cells. After heart perfusion, mice were sacrificed by decapitation. Dura meninges was removed to expose and extract the brain. Following extraction, tissues were cut into small pieces before digestion. The tissue was incubated in RPMI 1640 containing collagenase IV (1 mg/ml, Sigma-Aldrich), Dispase (0.5 mg/ml, Life Technologies, Invitrogen), and DNase I (0.25 mg/ml, Sigma-Aldrich) for 20 min at 37°C in a shaker incubator. Then, the mixture was filtered using 70-μm nylon cell strainers to capture a single-cell suspension. Mononuclear cells were isolated from the interphase between two Percoll (GE Healthcare) gradients (70% and 38%). Cells were washed twice in cold PBS containing 1% FCS and subsequently used for Ab staining. Colonic lamina propria (LP) lymphocytes were isolated as described previously (36). Briefly, the large intestine was removed, cleared of luminal contents and fat, cut into small pieces, and washed in chilled HBSS without Ca²⁺ or Mg²⁺. Minced tissue pieces were incubated in the presence of EDTA for 30 min and vortexed to remove epithelial cells, then incubated in RPMI 1640 containing a digestion mixture. LP lymphocytes were collected after passing through 70-μm strainers. For intracellular cytokine staining, cells were stimulated with PMA (50 ng/ml; Sigma-Aldrich) and ionomycin (750 ng/ml; Calbiochem) for 4 h in the presence of GolgiPlug (BD Pharmingen). For CD4⁺ T cell isolation from spleen, liver, and peripheral lymph nodes (PLNs), tissues were ground and filtered through 70-μm strainers. Single-cell suspensions of spleen and liver were treated with ACK (ammonium-chloride-potassium) lysis buffer to remove RBCs, followed by washing in cold PBS/1% FCS. Lung and kidney tissues were chopped and incubated in digestion mixture as stated before for 1 h at 37°C in a shaker incubator and passed through 40-μm strainers (37). For detection of intracellular cytokines and transcription factors, cells were fixed and permeabilized either in Foxp3 staining buffer (eBioscience) or BD permeabilization buffer. In all cases, Live/Dead fixable near-IR (infrared) dead cell stain (Invitrogen) was included prior to surface staining to exclude dead cells in flow cytometric analyses.

Diffusion-weighted mouse magnetic resonance imaging (MRI) was conducted with a Bruker BioSpec 9.4 T scanner running ParaVision 5.1 software (Bruker BioSpin, Billerica, MA). A Bruker 72-mm volume coil and Doty 24-mm surface coil (Doty Scientific, Columbia, SC) were used for signal excitation and reception, respectively. Mice were anesthetized with isoflurane gas and monitored with an MRI-compatible physiological monitoring system (SA Instruments, Stony Brook, NY). Mice were imaged in the prone position with bite and ear bars for head fixation, and heated water was circulated through the animal bed to maintain animal body temperature throughout the experiment. Diffusion-weighted imaging of the brain was accomplished with a two-dimensional spin-echo sequence using the parameters repetition time/echo time = 3500/30 ms, echo train length = 1, 1 average, field of view = 1.92 × 1.92 cm, and a 96 × 96 matrix for an in-plane resolution of 200 μm. Slice thickness was set at 1 mm, and seven slices were imaged with diffusion b-values of 0, 300, and 800 s/mm². Maps of the apparent diffusion coefficient were calculated using custom code written in MATLAB (MathWorks, Natick, MA). All images were transferred to our local personal computers and analyzed using the MATLAB Image Processing Toolbox.

Following the perfusion of brain with PBS, tissues were extracted and fixed in 4% formaldehyde overnight. Brains were embedded in paraffin block followed by cutting into 5- to 15-μm coronal sections. Slides were immersed in xylene for 5 min followed by immersion in a series of ethanol concentrations for 2 min each. Slides were then immersed in boiled citric acid (0.01 M) for 10 min followed by cooling and several washes in PBS. To avoid nonspecific binding, the brain sections were incubated with 3% serum for 1 h followed by overnight incubation with primary Abs (1:1000; rat anti-mouse CD4 from BD Biosciences) in a humidified chamber at 4°C. On the next day, slides were washed twice with PBS and stained with secondary Abs (1:400; streptavidin-conjugated goat anti-rat IgG-Alexa Fluor 488 from Invitrogen) for 2–3 h at room temperature in a humidified chamber followed by incubation in ProLong Gold antifade reagent with DAPI (Invitrogen).

Colon tissue samples obtained from proximal, middle, and distal portions of large intestines were fixed in 10% neutral buffered formalin, embedded in paraffin to prepare 5-μm sections, and stained with H&E. The tissue sections were examined and scored to evaluate tissue pathology, as previously described (38). In all scoring, the identity of specimens was concealed from the pathologist.

For astrocyte staining, mouse brains preserved in 4% paraformaldehyde were cut using a manual slicing machine to 10-μm thickness followed by rinsing in TBS. To quench endogenous peroxidase, sections were incubated with 3% H₂O₂ for 5 min at room temperature (RT) followed by incubation in a blocking solution containing 3% goat serum. The primary Ab (mouse anti–glial fibrillary acidic protein [GFAP] from Sigma-Aldrich at 1:1000) diluted in 0.5 M TBS-Triton X-100 (0.5%, pH 7.6) buffer was added in a humidified chamber on a shaker table in the dark (overnight) at RT. The slices were rinsed in TBST (5 min each) followed by the addition of secondary Ab (goat anti-mouse IgG-biotin from Sigma-Aldrich, 1:500) and kept at RT for 2 h in a humidified chamber. The tissue was rinsed three times in 20 ml of TBST. A tertiary streptavidin-HRP Ab (a part of the mouse ExtrAvidin peroxidase staining kit from Sigma-Aldrich at 1:1000) was added for 2 h, followed by diaminobendizine substrate reaction for 2–3 min. Finally, the tissue was rinsed to remove excess diaminobenzidine and mounted.

To determine the permeability of the cerebral vasculature, 4% Evans blue (Sigma-Aldrich) diluted in PBS was injected i.v. (3 ml/kg body weight) (39). One hour after Evans blue injection, all of the mice were anesthetized and transcardially perfused with cold saline as described before. The mice were killed and brains removed and imaged. Brain tissues were homogenized with 50% TCA and centrifuged at 1000 × g at 4°C for 20 min. Supernatants were collected and measured at 620 nm to detect the absorbance.

For analysis of myelin integrity, paraffin-embedded coronal brain sections were dipped in xylene twice for 5 min each followed by three changes in absolute alcohol and rinsed in deionized water for 1 min. Deparaffinized brain sections were placed in 0.1% Luxol fast blue solution at 56°C for 1 h in a water bath followed by a quick dip in 0.05% lithium carbonate solution and then in 70% alcohol. For counterstaining, slides were immersed in 0.5% periodic acid for 10 min followed by incubation in Schiff’s solution (Sigma-Aldrich) for 10 min followed by washing in water and dehydrated sequentially in 70% alcohol and in absolute alcohol, cleared in xylene, and mounted. Demyelination quantification was performed in ImageJ. Images obtained from an Olympus microscope were imported into the software. For each image, the scale was reset by setting the “distance in pixels” and “known distance” to 0 and the “unit of length” to pixels. Afterwards, color thresholding was applied. Subsequently, areas of different color contrast (e.g., myelinated versus demyelinated) were measured using the default settings. The normalized demyelinated area values were calculated in R.

For CD4⁺ T cell purification from brain, recovered lymphocytes from the Percoll gradient interface as described before were sorted by flow cytometry for the detection of Il17a, Ifnγ, and Ccr6. For Ccl20, microglia and astrocytes were isolated following previously published protocol by Agalave et al. (40). Brain tissue was homogenized in DMEM following perfusion and centrifuged after passing through 70-μm nylon cell strainers. After resuspension in 70% Percoll, the suspension of cell pellet was used to prepare gradients by layering of 50/35% Percoll and DMEM on top, respectively (at a 6:3:3:2 ratio). Following centrifugation at 2000 × g for 20 min without brake at RT, cells were recovered and pooled from interfaces of 70–50% and 50–35% Percoll gradients that predominantly contain the microglia and astrocytes. After washing with DMEM, cells were used for RNA extraction. Total RNA was isolated as per the manufacturer’s instructions (Qiagen). cDNA synthesis was performed with a SuperScript III First-Strand Synthesis System (Invitrogen), and real-time PCR was performed on QuantStudio 3 (Applied Biosystems) using PowerUp SYBR Green supermix along with the following primers: Il17a forward, 5′-TGAAGGCAGCAGCGATCA-3′, reverse, 5′-GGAAGTCCTTGGCCTCAGTGT-3′; Ifnγ forward, 5′-ACAATGAACGCTACACACTGCAT-3′, reverse, 5′-TGGCAGTAACAGCCAGAAACA-3′; CCR6 forward, 5′-CCTCTGTGCCCGGGTTTAC-3′, reverse, 5′-CATTATCATTTTCGACGGTCTCACT-3′; Ccl20 forward, 5′-TCAACTCCTGGAGCTGAGAATG-3′, reverse, 5′-CCATGCCAAAGCAAGGAAGA-3′; 18S rRNA forward, 5′- GCCGCTAGAGGTGAAATTCTTG-3′, reverse, 5′-CATTCTTGGCAAATGCTTTCG-3′. Relative levels of expression of genes of interest were calculated by using the 2^ΔΔCt method.

This test is used to asses anxiety-like behavior. The light/dark inclination assessment was executed by means of an automated tracking device consisting of a video camera tracker connected to a personal computer (Med Associates, St. Albans, VT). The test itself was done using a light/dark box made of plastic. Each mouse was positioned in the middle of a box subjected to normal illumination along with no illumination in the adjacent box and their movements were recorded for 10 min. Measurements included the total time variation between the two boxes (light versus dark) and number of entries to each compartment (41).

This experiment was used to evaluate depression and anxiety-related behavior in rodents. The maze is composed of four wooden arms of equal length with two closed arms and two opened arms and a central area between the arms. Anxious mice tend to shelter in the more secure area of the closed arms, whereas mice that do not show anxious behavior prefer to spend more time investigating the open arms. To perform the experiment each mouse was positioned in the center area and it was left to freely choose the area it wanted to stay in or explore for 5 min. The time spent in each area was calculated (42, 43).

This test was performed to explore approach/avoidance behavior as an indication of anxiety-like behavior according to Crawley’s standard protocol. The test takes place in a rectangular box consisting of three chambers (44). In one of the chambers, an unknown control mouse was placed in a wired cup, whereas the other room was left empty. In this test, the mouse was given a chance between inspecting the unknown mouse placed in the wired cup or inspecting the empty room. To ensure the accuracy of this test, the box was equally lit by general room lighting, and all mice were given time to acclimate for 30 min before the test started. Habituation (adaptation) was performed for 10 min. To assess anxiety, mice were monitored for 10 min in the chamber. The time consumed by the mouse in each respective chamber along with the frequency of entries to each room were calculated.

A Y-maze test was performed to assess short-term memory for mice suffering from colitis. The test setup consists of a Y-shaped plastic corridor. The test was performed in two phases. In the first phase one of the arms was closed with a wooden door, while the mouse was allowed to explore the other two arms for 8 min. In the second phase of the test (8 min), the opened and the closed corridors forming the “Y” arms are exchanged. Mice that do not show signs of spatial memory deficits tend to remember which arms they had explored and thus will explore the newly opened arm (45).

This measures the mouse’s ability to recall incidents through the retrieval of information from short-term memory storage. In this test, each mouse was subjected to an aversive unconditioned stimulus (e.g., short electrical shock) associated with a sound of 90 db for 30 s. The conditioning of the sound and the electric shock was repeated twice on the first day of the experiments with 90 s between the first shock and the second. On the second day, the ability to retrieve the association was determined by calculating the freezing time shown by the mice. To ensure the accuracy of the test, the mice were allowed to explore the fear condition chamber (Med Associates, Actimetrics chamber system) for 3 min on the first day, whereas on the second day, mice were allowed to explore the cage for 5 min to help retrieve short-term memory conditioning (43).

This test was performed to measure the ability of mice to differentiate between objects that are “familiar” versus “new” objects. The medium of this experiment is a square plastic field with closed sides. The arena is divided into two sides, with an object placed on each side. In the first phase, the mouse was put in the arena and left to explore the two sides containing the two objects for 10 min. In the next phase of the experiment, one object was substituted with a novel item and the mouse was left to freely explore the objects for 5 min. The duration of inspecting the novel object was compared with the time spent exploring the old object. Mice that do not have short-term deficits will spend a longer time with novel objects (42).

This test was used to measure muscle coordination by the cerebellum. In the test, mice have to stroll across a corridor consisting of a horizontal ladder that ends with a covered area. Mice tend to prefer to stay in the covered areas given the ability to move. The spaces between each rung are paced at 1 cm to make sure that the mouse can coordinate its movement without misstepping between the rungs. The number of missteps per each mouse was recorded, together with the total time to walk from one end of the ladder to the other.

All of the statistical analyses were done by GraphPad Prism software. The p values were calculated by a Pearson’s r test to measure linear correlation between two sets of data, a two-tailed unpaired Student t test, a two-tailed paired Student t test, and one-way ANOVA followed by a Tukey post hoc test as described in the figure legends. All p values ≤0.05 were considered significant.

Mouse studies conducted at the University of Alabama at Birmingham were approved by Institutional Animal Care and Use Committee, and all relevant ethics regulations were followed.

IBD is often associated with neurodegenerative disorders and neuropsychiatric manifestations such as depression, fatigue, and anxiety (1, 10, 46–48). To understand whether IBD drives brain inflammation, we investigated the development of brain inflammation in the adoptive T cell transfer model of chronic colitis in Rag1^−/− mice, which lack T and B cells. Although the brains of normal untransferred Rag1^−/− mice are devoid of CD4⁺ T cells, a substantial population of CD4⁺ T cells accumulates in the brain of colitic Rag1^−/− mice starting from the onset of colitis at 4 wk after CD45RB^hi CD4⁺ T cell transfer. At 8 wk posttransfer, when colitis becomes severe, ∼15% of the total live brain mononuclear cells were CD4⁺ T cells (Fig. 1A, 1B). Between 4 and 8 wk posttransfer, there was a >5-fold increase in the number of brain-infiltrating CD4⁺ T cells that was proportional to the severity of colitis (Fig. 1B, 1C). We found a significant correlation between the increase in the number of colonic CD4⁺ T cells and brain-infiltrating CD4⁺ T cells as severity of disease progressed over time (Fig. 1D). Among several organs analyzed in colitic mice, the number of CD4⁺ T cells increased by >20-fold only in the gut and in the brain during the peak phase of colitis (4–8 wk), whereas all other organs showed little increase (<2-fold) in the number of CD4⁺ T cells between the onset and development of severe colitis (Fig. 1E). This indicated that the number of CD4⁺ T cells in the brain increases proportionally to the expansion of effector CD4⁺ T cells in the gut. To understand whether there is an intrinsic defect in the BBB by Evans blue staining, we noted that BBB integrity is intact in precolitic Rag1^−/− mice at 2 wk posttransfer, suggesting that there is no intrinsic defect in the BBB permeability in precolitic mice. However, we noted a modest but significant disruption of the BBB permeability in colitic mice at 8 wk posttransfer compared with precolitic and normal mice (Fig. 1F, 1G). The data suggest that although there is no intrinsic defect in the BBB in precolitic mice, a modest disruption of the BBB occurs during the peak onset of colitis. To exclude the possibility that the observed phenomenon is not specific to colitis induction in lymphopenic Rag1^−/− mice, we analyzed the extent of CD4⁺ T cell infiltration with chronic DSS (cDSS)–induced colitis in normal lympho-sufficient B6 mice. Similar to the observation in the adoptive transfer model of colitis in Rag1^−/− mice, we found marked infiltration of CD3⁺CD4⁺ T cells in the brain of DSS-induced colitic mice where the frequency of brain-infiltrating CD3⁺CD4⁺ T cells was directly proportional to colitis severity that increased in each cycle of DSS treatment (Supplemental Fig. 1). After three cycles of DSS treatment, there was an ∼25-fold increase in the frequency of brain-infiltrating CD4⁺ T cells and an ∼300-fold increase in the total number of CD3⁺CD4⁺ T cells in the brain compared with vehicle-treated controls (0.3–0.5% CD4⁺ T cells, or absolute numbers of CD4⁺ T cells were ∼350 in age-matched normal wild-type [WT] mice). Taken together, our data indicated that CD4⁺ T cells infiltrate the brain of mice during colitis, and the number of infiltrating CD4⁺ T cells into the brain is proportional to progression in disease severity.

In behavioral neuroscience, quantitative responses to stressful environments such as those presented in the elevated plus maze (EPM), social interaction (sociability), and light/dark tests may be used to approximate assessment of emotional state. Aversion to stressful components of these tests has been suggested to evidence anxiodepressive-like behavior in rodents (49–52). To understand whether brain inflammation correlated with neurobehavioral abnormality, colitic mice were tested in these paradigms (53).

Colitic Rag1^−/− mice adoptively transferred with naive CD4⁺ T cells exhibited increased anxiodepressive-like behavior compared with controls in the EPM, light/dark test, and social interaction test (Fig. 2). Interestingly, despite these deficiencies, the performance of colitic mice was comparable to normal control mice (untransferred Rag1^−/−) with regard to learning and memory. Specifically, the ability to remember and identify novel objects was not affected in colitic mice, as evident from the novel object recognition test. Also, there was no effect on aversive memory formation or working memory as inducted by assessments in the fear conditioning and Y-maze tests, respectively. Similarly, the colitic Rag1^−/− mice did not show a significant difference in missteps compared with control mice in the horizontal ladder test, suggesting the absence of any defect in their motor function. Thus, colitic mice showed acute anxiety and depression-like behavior with no apparent deficits in cognition, working memory, spatial memory, cerebellum function, and muscle coordination. These data are consistent with the correlations between gastrointestinal inflammation and neuropsychiatric conditions observed in humans.

Dysregulation in the hypothalamus has been found to play a critical role in stress and depression during IBD (54). Targeting certain neurons in the hypothalamus rather than the whole brain potentially provides a more effective treatment for anxiety than conventional therapy (21). Neurobehavioral manifestations including anxiety and stress have been noted for many years in IBD patients (17, 46). As gastrointestinal disorders and HPA axis dysfunction are frequently observed in patients with major anxiety and depression, we analyzed the neurobehavioral outcome of the hypothalamus (55). Recently, it has been reported that the hippocampus neurogenesis is significantly affected in colitis. This alteration in neurodegeneration has been linked to anxiety-like behavior in a colitic mouse model (18). Thus, we specifically concentrated on these two main brain regions in our subsequent investigations.

As the behavioral studies suggested that the colitis impacted anxiodepressive-like parameters but not motor or memory functions, we focused on the limbic system circuitry controlling these emotions. Although the hippocampus can also contribute to depression and anxiety, as well as food intake, we found a substantial CD4⁺ T cell accumulation in the hypothalamus. However, a significantly lower number of CD4⁺ T cells migrated to the hippocampus compared with the hypothalamus (Fig. 3A, 3C). Interestingly, diffusion-weighted MRI of 10 wk posttransferred Rag1^−/− colitic mice revealed a small hyperintense signal in the hypothalamus, including the preoptic region near the anteroventral third ventricle (Fig. 3B). We could not detect the presence of CD4⁺ T cells in other areas of the brain investigated, including the thalamus (Fig. 3D and data not shown).

In the CNS, astrocytes respond rapidly to any type of neurologic insults through a process called astrogliosis by the upregulation of the intermediate filament GFAP indicating astroglial activation (56). As colitic Rag1^−/− mice showed a lesion-like structure in the hypothalamic region with accumulation of CD4⁺ T cells, we examined cross-sections of the brain of colitic mice for accumulation of reactive astrocytes at 8 wk posttransfer. In colitic mice, a marked expansion of the astrocytes occurred in the periventricular hypothalamic region, with pronounced upregulation of GFAP, cellular hypertrophy, and dispersed astrocyte proliferation, which are classical signs of ongoing inflammation in the brain but not in the hippocampus area (Fig. 3E, 3F). Astrocytes have long been related to demyelinating diseases of the CNS and as reactive components within and surrounding demyelinated lesions in MS (57). Analysis of the hypothalamic region of colitic mouse brain also revealed significant demyelination (Supplemental Fig. 2). The appearance of demyelination in human IBD patients is controversial. The vast majority of patients with UC present a form of neurobehavioral problems, including anxiety and depression (1–10). Nevertheless, it has been claimed that only 3% of patients suffering from IBD manifest visible demyelinating lesions during the course of the disease (11, 15). Additionally, it has been shown that 42–46% of IBD patients have small white matter lesions, compared with 16% in controls (1). Interestingly, astrogliosis and demyelination have been described in monkeys suffering from UC (13). Collectively, the colitic mice show a high degree of astrogliosis, demyelination, and lesion-like development in the hypothalamic area that are associated with neurobehavioral disorders. Therefore, it is likely that astrogliosis with signs of demyelination are critical steps in the initiation of T cell–mediated pathology in the CNS as well as indicative of ongoing inflammation in the brain during chronic colitis.

Both Th17 and Th1 cells are highly pathogenic effector cells in colitis and are comparable in their ability to induce colitis (58). Moreover, Th17 cells can cross the BBB more effectively than do Th1 cells, indicating that Th17 cells are more effective inducers of neuroinflammation (32). It has also been proposed that Th17 effector cells preferentially target astrocytes to promote neuroinflammation (33). Due to the presence of hypothalamic astrogliosis that correlated with the accumulation of CD4⁺ T cells in the brain of colitic mice, we analyzed the brain-infiltrating CD4⁺ T cells for expression of Th1, Th17, and Treg-specific transcription factors T-bet, RORγt, and Foxp3, respectively. CD4⁺ T cells retrieved from colonic LP of colitic Rag1^−/− mice expressed a high level of RORγt (>50%) with a moderate level of T-bet expression (<5%), indicating that RORγt-expressing CD4⁺ T cells were the overwhelmingly dominant population in the colon (Fig. 4A). Strikingly, similar to colonic CD4⁺ T cells, brain-derived CD4⁺ T cells of colitic mice also expressed a very high level of RORγt (>20%). However, CD4⁺ T cells examined from all other organs did not express RORγt (Fig. 4B, Supplemental Fig. 3). T-bet was only modestly expressed in the colon and brain CD4⁺ T cells (2–4%) and was lowly expressed in all of the other organs examined. However, Foxp3 expression was significantly higher in the colon and spleen and was nearly undetectable (<2%) in other organs, including in brain CD4⁺ T cells (Fig. 4A, 4B). This indicated that Foxp3⁺ Tregs were unable to migrate to the brain. Additionally, CD4⁺ T cells of all other organs such as spleen, liver, lung, and PLNs showed very low levels of RORγt and T-bet expression (Fig. 4B, Supplemental Fig. 3), indicating that only CD4⁺ T cells in the colonic LP and brain of colitic mice express a high level of RORγt and moderate level of T-bet. As RORγt and T-bet are the master transcription factors of Th17 and Th1 cells, respectively, which induce production of the proinflammatory cytokines IL-17 and IFN-γ, we analyzed their expression levels in different organs of colitic mice. In both IBD and MS, both Th17 cells and IFN-γ coproducing “Th1-like” Th17 cells are highly pathogenic (58, 59). However, the frequencies of IL-17 single-positive and IL-17/IFN-γ double-positive cells (∼5%) were highly induced in colonic CD4⁺ T cells. Similar to colon, there was a significant population of IL-17 single producers (>40%) and IL-17A/IFN-γ double producers in the brain (>12%) (Fig. 4C, 4D). However, the frequency of IFN-γ single-positive CD4⁺ T cells was significantly lower in the brain of colitic mice compared with the colon. As the frequency of pathogenic IL-17A/IFN-γ coexpressing CD4⁺ T cells was higher in the brain compared with the colon, it indicated that a significant proportion of brain-infiltrating Th17 cells transition to pathogenic IFN-γ–producing Th17 cells, a process that can occur independent of T-bet expression (59). In contrast, IL-17 and IFN-γ expression levels were undetectable from all other organs examined in colitic Rag1^−/− mice (Fig. 4C, 4D). Taken together, our findings show that during colitis, Th17 and Th1-like Th17 cells are the dominant helper CD4⁺ T cell population predominantly localized in the brain and in the colon, suggesting that proinflammatory characteristics of the LP-derived colonic CD4⁺ T cells are represented in the brain-infiltrating CD4⁺ T cells but absent from other major organs.

Because the brain-infiltrating CD4⁺ T cells overwhelmingly expressed RORγt, we investigated whether RORγt-deficient CD4⁺ T cells can selectively infiltrate into the CNS during colitis. In support of a previous finding, we found that adoptive transfer of RORγt^−/− CD45RB^hiCD4⁺ T cells transferred to Rag1^−/− mice caused mild colitis with a significantly lower inflammation score (Fig. 5A, 5B). RORγt is essential for induction of IL-17A from CD4⁺ T cells. We found that IL-17A expression from RORγt-deficient colonic CD4⁺ T cells was nearly absent in Rag1^−/− recipient mice (Fig. 5C, 5D). However, compared with WT colonic CD4⁺ T cells from reconstituted Rag1^−/− mice, IFN-γ was strongly induced from colonic RORγt-deficient CD4⁺ T cells isolated from Rag1^−/− mice that received RORγt-deficient naive CD4⁺ T cells. This supports a previous observation that both IL-17A single producers and IFN-γ/IL-17A coproducers are more pathogenic and cause greater colonic epithelial damage than do IFN-γ single producers (58). In contrast, the recovered CD4⁺ T cells from RORγt-deficient naive T cell recipient Rag1^−/− mice brain did not express either IL-17A or IFN-γ. RORγt-deficient naive T cells not only caused mild colitis, there was significantly lower numbers of CD4⁺ T cell infiltration (<10-fold) in the brain of Rag1^−/− recipient mice as detected both at 4 and 8 wk after adoptive transfer (Fig. 5E, 5F). To examine the induction of Th17-specific chemokines and cytokines in the brain of WT and RORγt-deficient naive T cell recipient Rag1^−/− mice at the mRNA level, we performed a comparative real-time RT-PCR of IL-17, IFN-γ, and CCR6 from sorted brain CD4⁺ T cells (Fig. 5G–I). Both CCR6 and IL-17 transcripts were highly elevated in WT brain-infiltrating CD4⁺ T cells with a significant increase in IFN-γ. However, we could not detect IL-17A or IFN-γ expression at the protein level in the brain RORγt-deficient CD4⁺ T cells by flow cytometry (data not shown). In contrast, there was marked enhancement of CCL20, the ligand of CCR6, from the pooled astrocyte and microglial population of WT brain relative to RORγt-deficient naive T cell recipient Rag1^−/− mice (Fig. 5J). In contrast to the brain, the number of colonic CD4⁺ T cells did not vary significantly in the colon or other organs such as spleen, lungs, PLNs, and liver between WT and RORγt-deficient CD4⁺ T cell recipient Rag1^−/− groups (Fig. 5K). Taken together, these data suggest that infiltration of CD4⁺ T cells in the brain is directly proportional to colitis severity and that colonic RORγt-expressing CD4⁺ T cells are essential for their homing from the colon to the brain during colitis.

To directly correlate the absence of colitis severity and low number of brain-infiltrating CD4⁺ T cells observed in RORγt^−/− naive CD4⁺ T cell recipient Rag1^−/− mice, we initially compared the extent of astrogliosis between WT and RORγt^−/− reconstituted groups. Accumulation of astrocytes in the hypothalamic region of the brain was significantly lower in Rag1^−/− recipient groups that received RORγt^−/− naive CD4⁺ T cells compared with WT (Fig. 6A). Moreover, the RORγt^−/− CD45RB^hi recipient Rag1^−/− mice showed no neurobehavioral abnormality compared with age-matched untransferred Rag1^−/− mice with comparable behaviors in the EPM, sociability, and light/dark tests, indicating the absence of anxiodepressive-like behavior (Fig. 6B). Accordingly, RORγt-expressing Th17 cells that are critical for aggravating colitis severity are also crucial for causing brain inflammation and neurobehavioral disorders.

As adoptively transferred RORγt^−/− naive CD4⁺ T cells caused mild to moderate colitis in Rag1^−/− mice without significant brain inflammation, we wanted to investigate the role of RORγt-expressing CD4⁺ T cells in promoting cDSS-induced colitis and colitis-induced brain inflammation in mice with an intact immune system. DSS-induced colitis is emerging as a reliable model to study effector and regulatory CD4⁺ T cells in lymphosufficient mice (60–63). Therefore, cDSS-induced colitis provides an excellent model for interrogating the effects of pathogenic Th1 or Th17 cells in the intestine of cDSS-treated mice that faithfully mimics human IBD. Induction of colitis by cDSS treatment induced severe colitis in both WT and Rorc^fl/fl Cd4-Cre mice (Supplemental Fig. 4). Unexpectedly, compared with WT mice with intact RORγt in the CD4⁺ T cell compartment, cDSS-induced colitis was significantly more severe in Rorc^fl/fl Cd4-Cre mice where RORγt is deleted from CD4⁺ T cells (Supplemental Fig. 4A, 4B). Rorc^fl/fl Cd4-Cre mice showed ∼6-fold higher induction of colonic IFN-γ⁺CD4⁺ T cells compared with WT colitic mice, whereas there was a 35-fold lower induction of the colonic IL-17⁺CD4⁺ T cell population (Supplemental Fig. 4C). Despite a higher degree of colonic inflammation compared with WT mice and a comparable number of colonic CD4⁺ T cells, a significantly lower frequency of CD4⁺ T cells infiltrated into the brain of colitic Rorc^fl/fl Cd4-Cre mice. In contrast, a substantial proportion of CD4⁺ T cells infiltrated the brain of colitic WT mice and expressed RORγt without significant induction of T-bet or Foxp3 (Supplemental Fig. 4D–F). These data further demonstrated that despite severe colonic inflammation in Rorc^fl/fl Cd4-Cre mice, a lower number of CD4⁺ T cells was able to infiltrate into the brain in the absence of RORγt. Accordingly, astrogliosis was more pronounced in the brain of WT mice (Supplemental Fig. 4G, 4H). The data strongly suggested that although severity of colitis is associated with brain inflammation, the inflammation is largely mediated by colitogenic RORγt-expressing Th17 cells, probably because of their greater ability to cross the BBB.

IBD patients suffer from several neuropathological comorbidities (1–9), but the mechanisms behind these neurologic manifestations during IBD remain largely unclear. Despite a growing field of research on the gut/brain axis, little is known about the mechanism by which gut pathology affects the brain. Our study found an intriguing mechanism that may directly link chronic intestinal inflammation to brain inflammation. In this study, we identified a cellular mechanism by which pathogenic RORγt-expressing CD4⁺ T cells are correlated with the pathogenesis of colitis-driven neuropathology. These findings may help to explain how gut pathology can directly influence brain inflammation leading to neurologic dysfunctions in IBD patients.

The brain/gut axis is a complex bidirectional system comprised of multiple connections between the nervous system and the gastrointestinal tract. It is believed that in IBD patients, psychological stress via the HPA axis increases intestinal motility and permeability, leading to a local inflammatory response in the gastrointestinal tract (22, 23). However, there is considerable debate on the “cause-and-effect” phenomenon between IBD and neurologic comorbidity (10, 22). Evidence from preclinical studies suggests that intestinal inflammation can influence psychological behavior and brain activity (22). Studies of animal models with colonic inflammation have shown region-specific changes in the CNS that correlated with expressions of inflammatory genes in the brain (25–27). In support of this, a study has shown that CD3⁺CD4⁺ T lymphocytes accumulate in a granuloma-like lesions in the brain of an IBD patient, suggesting that CD4⁺ T lymphocytes infiltrate the brain during IBD and cause inflammation (64).

Classically, the brain and its associated structures have been considered to be immune privileged (30). However, advances during the past decade have led to a reassessment of this assumption. New studies demonstrate the roles of a broad range of immune cells, including innate lymphoid cells and T cells, in regulating inflammation and neurologic diseases in the brain (30). Researchers recently discovered infiltration of T cells in the aging brain (65). Intriguingly, a pathogenic Th17 subset of CD4⁺ T cells found in the brain of MS patients is also found in the gastrointestinal lesions of IBD patients, making Th17 cells a common therapeutic target in both of these forms of autoimmune diseases (66–68). Th17 cells have been implicated in selective infiltration of the BBB and in causing neuroinflammation through the triggering of astrocyte activation (32–34). Several neurologic disorders associated with cognitive impairment have been linked to Th17 cells, including neurovascular disorders and neurodegenerative diseases. Besides MS, Th17 cells are linked to various brain diseases, including demyelination, dementia, ischemic brain injury, Parkinson’s disease, and Alzheimer’s disease (11, 14, 15, 34). However, little is known about the mechanism by which Th17 cells enter the brain to induce brain inflammation that might include either direct tissue destructive effects of IL-17 and IL-17/IFN-γ coproducers on brain cells or indirect effects mediated through neurovascular dysfunction. A study has shown that due to the presence of IL-17R on endothelial cells (ECs) of the BBB, Th17 cells transmigrate through the BBB with greater efficiency than Th1 cells to promote CNS inflammation during MS (32). Interestingly, IBD patients also suffer from MS-like demyelinating conditions (11, 13, 15, 69). Therefore, it is likely that pathogenic CD4⁺ Th17 cells that drive gut inflammation also contribute to inflammation of the brain during IBD.

Our study demonstrated that pathogenic Th17 cells, marked by the expression of their master transcription factor RORγt, accumulate in the brain during colitis and cause brain inflammation. In view of the evidence suggesting that pathogenic CD4⁺ Th17 cells traffic to the brain of colitic mice and trigger inflammation in the hypothalamus, we believe that neuropathological comorbidity in IBD patients may involve inflammation of the hypothalamus, which is one of the major regions regulating anxiety and depression (19–21). It is noteworthy that brain is usually well protected from uncontrolled influx of molecules from the periphery due to the restriction imposed by the BBB, a physical seal of cells lining the blood vessel walls. However, the BBB around the hypothalamus, which is located at the base of the brain, is a notable exception to this rule, as this region is surrounded by “leaky” blood vessels that might readily allow entry of the CD4⁺ T cells into the CNS (70, 71). Even though we found that the BBB integrity is modestly disrupted during the peak phase of colitis, whether it is a consequence of brain inflammation by the entry of pathogenic CD4⁺ T cells remains to be determined. It is possible that at a more advanced stage of disease progression there could be a greater disruption or increased permeability of the BBB. Despite our observation that lymphocytes accumulate in the hypothalamus, it is quite possible that the invading CD4⁺ T cells might eventually home to other regions of the brain at later stages of the disease and cause additional inflammatory changes. However, due to the limited duration of the progress of the disease in the mouse model of chronic colitis, it is difficult to validate the subsequent localization of CD4⁺ T cells in other parts of the brain.

We have not demonstrated how RORγt-expressing Th17 cells selectively enter the brain during colitis. Although the extent of CD4⁺ T cell infiltration and inflammatory changes in the brain is directly proportional to colitis severity, our data suggest that RORγt-expressing Th17 cells more efficiently cross the BBB compared with Th1 cells and are a prerequisite for efficient entry into the CNS. This was further highlighted by the observation that although colitis was modest after the adoptive transfer of RORγt-deficient T cells in lymphopenic recipient mice, which correlated with reduced brain infiltration and neuroinflammation, brain of colitic Rorc^fl/fl Cd4-Cre mice induced by cDSS showed a comparable lower frequency of CD4⁺ T cell infiltrates despite a more severe onset of colitis compared with WT mice. Intriguingly, in both of the models the number of colonic CD4⁺ T cells was comparable in the absence of RORγt in CD4⁺ T cells with significantly higher induction of IFN-γ–producing Th1 cells. The difference in the extent of disease severity in two different models of colitis in the absence of RORγt could be attributed to a difference in the mechanism of disease induction and/or a higher proinflammatory potential of Th1 cells in cDSS-induced colitis. Nevertheless, our data indicate that colitogenic RORγt-expressing Th17 cells causing intestinal inflammation are a critical population that promotes neuroinflammation. However, the brain-infiltrating potential of Th1 or Th1-like Th17 cells will require further investigation, preferably using a Th1 transfer model of colitis or by inducing chronic colitis in the T-bet–deficient mouse model.

It is possible that RORγt plays an intrinsic role in either regulating the expression of a specific adhesion molecule on Th17 cells that interacts with the ECs of the BBB or by interacting with IL-17R of the ECs via the production of IL-17 to facilitate their attachment and migration through the BBB. Despite our finding that the Th17-associated chemokine receptor CCR6 and its ligand CCL20 are highly expressed in the brain of colitic mice with intact RORγt in CD4⁺ T cells, we cannot rule out the possibility that Th17 cells could be preferentially recruited to the brain due to the production of specific chemokines or that Th17 cells may preferentially proliferate and/or expand at this site. Additionally, due to the inherent plasticity of Th17 cells, it is possible that they gradually become more pathogenic by increasingly transitioning to Th1-like IFN-γ coproducing Th17 cells at later time points. Why the infiltrating Th17 cells continue to actively produce the proinflammatory cytokines after their infiltration in the brain should be an area of active investigation to determine whether specific T cell clones are enriched in the brain by comparing their heterogeneity with the TCR clones of the colon during colitis. Identification of Th17 subsets with an overlapping TCR repertoire will suggest that specific microbial Ags in the colon trigger an aberrant T cell response leading to their expansion and subsequent migration to the brain where they cross-react with overlapping brain Ags, leading to their further expansion and retention of inflammatory potential.

This study suggests that the gut/brain immune axis is one of the principal corridors by which IBD drives the development of brain inflammation and behavioral neuropathology via migration of proinflammatory Th17 cells. We believe that the identification of specific molecules on Th17 cells that enable their infiltration into the brain by interacting with cognate receptors with the ECs of the BBB could lead to a novel therapeutic intervention that might be able to enhance the prognosis of both colitis and colitis-associated neuropathology.

The authors have no financial conflicts of interest.