The NAG/AG Appetite Subgroup HypothesisTwo immune pathways, two mechanisms, one convergent endpoint — why appetite phenotype may be a readout of which cytokine dominates the cascade

A bifurcation nobody stratified on

Efforts to find reproducible ASD subgroups have pursued regression history, GI comorbidity, immune markers, and copy-number variants — with limited replication across cohorts. One candidate variable has gone largely unexamined: appetite. It is directly observable, reliably reported by caregivers without specialist training, and doesn't require neuropsychological assessment — yet the field has generally filed appetite differences under sensory processing or food selectivity rather than asking whether a deeper immunological divergence sits underneath.

More than fifteen years of longitudinal clinical observation in community ASD support settings reveals a consistent, previously unexplained split: roughly half of individuals show persistent limited appetite and apparent absence of hunger drive; the other half show hyperphagia — constant food-seeking, inability to recognize satiety, and tendency toward overweight. These are not two ends of a continuum; they present as discrete poles. And the appetite split doesn't travel alone — the hyperphagic group tends toward better motor coordination, more normal gait, and stronger social engagement, and does not present the characteristic under-methylation phenotype seen in the food-refusing group.

The somatostatin polarity framework

Somatostatin is encoded by a single gene but processed into two bioactive forms with opposite territories: SST-28, predominating in the gut as a satiety signal, and SST-14, predominating in cortical interneurons as the pacemaker of gamma oscillation — the rhythm underlying social cognition, sensory gating, and executive function.

In the established IDA cascade, gut pH failure drives pepsin/DPP-IV disruption, producing opioid peptide excess and LPS translocation. LPS activates IDO1 and NF-κB; NF-κB sequesters the transcriptional co-activator CBP away from CREB, silencing SST-14. In parallel, gut SST-28 activity shifts — and here is the hinge of the whole framework: brain and peripheral somatostatin signaling exert opposite effects on circulating ghrelin via the SST2 receptor subtype. Central SST2 activation increases ghrelin; peripheral SST2 activation suppresses it. The direction of SST-28 dysregulation in the gut therefore determines appetite phenotype directly — hyperactive SST-28 suppresses ghrelin (no hunger signal, NAG), suppressed SST-28 lets ghrelin run high (insatiable hunger, AG).

The immune fork — convergent NF-κB, divergent architecture

Both pathways elevate NF-κB and both silence SST-14 through CBP sequestration from CREB — in that narrow sense they converge. But the routes to NF-κB elevation are structurally different in ways that produce profoundly different cascade persistence, metabolic consequences, and treatment responsiveness.

This distinction is clinically load-bearing, not academic. A locked cascade requires breaking the AhR/IDO1 feedback loop as a prerequisite to functional recovery — even after immune clearance. A dynamic cascade resolves more directly when the upstream immune signal is addressed. This predicts a faster, more complete IMIG response in AG, and the need for adjunctive IDO1-pathway intervention in NAG to reach an equivalent outcome.

The cAMP and adenylyl cyclase axis

The locked-vs-dynamic distinction becomes most precise at the level of adenylyl cyclase. Two Gαi-coupled inputs converge on its suppression: casomorphin/gliadorphin opioid peptides at mu-opioid receptors, and adenosine accumulation (from CD26/DPP-IV blockade by the same peptides) at A1/A2A receptors. Both inputs together substantially collapse cAMP production.

The amplitude of the opioid-peptide input depends on the severity of gut pH failure. In NAG, full gut pH dysregulation produces maximum opioid peptide excess — the Gαi input is fully engaged, adenylyl cyclase is substantially suppressed, and cAMP is effectively collapsed: PKA cannot activate CREB, and SST-14 transcription is near-zero. In AG, gut pH failure is proposed to be less complete — a lower opioid burden partially engages the Gαi input without saturating it, so adenylyl cyclase retains residual activity, cAMP is reduced but not collapsed, and SST-14 transcription is dampened but present. This is the molecular basis for AG's better motor coordination and stronger social engagement — not a separate phenomenon, but a direct expression of partially preserved cAMP/PKA/CREB/SST-14 signaling.

The forskolin prediction

Forskolin bypasses both Gαi inputs by activating the adenylyl cyclase catalytic subunit directly. In NAG, where adenylyl cyclase is maximally suppressed, forskolin provides a large gain from a deeply depressed baseline. In AG, where partial activity already exists, the same intervention produces a smaller delta from a higher starting point. This predicts forskolin produces a larger observable effect in NAG than in AG — a testable, clinically observable subgroup indicator available before the full biomarker panel.

The estrogen compensatory axis

A parallel route to SST-14 transcription — Gq-mER → Gαq → PLC → DAG → PKCδ → AC7 → cAMP → PKA → CREB — bypasses the Gαi block entirely and is proposed to explain the roughly 4:1 male-to-female ASD ratio (females carry a partial compensatory mechanism males lack). This bypass is predicted to be more effective in AG than NAG: against a partial Gαi block, the AC7 contribution restores a meaningful fraction of total cAMP; against a near-complete block, it represents a smaller fraction of what's been lost. This predicts the male-to-female ratio within AG will run closer to 2:1–3:1 rather than 4:1 — a checkable demographic prediction against existing appetite-stratified epidemiological data.

Differential hormonal profiles & the methylation intersection

The clinical observation that AG individuals lack NAG's characteristic under-methylation phenotype is explained by mechanistically different routes to methyl-donor depletion. In NAG, continuous tryptophan diversion into kynurenine draws methionine-cycle intermediates into compensatory reactions, producing systemic SAM/SAH depression — a global deficit visible in peripheral blood. In AG, IDO1 is inactive and tryptophan metabolism is preserved; instead, TNF-α-driven macrophage activation consumes SAM locally for histone methylation at inflammatory gene promoters, producing compartment-specific depletion that doesn't show up on standard peripheral methylation panels. The AG individual looks methylation-normal — not because methylation is healthier, but because the depletion is localized rather than global.

A second, independent appetite-control layer — the CCK/vagal satiety reflex — is also differentially affected: in NAG, SST-28 hyperactivation creates a "double satiety lock" by blocking CCK release entirely; in AG, CCK is released normally but TNF-α progressively desensitizes vagal CCK-1R signaling, so the post-ingestive satiety brake is present but blunted — a second, independent driver of AG hyperphagia alongside elevated ghrelin.

Proposed biomarker panel & testable predictions

Every marker below is measurable with existing clinical laboratory infrastructure — no novel assays required.

A KYN/TRP threshold of 5% has been reported with IDO1 activation present in roughly 58.7% of a published ASD cohort — notably close to the NAG prevalence this framework's clinical observations would predict. Combined with whole-blood serotonin, these two markers alone should correctly classify most individuals at intake.

Two headline testable predictions

• Forskolin response magnitude — larger observable cAMP-dependent effect in NAG (bigger delta from a depressed baseline) than in AG.
• Post-IMIG appetite direction — NAG individuals should show appetite emergence; AG individuals should show satiety-regulation emergence (recognizing fullness for the first time). If this directional split is observed in a mixed IMIG cohort, it confirms the hypothesis clinically without requiring the full biomarker panel.

Treatment implications

The core principle: standard ASD supplement protocols applied uniformly across both subgroups are likely misdirected in AG — and, in at least one specific case, potentially harmful.

Shared across both groups

Luteolin (NF-κB suppression regardless of which arm activated it), forskolin (adenylyl cyclase bypass — larger effect in NAG), sulforaphane (Nrf2/oxidative stress, plus IDO1-modulating properties that matter more in NAG), low-dose naltrexone (reduces mu-opioid receptor activation in both), and hydroxocobalamin (methylation support — methylcobalamin remains contraindicated in slow-COMT individuals regardless of subgroup).

NAG-specific priorities

IDO1 pathway modulation (EGCG, resveratrol) to break the AhR/IDO1 feedback loop — a prerequisite for cascade resolution that continues driving NF-κB even after immune clearance; tryptophan restoration only after IDO1 activity has been reduced, since premature supplementation simply feeds more kynurenine production; and aggressive systemic methylation support given the global methyl-donor depletion characteristic of the IDO1-active state.

AG-specific priorities

TNF-α modulation (omega-3, curcumin, low-dose naltrexone) as the primary upstream target; and caution around ghrelin normalization post-IMIG — if elevated ghrelin is providing partial neuroprotective compensation via hippocampal GHS-R1a signaling, cognitive function should be tracked alongside metabolic normalization rather than treating the drop as an unambiguous win.

Recovery trajectories differ, too

NAG — sequential: IMIG clears immune load → IDO1 subsides → AhR loop gradually breaks → NF-κB resolves → SST-14 recovers → appetite emerges as SST-28 normalizes. Each step depends on the prior one completing. AG — spring release: IMIG reduces TNF-α load → NF-κB resolves more directly, with no loop to break → SST-14 recovers → satiety regulation emerges, amplified by the partial cAMP/PKA axis already present. This is a more elastic recovery — the system retains enough residual signaling capacity to rebound relatively quickly.

The proposed study — and what isn't established yet

The hypothesis generates a critical experiment that doesn't yet exist in the published literature: a cross-sectional cohort study co-measuring fasting acyl ghrelin and KYN/TRP ratio in an appetite-stratified ASD cohort. No intervention, no novel assays — a single blood draw in 40–60 caregiver-classified NAG/AG participants, with primary outcome being whether KYN/TRP threshold status co-segregates with appetite classification and ghrelin direction as predicted. This is executable now within existing academic pediatric neurology infrastructure, without dedicated research funding beyond standard laboratory costs.

Related reading

• The Anatomy of Autism — the foundational cascade model this hypothesis builds from.
• Restoring the Somatostatin Signal in Immune-Dysregulated ASD — the primary SST-14 restoration framework.
• L1-79 and the Three-Intervention Model — where forskolin's differential NAG/AG response prediction connects to the broader intervention framework.
• The IMIG Clinical Trial — where NAG/AG stratification would inform trial design and endpoint selection.
• Autism and Alzheimer's — Two Sides of a Plasticity Problem — the shared plasticity biology underlying why SST-14 recovery differs by cascade state, relevant to both NAG and AG.