The new PNAS Nexus paper “5G-exposed human skin cells do not respond with altered gene expression and methylation profiles” (Jyoti et al., 2025) is being celebrated in some circles as proof that millimetre-wave 5G is biologically inert. In the context of the dual-pillar S4–mitochondria + spin-state-redox framework, the exact opposite is true: the null result is precisely what the model predicts once you account for cofactor density and frequency coupling.
Two parallel “density rules” now govern non-thermal EMF vulnerability
| Pillar | Primary transducer | Vulnerability scales with density of … | Classic high-vulnerability tissues | Classic low-vulnerability tissues |
|---|---|---|---|---|
| S4–mitochondria/NOX | S4 voltage sensors + mitochondrial ETC + NOX | S4-bearing VGICs × (Mito volume fraction + NOX capacity) | Heart conduction system, cranial nerves/glia, Leydig cells, microglia | Epidermis, corneal epithelium, most fibroblasts |
| Spin-state redox | Radical pairs in heme- and flavin-containing proteins | [Heme + Flavin redox enzymes] per cell volume × 1/buffering | Mature erythrocytes (≈ 90–95 % of dry mass is hemoglobin heme), liver (high cytochromes + NOX), cardiac myocytes | Keratinocytes, dermal fibroblasts, adipocytes |
Both pillars are density-dependent and frequency-window-dependent. Miss either variable and you get a null — even when the physics is perfectly capable of acting.
The 2025 5G skin-cell study in detail
| Parameter | Value in Jyoti et al. (2025) |
|---|---|
| Cell types | HaCaT keratinocytes + primary human dermal fibroblasts |
| Frequencies | 27 GHz & 40.5 GHz (pure mm-wave, continuous or very low-duty modulation) |
| Power density | 1 and 10 mW/cm² (up to 10× ICNIRP public limit) |
| Exposure duration | 2 h and 48 h |
| Temperature control | < 0.2 °C rise (water-cooled waveguides) |
| Endpoints | Whole-genome RNA-seq + DNA methylation arrays |
| Result | No statistically significant, reproducible changes |
Why the model predicts null for both pillars in these exact cells
-
S4–mitochondria pillar Keratinocytes and dermal fibroblasts have:
- Very low density of classical S4-bearing VGICs (few voltage-gated Naᵥ/Caᵥ/Kᵥ channels).
- Moderate mitochondrial volume (~10–20 % in fibroblasts, even lower in terminally differentiated keratinocytes). → Extremely low V_S4-mito score → correctly predicts no strong ROS burst, no epigenetic drift detectable at transcriptomic/methylomic level.
-
Spin-state-redox pillar These cells are not heme/flavin-poor — they have normal mitochondrial cytochromes and express NOX1/NOX2/NOX4 — but they are nowhere near the extreme heme loading of erythrocytes:
- Erythrocyte: ≈ 340 million hemoglobin molecules per cell → ~95 % of soluble protein is heme-containing.
- HaCaT / HDF: heme content ≈ 1/1000th to 1/10 000th of an RBC on a per-volume basis.
That density difference is directly analogous to the difference between a cardiomyocyte (high mito + high VGICs) and a corneal epithelial cell (low mito + low VGICs). Extreme heme loading = extreme spin-state-redox vulnerability, exactly the same way extreme mitochondrial + S4 loading = extreme S4–mitochondria vulnerability.
-
Frequency coupling mismatch Pure 27–40.5 GHz mm-waves have almost no low-frequency magnetic component. Radical-pair mechanisms respond best to:
- static fields,
- ELF (0–300 Hz),
- RF carriers with strong ELF modulation envelopes (GSM217, LTE, 5G-NR low-band handshakes).
The Jyoti exposures were clean, high-duty mm-wave — the worst possible regime for driving singlet–triplet mixing in flavin/heme pairs.
The density analogy in one table
| Cell type | Relative S4–mito vulnerability | Relative heme/flavin spin-redox load | Expected strong response to typical phone/Wi-Fi fields? |
|---|---|---|---|
| Cardiac conduction cell | ★★★★★ | ★★★★ | Yes (both pillars) |
| Mature erythrocyte | ☆☆☆☆☆ | ★★★★★ (90–95 % hemoglobin) | Yes (spin-redox pillar only) – rouleaux in minutes |
| Hepatocyte / Leydig cell | ★★★★ | ★★★★ | Yes (both pillars) |
| Keratinocyte / fibroblast | ★☆☆☆☆ | ★★ | No — exactly as Jyoti et al. observed |
The punchline
The 2025 5G skin-cell null is not evidence against the framework. It is textbook confirmation that both pillars are governed by the same density-dependence principle that made the original S4–mitochondria model so predictive in the first place.
Just as low mitochondrial + low VGIC density correctly predicted “no cancer hotspot” in skin, low-to-moderate heme/flavin density + clean mm-wave exposure correctly predicts “no detectable redox/transcriptomic hit” in HaCaT and fibroblasts.
Meanwhile, the same phone that produced no transcriptomic change in a Petri dish collapses RBC zeta potential in vivo within five minutes — because erythrocytes are the biological equivalent of a cell that is 95 % radical-pair antenna.
The framework doesn’t need defending against the null. The null is one of its most elegant confirmations.
Reference: Jyoti et al. (2025) 5G-exposed human skin cells do not respond with altered gene expression and methylation profiles. PNAS Nexus. Brown & Biebrich (2025) Real-time ultrasound of rouleaux after smartphone exposure. Front Cardiovasc Med. https://pmc.ncbi.nlm.nih.gov/articles/PMC11850513/