When the Gut Changes, the Brain Does Not Read the Diagnosis: What One Pediatric Autism/ADHD Case Taught Me About Phenols, Mucosal Immunity, FMT, Streptococcus, and Metabolic State

I have been working through a pediatric case that has made me think deeply about the way we approach children with autism, ADHD, childhood apraxia of speech, PANS-like symptoms, gastrointestinal dysfunction, histamine intolerance, unusual immune findings and what is often loosely described as mitochondrial dysfunction. I am deliberately removing all identifying information because the purpose of sharing this is not to discuss the child personally. What interests me is the biology that becomes visible when years of neurological findings, metabolic testing, stool studies, infections, nutritional data and treatment responses are placed on the same timeline and viewed as one connected system.

What became clear to me is that the diagnosis remained relatively stable while the biology surrounding it changed dramatically. This child had apparently typical early development followed by a clear developmental regression. Previously acquired language disappeared, facial expression changed, eye contact changed and spontaneous imitation decreased. Autism was subsequently diagnosed, ADHD was diagnosed later, and over time the language problem became recognizable as a much more specific motor planning disorder consistent with childhood apraxia of speech. Years later, however, a very different layer appeared. Repetitive behavior became prominent. There were rigid routines, unusual fears, vocalizations, tic-like movements, worsening food restriction, severe emotional dysregulation and eventually significant self-injury. At other times, many of these problems almost disappeared. That is where I think an important distinction has to be made. A neurodevelopmental condition may represent part of the underlying architecture of the nervous system, while metabolic, immunological, microbial and environmental factors can change how strongly that architecture is expressed at any particular time.

The first major clue came from comparing metabolic data obtained during very different clinical states. During one of the more difficult periods, an organic acids profile showed quinolinic acid at 16.98 mg/g creatinine with an upper reference limit of 5.50. L-kynurenine was 0.86 against an upper limit of 0.60, while the kynurenic acid to kynurenine ratio was only 2.12 when the laboratory expected a value above 4.00. Pyroglutamic acid was 77.95 against an upper reference limit of 27.60. Succinate was 13.20 against 2.10, fumarate was 1.41 against 0.70, malate was 2.44 against 1.90 and alpha ketoglutarate was 48.32 against 33.00. Several branched chain keto acids were elevated as well. Read as a whole, this did not look like one isolated abnormal metabolite. It looked like a period in which inflammatory metabolism, glutathione-related redox handling and mitochondrial intermediary metabolism were all under substantially greater pressure.

The kynurenine pathway immediately caught my attention. Tryptophan is often discussed only as the precursor for serotonin, but serotonin synthesis represents only one possible fate of tryptophan. During inflammatory signaling, more tryptophan can be redirected through the kynurenine pathway. Enzymes such as IDO can become more active, and downstream metabolism can shift toward different branches. One branch contributes to kynurenic acid production, while another can move toward quinolinic acid. Quinolinic acid is biologically interesting because it interacts with NMDA receptor signaling and participates in neuroexcitatory and inflammatory biology. A urinary measurement cannot tell us the concentration of quinolinic acid around neurons in the brain, so I would never make that leap. But when a child is in a severe neurobehavioral state and quinolinic acid is approximately three times the laboratory upper limit while kynurenine is also elevated and the kynurenic to kynurenine ratio is low, the pattern deserves attention. The useful question becomes what was pushing tryptophan metabolism in that direction. Was there inflammatory signaling, microbial antigen exposure, oxidative stress, nutritional limitation, infection, or several of these processes acting at the same time?

What made this much more compelling was what happened later. Several years afterward, during a period when the child was functioning substantially better, quinolinic acid was 4.9, kynurenic acid was 0.97, pyroglutamate was 36 and the major TCA cycle intermediates were within that laboratory’s reference ranges. Lactate and pyruvate were normal. The fatty acid oxidation markers were unremarkable. The neurodevelopmental diagnosis had not disappeared, but the metabolic state surrounding it had changed dramatically. That is one of the central observations in the case. It suggests that we need to distinguish biological traits that remain relatively stable from biological states that can change considerably. Autism, ADHD and apraxia may form part of the underlying neurodevelopmental architecture, while inflammatory metabolism, oxidative stress, microbial fermentation, histamine reactivity and severe behavioral dysregulation can rise and fall on top of that architecture.

The microbial metabolites from the earlier profile made the picture even more interesting. P-cresol sulfate was 537.4 with an upper reference limit of 127. Trimethylamine was 712 with an upper limit of 200. TMAO was 150 against 70. Hippuric acid was 759.26 against 363, and another aromatic microbial metabolite, 4-hydroxybenzoic acid, was also elevated. The p-cresol finding was particularly striking. P-cresol can be generated through microbial metabolism of aromatic amino acids, especially tyrosine. When substantial protein substrate reaches the distal intestine, certain organisms can increasingly use amino acids for fermentation rather than relying primarily on carbohydrates. Tyrosine metabolism can then generate phenolic compounds including p-cresol. Once absorbed, p-cresol has to be processed by the host, and a substantial portion becomes conjugated into compounds such as p-cresol sulfate before elimination.

A value of 537.4 against an upper limit of 127 does not mean p-cresol caused autism. That interpretation would go far beyond what the data support. What it tells me is that microbial processing of aromatic amino acids was unusually active during that particular biological state. That becomes much more interesting when it occurs beside abnormal kynurenine metabolism, glutathione-related stress and multiple abnormal TCA cycle intermediates. Instead of asking whether one metabolite explains the developmental diagnosis, the more useful question is whether a changing microbial metabolic environment can modify neuronal threshold, immune signaling, redox balance and behavior in a nervous system that is already developmentally vulnerable.

This is where protein digestion becomes much more important than the simple word dysbiosis. Protein digestion begins long before the colon. Gastric acid denatures dietary protein and supports activation of pepsin. Pancreatic proteases continue breaking larger proteins into smaller peptides. Brush border enzymes complete much of that process, and amino acids and small peptides are absorbed through the small intestine. If that upstream processing is inefficient, more nitrogen-containing substrate can reach the colon. Then the microbiota inherit the problem. Certain organisms can increasingly ferment amino acids and generate phenols, indoles, ammonia, amines and sulfur-containing metabolites. What looks like a microbial problem may therefore begin farther upstream with host digestive capacity.

That is why I am cautious when an abnormal stool test immediately leads to a discussion about which bacterium needs to be killed. Sometimes the more useful question is why the substrate reached that bacterium in the first place. Was gastric digestion adequate? Was pepsin activity adequate? Was pancreatic digestion intact? Was transit abnormal? Was food diversity too narrow? Was there enough fermentable carbohydrate available to support saccharolytic metabolism? Was the bile acid environment selecting for particular organisms? Microbial metabolism cannot be separated from host physiology.

The trimethylamine findings illustrate the same principle. Intestinal microorganisms can generate trimethylamine from substrates such as choline and carnitine. The liver then converts much of that trimethylamine into TMAO through flavin-containing monooxygenases. Trimethylamine was 712 when the laboratory’s upper limit was 200, and TMAO was 150 against 70. These measurements do not tell us how much of either molecule was present inside the brain, but they tell us that microbial substrate utilization and host hepatic metabolism were behaving very differently during that period. When those findings occur beside high quinolinic acid, high p-cresol sulfate, elevated pyroglutamate and abnormal TCA intermediates, they begin to describe a biological state rather than an isolated laboratory abnormality.

The longitudinal comparison is what made this convincing to me. During the more difficult state, p-cresol sulfate was 537.4, quinolinic acid was 16.98, trimethylamine was 712, TMAO was 150 and pyroglutamate was 77.95, with several abnormal TCA intermediates. During a later and clinically much better period, quinolinic acid was 4.9, 4-cresol was 59 and within range, hippuric acid was 241, pyroglutamate was within range and the TCA cycle pattern had normalized. The child was clinically better at the same time. That does not prove causation, but longitudinal biological change is much more informative than one abnormal result taken in isolation.

The stool data told a similarly dynamic story. An earlier gastrointestinal study showed secretory IgA at only 210 µg/g, below the laboratory range. Akkermansia was present. Prevotella was present. Methanogenic organisms were present. Lactobacillus was present. Several years later, the ecosystem looked almost completely different. Secretory IgA was greater than 7,500 µg/mL and had exceeded the upper limit of the assay. Calprotectin was mildly elevated at 62.3 µg/g. Akkermansia was no longer detected. Prevotella was no longer detected. Methanogenic organisms were absent. Bifidobacterium represented approximately 17.2 percent of the reported community and Faecalibacterium prausnitzii approximately 21.2 percent, while Lactobacillus was only 0.001 percent.

The earlier and later stool tests used different analytical technologies, so the numerical values cannot be compared as if they were generated by the same method. But the ecological transition is still remarkable. The intestinal immune environment moved from low secretory IgA output to an extraordinarily high output. Organisms that were once clearly present disappeared. Other organisms became dominant. The later microbiome was not simply depleted. It had become highly selective.

This is where the idea of good bacteria and bad bacteria becomes inadequate. Bifidobacterium is generally considered beneficial. Faecalibacterium is generally considered beneficial. Yet together they represented almost 38 percent of the reported relative community. That does not make either organism harmful. It means abundance alone is not enough to understand function. Microbes live through metabolic relationships. Bifidobacteria commonly ferment carbohydrates and generate acetate and lactate. Faecalibacterium can use acetate and other cross-fed metabolites and generate butyrate. In the later stool profile, predicted lactate-producing capacity was 17.2 percent and predicted butyrate-producing capacity was 26.7 percent. Those findings can coexist if the community is operating through cross-feeding. Bifidobacterium may be generating acetate and lactate while Faecalibacterium consumes part of that substrate and produces butyrate. The stool pH was 6.5, which also makes severe accumulation of unconsumed lactate less obvious.

The disappearance of Akkermansia became equally interesting. Akkermansia occupies the intestinal mucus niche and uses mucin-associated glycans as substrates. In this child, Akkermansia became undetectable and predicted mucin degradation also fell to zero. It would be easy to say the child simply lacked mucus, but mucosal ecology is much more complicated. Akkermansia depends on the structure and chemistry of mucus, including the glycan architecture of MUC2 and how those glycans are fucosylated, sialylated and sulfated. Mucus turnover matters. Diet matters. Oxygen tension matters. Bile acids matter. Other microbes matter. Bacteriophages matter. Secretory IgA targeting matters. This child’s intestine had supported Akkermansia previously and later no longer did. To me, the interesting question is not simply how to add Akkermansia back. It is what changed in the host environment that made the niche no longer supportive.

The secretory IgA finding may be part of that story. Going from a low secretory IgA of approximately 210 to a value above 7,500 represents an enormous transition in mucosal immune activity. Secretory IgA is one of the main interfaces between the immune system and the microbiome. It can coat microorganisms, alter adherence, agglutinate them, influence replication and exert highly selective pressure against particular strains. This became especially interesting because the child underwent fecal microbiota transplantation before the later stool analysis.

The FMT involved oral donor material followed by rectal administration. The child was already unstable before the procedure, but afterward the family saw a major increase in agitation, histamine-type symptoms and self-injurious behavior. I do not think simply calling this a die-off explains very much. FMT is not the replacement of one microbiome with another. It is an ecological invasion. Donor organisms enter a recipient environment that already has its own microbes, mucus chemistry, bile acids, antimicrobial peptides, bacteriophages, oxygen gradient, nutrient availability and immune system. Some organisms establish. Others disappear. Some resident organisms expand when competition changes. Some incoming strains may become heavily coated with IgA and fail to persist. When extremely high secretory IgA appears alongside major expansion of certain organisms and disappearance of others, selective ecological filtering becomes much more interesting than the idea of complete donor engraftment.

The sulfur side of the microbiome added another layer. Bilophila wadsworthia was elevated at approximately 0.553 percent. Bilophila is interesting because it sits at the intersection of bile acid metabolism, taurine and hydrogen sulfide production. The liver produces primary bile acids from cholesterol and conjugates many of them with taurine or glycine. Microbial bile salt hydrolases can remove those conjugated groups. When taurine-derived sulfur becomes available, organisms such as Bilophila can metabolize it and ultimately generate hydrogen sulfide.

Hydrogen sulfide is not simply a toxin. At physiological concentrations it has important signaling functions. The problem is concentration and location. At sufficiently high local exposure, hydrogen sulfide can interfere with mitochondrial respiratory enzymes including cytochrome c oxidase, also known as Complex IV. This becomes particularly important in colonocytes because these cells normally consume large amounts of butyrate and oxygen. Their mitochondrial metabolism helps maintain the low oxygen environment of the colon. If excessive sulfide suppresses colonocyte mitochondrial respiration, butyrate oxidation can decline and epithelial oxygen consumption can change. The oxygen gradient at the intestinal surface can then shift and create new selective pressures on the microbial community. Microbial sulfur metabolism can therefore alter host epithelial metabolism, while altered host metabolism can then reshape the microbiome. The direction of influence goes both ways.

Even the high Bifidobacterium abundance becomes more interesting when viewed through this pathway. Some Bifidobacterium strains possess bile salt hydrolase activity. Theoretically, one function of a highly abundant Bifidobacterium population could participate in deconjugating taurine-containing bile acids and increasing taurine-derived substrate available to sulfur-metabolizing organisms. At the same time, another function of that same broader population could generate acetate and lactate that support Faecalibacterium and butyrate production. The same microbial ecosystem can therefore support potentially favorable and potentially unfavorable pathways simultaneously. What matters is not whether a genus has a good reputation. What matters is metabolic flux.

The immune history then added a completely different dimension. During one period, the child developed repetitive behavior, rigid routines, severe fears, food restriction, vocalizations and tic-like activity. Later, after a relatively stable period, severe self-hitting abruptly returned at approximately the same time as a documented Group A Streptococcus infection. The tonsillar area was markedly enlarged and the Strep test was positive. Subsequent ASO was 299 IU/mL against a reference below 200, and anti-DNase B was 266 against a reference below 240. An earlier ASO had already been 320. This does not automatically prove PANDAS, but it creates a legitimate neuroimmune question.

Streptococcal immune responses can involve adaptive immunity, Th17 pathways and, in susceptible individuals, potentially neurovascular and cortico-striatal signaling. The basal ganglia are particularly interesting because they participate in habit formation, motor sequencing, tic expression, behavioral selection and dopaminergic regulation. They also interact with circuits involved in speech motor programming. What made the infection history even more useful was what happened shortly afterward. The child developed extensive chickenpox. The infection was visually dramatic, yet there was no comparable behavioral collapse. Mood remained relatively good, severe self-injury did not reappear in the same way and physical activity was reduced only briefly.

That is an important natural experiment. If the mechanism were simply that any infection creates inflammation and inflammation destabilizes the brain, then a substantial viral infection might have been expected to reproduce the same phenotype. It did not. That makes pathogen-specific adaptive immunity, different cytokine patterns, molecular mimicry, infection location or simple coincidence more interesting possibilities than a generic inflammation model.

The neurological history also became much more informative after the original EEG was recovered. At approximately two years and eight months, the child underwent a wake and sleep EEG. It showed persistent left-central spike and sharp-wave discharges with occasional generalization during the transition into sleep. The EEG was classified as pathological, grade III. A possible Rolandic or centrotemporal focus was considered. There was no ESES or CSWS, and no clinical seizure was captured.

This finding matters because the child’s remaining language difficulty today is increasingly motoric rather than simply linguistic. He now speaks in complete sentences. His vocabulary has expanded substantially and social motivation has increased. Yet speech remains difficult to execute. Intelligibility is reduced, speech can become excessively fast, prosody is atypical and some voluntary facial movements remain effortful. Speech production depends on a large neural network involving auditory representation, phonological planning, premotor programming, supplementary motor areas, basal ganglia, cerebellar timing and corticobulbar execution. A left-central epileptiform focus does not prove that the EEG abnormality caused the apraxia, but it is objective neurophysiological information and deserves to be treated as its own domain rather than forcing the speech problem into a gut or mitochondrial explanation.

The anemia provided another lesson in how longitudinal physiology can change the question. Earlier in the case, iron depletion was obvious. Ferritin was 16 µg/L and transferrin saturation 11.6 percent. Over time, the iron indices improved. By September 2026, serum iron was 125 µg/dL, transferrin saturation was 31 percent and ferritin was 36 µg/L. Yet the red blood cell count remained low at 3.71 T/L, hemoglobin was 11.2 g/dL and hematocrit was 31.2 percent with an MCV of 84.1. At that point, continuing to assume iron deficiency was the entire explanation no longer made sense.

The physiological question changes from asking how much iron the child is receiving to asking whether the bone marrow is producing enough new red cells or whether red cells are being lost or cleared prematurely. That is why something as simple as a reticulocyte count can sometimes be more informative than another complex test. If reticulocytes are low or inappropriately normal, production becomes the problem. If reticulocytes are elevated, loss or shortened red cell survival becomes more likely. That directs the next stage toward blood film, haptoglobin, bilirubin, LDH, G6PD, hemoglobin traits, occult blood loss and other targeted possibilities.

The nutritional data reinforce the same principle. There had been historical iron depletion, low vitamin D, mildly low zinc and later mildly low selenium. A urinary vitamin C marker had also been extremely low at one point. Vitamin B6, however, later measured 47.5 µg/L against an upper reference limit of 30. That is exactly why I resist supplementing biochemical pathways simply because I know what cofactors an enzyme uses. Yes, glutamate decarboxylase requires B6. Yes, mitochondrial enzymes require B vitamins. Yes, neurotransmitter synthesis depends on several micronutrients. But if B6 is already elevated, adding more P5P simply because the pathway uses it is not precision. It is ignoring the child in front of us.

The same principle applies to probiotics. If Bifidobacterium already represents around 17 percent of the measured ecosystem, automatically adding more Bifidobacterium because it is usually considered beneficial may not be the most intelligent intervention. If predicted butyrate production is already high, automatically adding more butyrate may not address the actual bottleneck. If Akkermansia is absent because the mucus niche is no longer supportive, simply adding Akkermansia may accomplish very little. If hydrogen sulfide is interfering with colonocyte oxidation, increasing butyrate substrate may not solve the problem if the epithelial cell cannot efficiently use what is already being produced.

This is what I mean when I talk about Host Capacity. The question is not simply what compound can be added. The question is whether the host is capable of using it. Can the stomach digest the meal? Can the small intestine absorb the nutrients? Can the mucosal immune system tolerate the microbial community? Can colonocytes maintain the metabolic environment necessary to support obligate anaerobes? Can the liver process microbial metabolites? Can red blood cells maintain adequate antioxidant protection? Can the child obtain sufficient iron, selenium, zinc, vitamin C and other nutrients without creating another imbalance? Is sleep being disrupted by snoring or airway obstruction? Is cortical hyperexcitability still present? Is the motor speech system receiving the type and intensity of therapy required for motor learning? Those questions often come before another supplement.

This case also contained a long history of interventions, and the responses were remarkably inconsistent. One antimicrobial protocol was tolerated while another produced substantial gastrointestinal deterioration. A very small amount of inulin caused a dramatic reaction during an earlier period. FMT was followed by severe worsening. Cryptolepis coincided with a rapid disappearance of self-injurious behavior. Hyperbaric oxygen therapy preceded a period of deterioration. Repeated rectal ozone was eventually followed by greater irritability and emotional dysregulation.

None of those observations proves the mechanism for which the intervention was originally prescribed. A reaction to an antimicrobial is not automatically a Herxheimer reaction. A response to an herb does not prove that the organism it was intended to target was responsible for the symptoms. A worsening after FMT does not prove that every donor organism was harmful. But together these observations tell us that the system was strongly state dependent and unusually sensitive to biological perturbation.

When a biological system behaves that way, adding five interventions at the same time makes the case almost impossible to read. Sometimes the most sophisticated intervention is not another compound. It is reducing variables and restoring enough physiological stability that each intervention becomes interpretable.

What continues to fascinate me is how much better the child is now. He speaks in complete sentences. His vocabulary has expanded. His facial expression has improved. He seeks out other children. He spends hours outside. He rides a bicycle and scooter, swims for long periods and has begun playing football with peers. Severe self-injury is no longer dominating the picture. The previous intense fears have largely disappeared. The dramatic histamine-type episodes and recurrent nosebleeds have subsided. Gastrointestinal function is substantially better.

Yet some objective abnormalities remain. Mild anemia remains. Eosinophilia remains intermittently present. Some developmental limitations remain. The motor speech disorder remains. That discrepancy teaches us something important. If a biomarker remains abnormal while the child becomes dramatically better, that biomarker cannot by itself explain the severe earlier state. If a biochemical pathway moves toward normal at the same time that clinical function improves, that pathway becomes more interesting. If one infection coincides with neurobehavioral deterioration while another major infection does not, immune specificity becomes more interesting. If the microbiome changes dramatically after FMT while mucosal IgA changes from very low to beyond the assay ceiling, host microbial selection becomes more interesting than the simple idea of replacing a microbiome.

This is why longitudinal analysis matters so much. One test tells us what was happening on one day. A timeline tells us how the system behaves. When I look at complex autism, ADHD, PANS-like, histamine and gastrointestinal cases now, I am increasingly less interested in finding one master explanation. I am interested in reconstructing the sequence. What appeared first? What appeared later? What changed immediately before a regression? What remained abnormal during recovery? What normalized when behavior improved? Which interventions changed the state? Which interventions made it worse? Which findings belong to the child’s underlying developmental architecture and which represent temporary physiological pressure placed on that architecture?

That is a very different way of approaching these cases. It does not reduce autism to the gut. It does not reduce ADHD to dopamine. It does not reduce PANS to one antibody. It does not reduce dysbiosis to one bacterium. It does not reduce mitochondrial biology to one supplement. It asks how the systems interact.

A neurodevelopmentally vulnerable brain can still be influenced by sleep, infection, intestinal metabolites, immune signaling, nutrition, oxidative stress and cortical excitability. A microbiome can be influenced by gastric digestion, bile acids, mucus glycans, food diversity, oxygen gradients and immune selection. An immune system can respond very differently to different organisms. Mitochondrial stress can be secondary and reversible rather than genetically fixed. A biochemical pathway can be severely abnormal during one period and return toward normal several years later.

That is the part of this work I have become increasingly interested in. Not simply identifying what is abnormal, but understanding why it became abnormal, what sits upstream of it, what keeps it active, what makes it disappear and whether changing it actually changes the person in front of us.