The pupillary light reflex (PLR) is a candidate intermediate
phenotype associated with autism. In this study, we aimed to extend
previous research supporting the PLR as a candidate intermediate
phenotype for autism by exploring epigenetic variability associated with
individual differences in early PLR development.
The researchers measured a brain reflex that appears very early in
life, called the PLR: “Pupillary Light Reflex”. Instead of
searching for mutations, the study investigates how genes are being
regulated from the very beginning of life.
What does this reflex do?
When light increases, the pupil constricts. The faster and the
stronger this constriction, the more efficiently a specific brain
circuit is functioning.
This circuit involves:
This is one of the most basic circuits of the human brain, which makes it ideal for studying early neurological development.
Because autistic infants already show differences in this
reflex months before any behavioral signs appear.
From previous studies, researchers already knew that:
| Trajectory of the pupillary response in children later diagnosed with autism | |
| Changes in the pupillary light reflex across development | |
| Developmental stage | What happens in the pupil |
|---|---|
| At 9–10 months | The pupil constricts more strongly |
| From 9 → 14 months | The pupillary response begins to weaken |
| From 14 → 24 months | The pupillary response becomes slower |
| Summary based on longitudinal studies of the pupillary light reflex (PLR). | |
And this means that the circuit that controls the pupil is
developing differently.
The central question of this research was: “Can early
differences in the pupillary light reflex (PLR) be explained by
epigenetic alterations—specifically DNA methylation—rather than by
mutations in the DNA sequence?”
(If you’re not familiar with DNA methylation, think of it as a
“chemical tag” that can increase or decrease the activity of certain
genes, helping regulate when and how much a gene is expressed, without
changing the DNA sequence.)
To address this, the researchers collected:
- DNA from the infants’ buccal (cheek) cells at 9 months;
- Pupillary measurements at 9, 14, and 24 months.
The guiding question during data collection was:
“Do DNA marks at 9 months predict how the pupil will behave later
on?”
For pupil speed (latency), they found very strong results:
- 4 extremely significant DNA regions;
- 13 larger regions (DMRs) that were also significant.
This effect was especially strong between 14 and 24 months, which is
precisely the period when autistic traits begin to emerge.
These regions regulate genes involved in:
- Neuronal formation;
- Growth of neural circuits;
- Synaptic organization.
| Table 1. Summary of probe-level EWAS analyses and identified DMRs for each phenotype | |||||
| Epigenome-wide analysis of the pupillary light reflex in infants | |||||
| N | λ |
DMP analysis associated with PLR
|
DMR analysis associated with PLR
|
||
|---|---|---|---|---|---|
| Probes with strict significance | Significant probes at the discovery stage | Bonferroni-adjusted significant DMRs | |||
| Latency (ms) | |||||
| 9 months | 48 | 1.16 | 0 | 28 | 1 |
| 14 months | 47 | 1.14 | 2 | 27 | 1 |
| 24 months | 40 | 1.02 | 1 | 21 | 1 |
| 9–14 months | 44 | 0.94 | 0 | 6 | 1 |
| 14–24 months | 37 | 1.20 | 1 | 60 | 9 |
| Amplitude (%) | |||||
| 9 months | 46 | 1.07 | 0 | 11 | 1 |
| 14 months | 47 | 1.03 | 0 | 21 | 1 |
| 24 months | 37 | 1.05 | 0 | 28 | 5 |
| 9–14 months | 42 | 1.06 | 0 | 28 | 3 |
| 14–24 months | 34 | 0.96 | 0 | 15 | 7 |
| λ = genomic inflation factor; sig. = significant; Bonferroni = Bonferroni correction. | |||||
Two key genes emerged, both considered high risk for autism in the SFARI
database:
- NR4A2: Responsible for dopaminergic neurons (motivation, focus,
sensory responsiveness);
- HNRNPU: Responsible for synapse formation and neural network
organization.
Based on this, they concluded that the pupillary reflex is
connected to genes that build the autistic brain.
Regarding how much the pupil constricts (amplitude),
the effect was more diffuse. They did not identify a single dominant
gene, but rather a large network involving:
- Immune system;
- Metabolism;
- Calcium signaling;
- Biological clock;
- Cellular signaling.
This means that the strength of the pupillary response is regulated
by many systems simultaneously, not only by the reflex circuit itself,
which explains why the epigenetic signals are more widespread.
(Although NR4A2 and HNRNPU were not identified as direct epigenetic
targets in this study, both are part of the enriched biological
processes identified by functional analyses, particularly those related
to neurodevelopment, gene regulation, and nuclear organization.)
They appear in the study indirectly, as components of the biological
networks represented by Gene Ontology terms.)
This study shows us something very profound: autism begins in
basic sensory circuits, not in behavior.
In other words, before:
- Language;
- Socialization and
- Cognition,
there is already a brain that responds differently to light,
regulated by neurodevelopmental genes that are modulated by DNA
methylation.
This means that autism begins with a difference in how the
brain calibrates the sensory world.
In summary, these early differences in sensory calibration can be
observed very early through the pupillary light reflex, possibly
associated with epigenetic alterations such as DNA
methylation.
It is revolutionary because these measurements were taken at 9 months
of age, that is:
- Long before any diagnosis;
- Using an automatic reflex;
- Linked to chemical marks on DNA.
This means that in the future it may be possible to:
- Identify risk;
- Understand the type of brain involved;
- Offer early sensory support without labeling and/or
pathologizing.
And this is not a “defect,” but rather a measurable style of neurodevelopment at the molecular level.
Table 2 is central because it shows the finest level of the
epigenetic signal (CpG by CpG) before “scaling up” to DMRs and
biological interpretation.
It lists individual CpG probes whose methylation is significantly
associated with pupillary light reflex (PLR) latency at different ages
in early childhood.
In other words, it shows which specific points in the genome where
DNA methylation is related to the time it takes for the pupil to respond
to light.
| Table 2. Differentially methylated probes associated with pupillary light reflex latency | |||||
| p < 2.4 × 10⁻⁷ | |||||
| Probe | Latency phenotype (months) | Genomic location (hg19) | Illumina gene annotation | Effect size (β) | p-value |
|---|---|---|---|---|---|
| cg05148717 | 14 | chr6:28829171 | −0.00055 | 3.74 × 10−8 | |
| cg22367466 | 14 | chr12:6840532 | COPS7A | −0.00035 | 4.28 × 10−8 |
| cg09732535 | 24 | chr16:89331997 | 0.00066 | 1.85 × 10−8 | |
| cg15130433 | 14–24 | chr6:159240081 | EZR | 0.00025 | 2.21 × 10−7 |
| ¹ A negative effect size (−β) indicates hypermethylation associated with faster latency (cross-sectional measures) or increasing latency over time (change scores). | |||||
This table identifies individual CpG probes with highly significant associations between DNA methylation and pupillary light reflex (PLR) latency, highlighting localized epigenetic signals that precede and support the regional analysis based on DMRs.
Table 3 is the epigenetic heart of the paper because it
moves beyond the “point-by-point” level (Table 2) and demonstrates
coordinated regulation across genomic regions.
It presents Differentially Methylated Regions (DMRs)
whose methylation levels are associated with PLR
latency at different ages or developmental intervals.
This table supports the conclusion that the signal is not
statistical noise, but rather reflects structured, biologically
meaningful epigenetic regulation.
| Table 3. DMRs significantly associated with each PLR latency phenotype | ||||||
| Latency phenotype (months) | Genomic location (hg19) | Number of probes | Probes | Illumina gene annotation | Effect size (β) | Bonferroni-adjusted p-value |
|---|---|---|---|---|---|---|
| 9 | chr10:74927623–74927863 | 8 | cg15213114; cg04167018; cg21416602; cg25138168; cg08571229; cg04749667; cg16124546; cg12276298 | FAM149B1; ECD | −0.007 | 7.25 × 10−3 |
| 14 | chr1:103573700–103573772 | 2 | cg26436330; cg20847625 | COL11A1 | 0.010 | 6.33 × 10−4 |
| 24 | chr10:45495981–45496216 | 3 | cg18382353; cg16512882; cg15078013 | C10orf25; ZNF22 | −0.020 | 8.49 × 10−4 |
| 9–14 | chr3:52569053–52569169 | 3 | cg18337363; cg08365687; cg13284614 | NT5DC2; LOC440957 | −0.009 | 7.35 × 10−3 |
| 14–24 | chr17:2699706–2699718 | 2 | cg05890550; cg25373595 | RAP1GAP2 | 0.020 | 6.53 × 10−5 |
| ¹ A negative effect size (−β) indicates hypermethylation associated with faster latency (cross-sectional measures) or increasing latency over time (change scores). | ||||||
| ² This row details the most significant DMR associated with latency change between 14–24 months (see Supplementary Table 2 for all significant DMRs). | ||||||
Conceptually, Table 3 shows that the PLR is not merely a peripheral
reflex; it reflects regional epigenetic states linked to
neurodevelopment.
Temporally, it reveals distinct genomic regions regulated at different
ages, reinforcing the idea of sensitive epigenetic windows.
Methodologically, it justifies downstream functional analyses (GO,
SFARI).
The identified DMRs indicate that PLR latency is associated with
coordinated regional patterns of DNA methylation, suggesting dynamic
epigenetic regulation of genes involved in neurodevelopment across the
first two years of life.
Table 4 presents the DMRs (Differentially Methylated Regions) most
significantly associated with the amplitude of the pupillary light
reflex (PLR) at different ages or developmental intervals.
It confirms and strengthens the CpG-by-CpG findings, showing that the
observed alterations are not isolated events, but rather part of
biologically coherent regional patterns.
| Table 4. Most significant DMRs associated with each PLR amplitude phenotype | ||||||
| Amplitude phenotype (months) | Genomic location (hg19) | Number of probes | Probes | Illumina gene annotation | Effect size (β) | Bonferroni-adjusted p-value |
|---|---|---|---|---|---|---|
| 9 | chr13:103452556–103453215 | 4 | cg06518779; cg21251000; cg15186648; cg15193473 | BIVM; KDELC1 | 0.04 | 0.000213 |
| 14 | chr11:111249659–111250201 | 5 | cg19126910; cg17390301; cg24049888; cg18316498; cg11362935 | POU2AF1 | -0.03 | 0.022200 |
| 24 | chr5:77253833–77253990 | 3 | cg25051331; cg09048251; cg07595776 | NA | -0.08 | 0.000823 |
| 9–14 | chr10:102295134–102295549 | 5 | cg07690778; cg26303175; cg07080220; cg08314679; cg07510080 | HIF1AN | 0.07 | 0.002670 |
| 14–24 | chr15:89786761–89787223 | 4 | cg22813622; cg01741397; cg15769724; cg06870609 | FANCI | -0.04 | 0.001220 |
Regional methylation changes in key genes and regulatory regions are associated with the amplitude of the pupillary light reflex across early development, indicating that coordinated epigenetic mechanisms contribute to the maturation of early sensory responses—an intermediate phenotype relevant to autism.
| Levels of epigenomic analysis presented in the study tables | ||
| Table | Level of analysis | What it shows |
|---|---|---|
| Table 2 | Individual CpG (DMP) | Point-wise associations |
| Table 3 | Integration / annotation | Functional context |
| Table 4 | Regions (DMRs) | Coordinated and robust effects |
Here is where the study “moves up a level”, transforming CpGs/DMRs
into biological processes: it presents the Gene Ontology (GO) terms that
are significantly enriched among genes associated with pupillary light
reflex (PLR) phenotypes.
PLR latency: how quickly the pupil responds. PLR amplitude: how
strong the constriction is.
Thus, instead of asking “Which CpG changed?”, the question here becomes: “Which biological processes appear more often than expected among genes associated with PLR?”
| Table 5. Significantly enriched Gene Ontology (GO) terms for each PLR phenotype | ||||
| Number of significant GO terms | Most significantly enriched GO term (GO ID) | Fold enrichment | p-value (FDR) | |
|---|---|---|---|---|
| PLR latency | ||||
| 9 | 12 | Homophilic cell adhesion via plasma membrane adhesion molecules (GO:0007156) | 4.20 | 3.17 × 10−7 |
| 14 | 34 | Negative regulation of developmental process (GO:0051093) | 1.84 | 6.95 × 10−3 |
| 24 | 0 | – | – | – |
| 9–14 | 16 | Homophilic cell adhesion via plasma membrane adhesion molecules (GO:0007156) | 6.62 | 8.27 × 10−22 |
| 14–24 | 16 | Regulation of cellular component organization (GO:0051128) | 1.50 | 5.00 × 10−3 |
| PLR amplitude | ||||
| 9 | 0 | – | – | – |
| 14 | 73 | Regulation of glucuronosyltransferase activity (GO:1904223) | 24.41 | 3.86 × 10−8 |
| 24 | 0 | – | – | – |
| 9–14 | 22 | System development (GO:0048731) | 1.41 | 1.30 × 10−3 |
| 14–24 | 0 | – | – | – |
| sig. = significant after FDR (False Discovery Rate) adjustment. | ||||
PLR latency is strongly associated with processes related to cell adhesion, structural organization, and regulation of neural development, especially during the first year of life. In contrast, PLR amplitude shows more localized associations, highlighting metabolic processes and systemic developmental pathways within specific developmental windows. In other words, latency reflects neural connectivity and architectural organization, whereas amplitude reflects functional and metabolic modulation.
Table 6 is one of the most important in the study because it
establishes a direct bridge between the epigenetic findings related to
the PLR and genes already implicated in autism.
It lists Differentially Methylated Regions (DMRs) that are significantly
associated with PLR phenotypes (latency or amplitude), overlap with
genes cataloged in the SFARI database—which compiles genetic evidence
for autism—and remain significant after rigorous statistical
corrections.
These are not isolated CpGs, but regional epigenetic signals affecting
genes with well-established relevance to ASD.
| Table 6. Significantly associated DMRs annotated to autism-related genes | |||||
| DMR genomic location (hg19) | Probes | Associated phenotype | Gene | SFARI gene score | Evidence level |
|---|---|---|---|---|---|
| chr11:19372012–19372234 | cg23330281; cg24137774; cg25909885 | Latency 14–24 months | NAV2 | 2 | Strong candidate |
| chr7:2968559–2968595 | cg22989995; cg23770265 | Amplitude 24 months | CARD11 | 2 | Strong candidate |
| chr13:38444227–38444490 | cg16409955; cg23021771 | Amplitude 9–14 months | TRPC4 | 3 | Suggestive evidence |
| ¹ The SFARI gene score reflects the strength of evidence for a gene’s relevance to autism. | |||||
The genes NAV2, TRPC4, and CARD11 act at complementary levels of neurodevelopment—structural organization, neuronal excitability, and neuroimmune modulation—providing a biologically plausible framework for the association between DNA methylation, the pupillary light reflex, and risk for ASD.
| Functional integration of the three analyzed genes | ||
| Gene | Primary level | What it regulates |
|---|---|---|
| NAV2 | Structural | Assembly and guidance of neural circuits |
| TRPC4 | Functional | Neuronal excitability and intracellular signaling |
| CARD11 | Modulatory | Inflammatory state and autonomic modulation |
Probes are short synthetic DNA sequences
fixed on the array chip that hybridize to a specific genomic location,
allowing the measurement of the methylation level of a cytosine
(CpG) at that site.
Probes capture gene regulation, not mutations.
Each probe generates a value called β (beta value), calculated as:
β = relative methylation (0 = unmethylated | 1 = fully
methylated)
This value is interpreted as:
- Negative β: the region is more methylated in a specific group.
- Positive β: the region is less methylated.
What does it mean when several probes appear together in a
study?
In the table, entries such as cg23330281 / cg24137774 / cg25909885
indicate that all these CpG probes are:
- Located close to each other in the genome;
- Within the same Differentially Methylated Region (DMR);
- And together show a significant change in methylation, reinforcing
that this is not noise, but a regional epigenetic signal (which
is precisely what defines a DMR - Differentially Methylated
Region).
The probes capture gene regulation rather than mutations, many DMRs
are found in:
- Cell adhesion genes;
- Axon guidance pathways;
- Synaptic development processes.
This reflects sensory calibration of the developing brain, which
is exactly what this article demonstrates.
“Reinforcing that this is not noise, but a regional
epigenetic signal”
But Jânice, what does that actually mean?
Let’s break it down.
1. What does “noise” mean in this context?
In methylation data, noise refers to:
- Random signal variation;
- Technical array errors;
- Statistical fluctuation at a single CpG site;
- Spurious effects that do not replicate and lack biological
coherence.
An example of noise: A single CpG probe appears “significant,”
but neighboring probes show no similar pattern.
(This is common in EWAS - Epigenome-Wide Association Studies - and
usually lacks robust biological meaning.)
2. What characterizes a real epigenetic
signal?
A biologically relevant epigenetic signal shows:
- Multiple CpG probes located close together;
- A consistent direction of effect (all more or less methylated);
- Statistical significance after correction for multiple testing;
- Localization consistent with regulatory regions such as promoters,
enhancers, and gene bodies.
This indicates coordinated regulation, not chance.
3. Why is it called “regional”? Because the
alteration does not occur at a single point, but across a genomic region
involving multiple contiguous CpGs that are typically regulated
together.
DNA methylation acts in regulatory blocks, not point by
point. (Which is exactly the definition of a
DMR.)
4. Why do multiple probes strengthen the conclusion that this
is not noise?
Because statistically, the probability that several adjacent CpGs change
in the same direction with significance in the same group by chance
alone is extremely low.
Biologically, this reflects the action of:
- DNMTs (DNA methyltransferases);
- TET enzymes (Ten-Eleven Translocation);
- Chromatin remodelers acting across an entire functional region.
(In other words, an active regulatory program, not random
fluctuation.)
What is the relationship between probes, DMPs, and DMRs?
| Relationship between probes, DMPs, and DMRs | |
| Term | Definition |
|---|---|
| Probe | Measures the methylation level of a single CpG site |
| DMP (Differentially Methylated Position) | Individual CpG site with a statistically significant difference in methylation between groups |
| DMR (Differentially Methylated Region) | Genomic region containing multiple CpG sites with coordinated methylation changes |
| CpG: a DNA region where a cytosine is followed by a guanine; methylation at these sites is a key mechanism of epigenetic regulation. | |
EWAS (Epigenome-Wide Association Study) is
a study design that investigates associations between epigenetic marks
on DNA (primarily DNA methylation) and biological or clinical
phenotypes.
An EWAS analyzes the genome CpG site by CpG site, testing whether
differences in DNA methylation are associated with a given phenotype,
using:
In this study, the use of EWAS made it possible to detect very early
epigenetic signatures prior to the clinical diagnosis of ASD, linked to
a basic sensory function (the pupillary light reflex, PLR), without
relying on rare genetic mutations. This strongly supports the idea that
autism may involve early regulatory differences, rather than only
fixed genetic alterations.
DMPs and DMRs within EWAS
DMP (Differentially Methylated Position): a single CpG
site whose methylation level is associated with the phenotype.
DMR (Differentially Methylated Region): a cluster of
nearby CpG sites showing coordinated methylation changes, representing a
more robust and biologically interpretable signal.
Within EWAS, DMRs are particularly valuable because
they reduce false positives and reflect genuine regional epigenetic
regulation rather than isolated statistical fluctuations.
EWAS is not direct causality!
EWAS identifies associations, not causality by itself.
However, it reveals biologically plausible mechanisms that can be
further tested using cellular models, longitudinal studies, and
integrative analyses with genetic data.
Gene Ontology (GO) is a standardized classification
system used in molecular biology and bioinformatics to consistently
describe what genes and their products (RNAs, proteins) do, where they
act, and which biological processes they participate in—independently of
species.
GO functions as a controlled vocabulary combined with a
hierarchical structure, enabling cross-species gene comparison
and the interpretation of large-scale datasets such as RNA-seq,
proteomics and DNA methylation studies.