PMID 35429626-paper-review

Key Findings

Finding 1 — The core result: exposure-graded, DMN-specific topology reorganization at the connection-probability level

The paper's central claim is that pesticide exposure is associated with default mode network-specific topological reorganization, and that this signal appears specifically at the level of whether a connection exists (connection probability), not how strong existing connections are, and specifically within the DMN, not across the whole brain.

Study system. The cohort comprised n = 78 children (48 rural FW, 30 urban NFW), all Latinx, mean age ≈ 8.3–8.4 years. Brain networks were built from the Shen 268-region functional parcellation, with edges defined by Pearson correlation between regional time series (negative correlations set to zero). The DMN was represented by 37 regions of interest. Analysis used the two-part mixed-effects network regression implemented in WFU_MMNET: Part I models connection probability with a logistic link; Part II models connection strength. Models were adaptive-FDR corrected and covaried for maternal education and race, preschool education, parents' occupation, and inter-ROI anatomical distance plus distance².

Primary results (FW vs. NFW contrast, within DMN). The probability–clustering interaction was β_{r,COI×DMN×Clust} = 0.4790 (p < 0.0001), with a contrast sum β_{r,COI×Clust} + β_{r,COI×DMN×Clust} = 0.5226 (p < 0.0001) indicating that FW children's DMN connections are more positively governed by clustering. The probability–global-efficiency interaction was β_{r,COI×DMN×Eglob} = −0.4793 (p < 0.0001), contrast sum = −0.5058 (p < 0.0001), indicating FW children's DMN connections are more negatively related to global efficiency.

Secondary results (continuous historic exposure, FW only, prenatal–96 months). The same directional pattern held: probability–clustering β = 0.1444 (p < 0.0001), contrast = 0.1284 (p = 0.0003); probability–efficiency β = −0.0705 (p = 0.0304), contrast = −0.0759 (p = 0.0466). Low-exposure FW children resembled the NFW group; high-exposure FW children resembled the FW group as a whole.

Specificity / negative controls. Every connection-strength term was non-significant (p = 0.36–0.99), and every non-DMN global term was non-significant. This is what makes the finding interpretable: the effect is confined to DMN connection probability.

Directional interpretation. Higher exposure → DMN connection probability increasingly governed by clustering (local segregation) and decreasingly by global efficiency (long-range integration) — a shift toward a more segregated, less integrated DMN.

The authors' own verbatim framing captures both the result and its associational character:

"We found that network topology was differently associated with the connection probability between FW and NFW children in the DMN. Our results also indicated that, among 48 FW children, historic reports of exposure to pesticides from prenatal to 96 months old were significantly associated with DMN topology, as hypothesized." — PMID: 35429626

"the observed associations between network connectivity/topology and historic exposure reports in FW children provide compelling evidence for a contribution of pesticide exposure on altering the DMN network organization in this vulnerable population" — PMID: 35429626

Finding 2 — Convergent prior evidence: organophosphate exposure alters DMN-constituent structure/connectivity, and the same farmworker-network signature appears in adults

Three independent human neuroimaging lines converge on DMN-region vulnerability and contextualize this paper's rationale and result:

1. Rauh et al. 2012 (PMID: 22547821) — structural MRI of 40 children found prenatal chlorpyrifos associated with cortical enlargement including the cuneus/precuneus and mesial superior frontal gyrus — core DMN midline hubs. Verbatim: "enlarged superior frontal gyrus, gyrus rectus, cuneus, and precuneus along the mesial wall of the right hemisphere."

2. van den Dries et al. 2020 (Generation R; PMID: 32805249) — in n = 518 preadolescents, prenatal dimethyl-OP metabolites were associated with lower fractional anisotropy (B = −1.00, 95% CI −1.80, −0.20 per 10-fold increase) in white-matter tracts interconnecting DMN regions. Verbatim: "a 10-fold increase in averaged DM metabolite concentrations across pregnancy was associated with lower FA."

3. Bahrami et al. 2017 (PMID: 28583619) — the same research group's adult Latino farmworker study found "brain networks in farmworkers had more clustered and modular structures when compared to non-farmworkers," with erythrocyte cholinesterase activity tracking the differences. This mirrors the increased-segregation signature seen in the children's DMN and ties it to the cholinergic molecular anchor.

Together these establish that (a) DMN hub regions are structurally and microstructurally sensitive to prenatal OP exposure, and (b) the increased-segregation network phenotype is reproducible across the lifespan within this research program.

Finding 3 — Disease relevance: direct for OP developmental neurotoxicity/ADHD-cognitive risk, indirect for psychiatric DMN disorders, speculative for Parkinson disease

The paper diagnoses no disease: only 3 of 78 children had any parent-reported learning disability, and no cognitive outcomes were measured. Disease relevance is therefore mechanistic/exposure-pathway based, graded as follows:

Supporting verbatim evidence:

• "Most of the studies evaluating prenatal exposure observed a negative effect on mental development and an increase in attention problems in preschool and school children." — PMID: 24291036
• "We relate DMN dysfunction to dementia, schizophrenia, epilepsy, anxiety and depression, autism and attention deficit/hyperactivity disorder" — PMID: 18824195
• "Exposure to paraquat or maneb/mancozeb was associated with about a 2-fold increase in risk." — PMID: 23713084
• "Epidemiologic studies support a connection between organophosphate pesticide exposures and subsequent risk of Parkinson's disease" — PMID: 21968025

The paper itself does not diagnose any disease in its participants — this must be emphasized in any knowledge-base encoding.

Finding 4 — Ontology mapping (verified) and the acetylcholinesterase molecular-initiating-event anchor

To represent this paper in a structured knowledge base, the following CURIEs were verified via EBI OLS4:

Flagged correction: earlier iteration-1 guesses (ECTO:9000078, ECTO:9000541, UBERON:0002976, UBERON:0022352) were incorrect and are superseded by the verified IDs above.

The molecular-initiating-event anchor is acetylcholinesterase inhibition, the canonical OP mechanism, broadened by evidence that OP esters also act through non-cholinergic pathways (cAMP, glutamate/GABA, endocrine). Verbatim: "organophosphate esters can alter neurodevelopmental processes by interfering with noncholinergic pathways at environmentally relevant doses" (PMID: 34612677). Lower-level mechanistic models (absent from the paper, present in related work) include zebrafish OP developmental neurotoxicity — chlorpyrifos → AChE inhibition + down-regulated mbp/syn2a/α1-tubulin + locomotor deficits (PMID: 32730898, PMID: 29605586, PMID: 27018022). Verbatim zebrafish anchor: "The STC test accurately detected the effect of 8 out of 11 neuroactive substances (chlorpyrifos, chlorpyrifos-oxon, diazinon..." (PMID: 32730898).

Finding 5 — The residential-proximity exposure proxy is externally validated against autism, ADHD, and IQ loss

PMID:35429626 defines exposure as parent-reported months living next to agricultural fields — a residential-proximity proxy. Independent studies using the same proxy class or OP biomarkers link this exposure to downstream neurodevelopmental disease, strengthening the exposure arm of the inferred pathway:

Verbatim anchors: - "Proximity to organophosphates at some point during gestation was associated with a 60% increased risk for ASD, higher for third-trimester exposures" — PMID: 24954055 - "We observed a decrease of 2.2 points [95% confidence interval (CI): −3.9, −0.5] in Full-Scale IQ" — PMID: 28557711 - "A 10-fold increase in DMAP concentration was associated with an odds ratio of 1.55" — PMID: 20478945

Study Models / Systems

Models used in this paper

Models in the wider literature (NOT in this paper)

• Human birth cohorts: CHAMACOS (PMID: 29701446), Columbia CCCEH (Rauh, PMID: 22547821), Generation R (PMID: 32805249), CHARGE (PMID: 24954055), ECHO, Rhea — these use urinary DAP biomarkers or pesticide-use-registry proximity, measures this paper's proxy approximates.
• Rodent: chlorpyrifos/diazinon/parathion gestational and neonatal exposure (e.g., PMID: 18941570, PMID: 20026089).
• Zebrafish: chlorpyrifos and OP flame retardants → locomotor and transcriptional neurodevelopmental deficits (PMID: 32730898, PMID: 29605586, PMID: 27018022).
• In vitro: PC12 and neuronal cultures revealing cAMP-cascade convergence (PMID: 20026089, PMID: 21968025).

Biological Processes and Endpoints Measured

No cognitive, behavioral, or exposure-biomarker endpoints were measured in the paper. The molecular-initiating event (not measured here) is acetylcholinesterase inhibition (CHEBI:38462). [inference] The CURIEs above are proposed mappings; the DMN has no UBERON class, so MeSH D000082702 is used.

Mechanistic Model / Interpretation

The findings assemble into a candidate Adverse Outcome Pathway (AOP) in which this paper supplies one specific, empirically supported edge (the DMN-reorganization step):

   Pesticide exposure            MOLECULAR INITIATING EVENT        NEURODEVELOPMENTAL         NETWORK-LEVEL                 COGNITIVE / CLINICAL
   (residential proximity,  -->  Acetylcholinesterase inhibition   PERTURBATION         -->  KEY EVENT                -->  OUTCOME
    prenatal-96 mo)              (± non-cholinergic: cAMP,     -->  (altered maturation        Reorganized DMN topology:     (future risk; NOT
    [ECTO:0000530 /              glutamate/GABA, endocrine)         of late-maturing           ↑ clustering/segregation,     measured in this paper)
     ECTO:9000284]               [CHEBI:38462]                      DMN hubs)                  ↓ global efficiency/          ADHD, cognitive deficit,
                                                                                               integration                   ASD traits, ...
    -- validated externally --   -- rodent/zebrafish/in vitro --    -- Rauh, Generation R --   <== THIS PAPER (35429626) ==  -- epidemiology --

The direction of the network change — more clustering, less global efficiency — denotes a DMN that is more locally segregated and less globally integrated. Because DMN hubs (precuneus/PCC, mPFC, IPC) are late-maturing and support self-referential and higher-order cognition, a developmentally-timed shift toward segregation is a biologically plausible substrate for the attention and cognitive phenotypes repeatedly linked to OP exposure in the epidemiological literature (Findings 3 and 5) [inference]. The adult farmworker replication of the increased-segregation signature (PMID: 28583619), together with its link to erythrocyte cholinesterase activity, strengthens the mechanistic plausibility of the cholinergic anchor.

Importantly, the knowledge-base representation should encode this as an association edge with a proxy exposure node, not a causal/diagnostic edge. The paper supports the middle of the AOP; the molecular-initiating event and the clinical outcome are imported from other model systems and cohorts.

Evidence Base

Limitations and Knowledge Gaps

1. Cross-sectional design — cannot establish causality or temporal ordering; no longitudinal trajectory.
2. Proxy exposure — parent-reported residential proximity, not a measured pesticide biomarker; no chemical speciation, no intensity/dose; subject to recall bias.
3. Confounding by rurality and SES — FW vs. NFW differ in urban/rural residence and socioeconomic status; covariate adjustment cannot fully disentangle pesticide exposure from correlated deprivation and environmental factors.
4. No clinical or cognitive outcomes — the paper measures network topology only; it diagnoses no disease and cannot link the DMN change to any behavioral phenotype in these children (only 3/78 had any parent-reported learning disability).
5. Modest sample — 78 total (30 NFW) limits power and generalizability.
6. Analytic choices — Shen-268 parcellation, correlation thresholding, and negative-edge zeroing all influence topology; robustness to alternative pipelines is untested here.
7. Specificity but not localization — the effect is DMN-specific and probability-specific, but the paper does not identify which individual DMN edges/hubs drive it.

Proposed Follow-up Experiments / Actions

1. Longitudinal re-imaging of the PACE5 FW cohort to test whether the DMN segregation shift persists, worsens, or recovers, and whether it predicts later cognitive/attention outcomes.
2. Add pesticide biomarkers (urinary DAP metabolites, erythrocyte cholinesterase activity) to convert the proximity proxy into a quantified, speciated exposure and test dose-response at the molecular level.
3. Couple imaging to cognition — administer standardized attention/executive and IQ batteries (SRS-2, Conners, WISC) to directly connect DMN topology to clinically meaningful phenotypes.
4. Edge-level localization — identify which DMN hubs (precuneus/PCC vs. mPFC vs. IPC) drive the connection-probability effect to sharpen the UBERON-level knowledge-base annotation.
5. Better control groups — recruit rural non-farmworker and low-exposure rural children to decouple pesticide exposure from rurality/SES.
6. Cross-species anchoring — align the human DMN-segregation readout with zebrafish/rodent AChE-inhibition and connectivity endpoints to formalize the AOP from MIE to network key event.
7. Knowledge-base encoding — represent this as an association edge: exposure to organophosphorus compound (ECTO:9000284) → altered DMN (MeSH:D000082702) connection-probability topology, with the MIE node acetylcholinesterase inhibitor (CHEBI:38462) imported from related models and an explicit "no disease diagnosed" flag.

What this paper can and cannot support as evidence

Can support: a specific, exposure-graded association between probable historic pesticide exposure and DMN-specific connection-probability topology (↑ clustering/segregation, ↓ global efficiency/integration) in developing Latinx children — a mechanistic key-event edge for an OP developmental-neurotoxicity knowledge base.

Cannot support: any disease diagnosis, any causal claim, any cognitive/behavioral outcome, or any molecular measurement — these must be imported from other cohorts and model systems.

Usable verbatim evidence snippets (from the paper): - "We found that network topology was differently associated with the connection probability between FW and NFW children in the DMN." - "among 48 FW children, historic reports of exposure to pesticides from prenatal to 96 months old were significantly associated with DMN topology, as hypothesized." - "the observed associations between network connectivity/topology and historic exposure reports in FW children provide compelling evidence for a contribution of pesticide exposure on altering the DMN network organization in this vulnerable population."

Artifacts

• OpenScientist final report
• OpenScientist final report

Reference Validation

Checked with linkml-reference-validator 0.5.1.

All extracted references resolved successfully.

Term Validation

Checked with linkml-term-validator 0.4.5, through the ols: adapter.

Terms the report names something else

These identifiers resolve, so nothing about them looks wrong, and the ontology calls them something unrelated to what the report calls them. That usually means the identifier is not the one the sentence needs:

• UBERON:0022353 (2 mentions) - the report calls it "PCC"; UBERON calls it posterior cingulate cortex
• UBERON:0006088 (2 mentions) - the report calls it "IPC"; UBERON calls it inferior parietal cortex

Terms whose name is worth a second look

The report's name for these is recognisably related to the term's own name without being one of them. A loose paraphrase reads the same way as a citation of the wrong sibling term - and so does a related synonym, which the ontology records precisely because it names something adjacent rather than the same thing - so these are listed rather than judged:

• CHEBI:38462 (4 mentions) - the report calls it "acetylcholinesterase inhibitor", "acetylcholinesterase inhibition"; CHEBI calls it EC 3.1.1.7 (acetylcholinesterase) inhibitor, and lists "acetylcholinesterase inhibitor" among its other names

Terms named inconsistently

The report gives these identifiers more than one name of its own:

• CHEBI:38462 - called "acetylcholinesterase inhibitor", "acetylcholinesterase inhibition"