Bilateral prefrontal stimulation as a lever on vaccine memory durability

The two halves of the brain continuously inhibit each other through the corpus callosum, and when that push-pull balance breaks down — after a stroke, or in depression — one hemisphere becomes hyper-excitable and suppresses the other. Non-invasive brain stimulation can restore the balance either by exciting the under-active hemisphere or by inhibiting the over-active one, and the strongest version of the intervention does both at once: an excitatory electrode over one prefrontal cortex and an inhibitory electrode over the other. Separately, a large immunology literature tracks how vaccine protection decays over months and years: circulating antibodies fall fast, often halving within six months, while memory T and B cells settle onto a stable plateau that can last five years or more. What almost nothing connects the two fields is any experiment that perturbs the hemispheric balance while measuring what happens to adaptive immune memory over a long horizon.

The one bridge in the literature is narrow and important: a double-blind, sham-controlled trial found that stimulating the left prefrontal cortex with 10 Hz rTMS in depressed patients shifted peripheral cytokines — interleukin-2 and C-reactive protein rose over 12 weeks, early clinical improvement tracked falling TNF-α (r = −0.272), and sustained recovery tracked rising IL-2 (r = −0.340) Inflammatory Cytokines Changed in Patients With Depression Before and After Repetitive Transcranial Magnetic Stimulation Treatment. That paper used a purely unilateral montage, followed cytokines for only 12 weeks, and never measured anything adaptive — no neutralizing antibodies, no memory B or T cells. The open question this proposal takes on is whether bilateral push-pull stimulation, which the stroke literature shows is the montage that actually rebalances interhemispheric competition, changes the multi-month decay kinetics of vaccine-induced immune memory when a unilateral montage does not.

Core Claim

The claim is a causal one about a central-peripheral coupling. Interleukin-2 is a master cytokine governing T-cell proliferation and survival, and the depression rTMS trial is direct evidence that cortical stimulation can move IL-2 in the periphery. Since IL-2-dependent maintenance signals are precisely what sustains memory T-cell survival after vaccination, a stimulation montage that more effectively rebalances prefrontal circuitry should shift the shape of post-vaccination immune decay curves — either flattening the exponential antibody decline, raising the cellular memory plateau, or both. The bilateral-versus-unilateral contrast is the key ingredient, because the stroke literature shows unilateral excitation fights against persistent transcallosal inhibition from the hyper-excitable opposite hemisphere, whereas a bipolar montage (anode over left DLPFC, cathode over right) directly rebalances it and measurably raises interhemispheric functional connectivity Increased both cortical activation and functional connectivity after transcranial direct current stimulation in patients with post-stroke: A functional near-infrared spectroscopy study. No published study has run this contrast with any immunological endpoint beyond circulating cytokines, and none has tracked adaptive immunity under neuromodulation at all.

Research Program

The centerpiece is a randomized, double-blind, sham-controlled trial built around a scheduled vaccine event so that the immune system has a well-characterized memory curve to perturb. Adults presenting for a standard seasonal influenza or SARS-CoV-2 booster vaccination would be randomized to three arms, all receiving identical session counts and durations so that the comparison isolates the montage: a bilateral arm (anodal tDCS over left DLPFC at F3, cathodal over right DLPFC at F4, 2 mA, 30 minutes — the exact parameters of the stroke cognitive trial), a unilateral arm (active anode at F3, sham cathode at F4 — this is simultaneously the ablation that isolates the bilateral ingredient), and a full sham arm. Stimulation begins on the day of vaccination and runs for 14 daily sessions, a dose chosen deliberately: Yang et al. found that interhemispheric functional connectivity shifts were undetectable after 7 sessions and only emerged after 14 Increased both cortical activation and functional connectivity after transcranial direct current stimulation in patients with post-stroke: A functional near-infrared spectroscopy study, so a shorter protocol risks a null result from under-dosing.

Immune tracking follows the design grammar of the longitudinal immunity literature rather than the cytokine-snapshot grammar of the neuroimmune literature. Neutralizing (or hemagglutination-inhibition, for influenza) antibody titers would be measured at baseline, day 7, month 1, month 3, month 6, and month 12. Memory B-cell and antigen-specific T-cell responses would be measured by standard ex-vivo intracellular cytokine staining at months 1, 6, and 12 — and, critically, by an eight-day in-vitro expansion assay of the EpiMax type at months 6 and 12. This dual-assay design is forced by a methodological tension the Ebola literature exposed clearly: in the PREVAC ancillary study, standard ex-vivo staining suggested rVSV-induced T-cell responses waned after one year, while the expansion assay showed memory T cells persisting through 60 months with no decline between months 24 and 60 Long-term cellular immunity of vaccines for Zaire Ebola Virus Diseases. A neuromodulation study relying on ex-vivo assays alone could falsely conclude that stimulation erased cellular memory when it merely silenced the easily detectable effector pool. The seven ELISA cytokines of the depression trial would also be collected at matched time points to anchor the results to the closest prior work.

The mechanistic sub-study closes the loop between the two directions. A subset of participants would undergo 52-channel functional near-infrared spectroscopy during a verbal fluency task at baseline, mid-protocol, and session 14, replicating the measurement that detected connectivity gains in the stroke cohort Increased both cortical activation and functional connectivity after transcranial direct current stimulation in patients with post-stroke: A functional near-infrared spectroscopy study. The decisive analysis is whether the individual change in interhemispheric connectivity predicts the individual immune decay slope — that is, whether the physiological variable the bilateral montage is supposed to move is the same variable that carries the immunological effect. If cytokine changes occur without connectivity changes, or vice versa, the proposed coupling is wrong and the trial will say so.

A secondary computational analysis extends the predictive side using HR-VILAGE-3K3M: A Human Respiratory Viral Immunization Longitudinal Gene Expression Dataset for Systems Immunity, the harmonized repository of 66 studies and 14,136 longitudinal gene-expression profiles. Its benchmark for predicting vaccine responsiveness — PCA-based logistic regression at F1 = 0.822, AUC = 0.748 — sets the quantitative reference point for asking whether pre-vaccination transcriptomic features combined with hemispheric-balance features from our baseline fNIRS/EEG recordings improve prediction of who responds. One caution carries over from that dataset's own findings: aggressive batch correction such as ComBat suppresses genuine biological variance between responders and non-responders, so the analysis would use batch-aware validation rather than blindly harmonized data.

The statistical core is a linear mixed-effects model on log-transformed titers, where the treatment effect of interest is the interaction between time and arm:

$$\log(\text{titer}_{it}) = \beta_0 + \beta_1 t + \beta_2\,\text{arm}_i + \beta_3\,(t \times \text{arm}_i) + u_i + \epsilon_{it}$$

where $t$ is months since vaccination, $\text{arm}_i$ indexes stimulation condition, and $u_i$ is a per-participant random intercept. The linear-in-log-time form matches the exponential decay documented in the inactivated-vaccine cohort, where wild-type neutralizing titers fell from 17.95 at month 3 to 8.00 at months 9–12 Immune Persistence against SARS-CoV-2 after Primary and Booster Immunization in Humans: A Large-Scale Prospective Cohort Study; $\beta_3$ is then directly the change in decay half-life attributable to the montage, and the same model applied to memory T-cell frequencies tests for a shifted plateau rather than a shifted slope.

Baselines

The naive reference point is the natural-history decay curve: what happens to vaccine memory with no intervention at all. The best-quantified version comes from the prospective cohort of Li et al., where neutralizing titers against wild-type virus declined from a geometric mean of 17.95 at 3 months to 8.00 at 9–12 months after two inactivated-vaccine doses, and roughly half of participants in the 2009 H1N1 cohort had lost protective neutralizing activity by day 180 Antibody Dynamics of 2009 Influenza A (H1N1) Virus in Infected Patients and Vaccinated People in China. Our sham arm reproduces this reference in-sample; agreement between our sham decay slope and these published values would validate the assay pipeline, and any treatment claim must show a slope measurably shallower than this in-sample control under the same assays, the same lab, and the same follow-up schedule.

The strongest existing method is unilateral left-DLPFC stimulation, because it is the only intervention in the literature with demonstrated peripheral immune effects in a sham-controlled design. In the Pudong trial, unilateral 10 Hz rTMS over left DLPFC significantly raised IL-2 and high-sensitivity CRP at both 2 and 12 weeks and produced superior Hamilton Depression scores at 2 weeks (14.41 vs. 17.14, p = 0.001), with 12-week recovery correlating with IL-2 elevation (r = −0.340, p = 0.010) Inflammatory Cytokines Changed in Patients With Depression Before and After Repetitive Transcranial Magnetic Stimulation Treatment. Our unilateral arm matches this competitor as closely as the modality allows — same target, same daily-session structure, same 14-session course and 30-minute duration, matched blinding, and identical immunological assays and time points — so the comparison is fair on everything except montage geometry and modality. (Delivering a true unilateral rTMS comparator arm alongside tDCS arms would make the trial four-armed; if resources allow it, we would add it, since modality is a genuine confound between our tDCS design and the rTMS evidence base.) Beating this baseline means the bilateral arm produces a shallower titer decay slope or a higher month-12 memory T-cell plateau than the unilateral arm, not merely than sham — a difference only on sham would fail to establish that the bilateral ingredient, rather than stimulation per se, carries the effect.

The third baseline is peripheral rather than neural: booster-timing optimization, the standard immunological tool for managing decay. Li et al. showed a booster raised geometric mean titers 3.5- to 6.8-fold, with a 6-month primary-to-booster interval optimal for wild-type titers — but intervals from 3 to 12 months made no significant difference against immune-evasive variants, and Omicron titers stayed lowest regardless of schedule Immune Persistence against SARS-CoV-2 after Primary and Booster Immunization in Humans: A Large-Scale Prospective Cohort Study. We would re-parameterize decay slopes from that published schedule data as a comparison trajectory: the question our trial answers that schedule optimization cannot is whether a host-side intervention moves the decay curve itself rather than re-starting it.

Finally, the within-trial ablation is the unilateral-versus-bilateral tDCS contrast already built into the design: identical current, duration, session count, and blinding, differing only in whether the cathode is active. That contrast alone determines whether the proposal's core ingredient — push-pull rebalancing rather than unilateral excitation — is doing any work.

Why not just…?

Why not just use unilateral DLPFC stimulation, since it already changes cytokines? This is the most serious objection, and the honest answer is that the trial could come back showing no difference between montages. What the existing evidence says against settling for unilateral is that unilateral excitation operates against ongoing transcallosal inhibition from the opposite hemisphere: Chen et al. showed that inhibiting one motor cortex with cTBS released the other hemisphere and significantly improved contralateral-hand motor function (p < 0.05), with state-occupancy changes in the unstimulated hemisphere persisting 80 minutes Tracking the Immediate and Short-Term Effects of Continuous Theta Burst Stimulation on Dynamic Brain States, and Yang et al. showed the bipolar montage raised interhemispheric connectivity where a unilateral approach would leave the hyper-excitable hemisphere unchecked Increased both cortical activation and functional connectivity after transcranial direct current stimulation in patients with post-stroke: A functional near-infrared spectroscopy study. The unilateral arm inside our design settles the question empirically rather than by argument: if it matches the bilateral arm on every immune endpoint, the bilateral ingredient is not needed and the simpler protocol wins.

Why not just optimize booster timing or adjuvant formulation instead of stimulating the brain? Because the published evidence shows schedule optimization has hit a ceiling exactly where it matters: against antigenically drifted variants, booster intervals from 3 to 12 months made no significant difference in restored titers Immune Persistence against SARS-CoV-2 after Primary and Booster Immunization in Humans: A Large-Scale Prospective Cohort Study, and new adjuvants such as CpG-emulsion raise acute immunogenicity in mice Immunogenicity and cellular response of a herpes zoster virus gEgI fusion protein adjuvanted with CpG-emulsion in mice without addressing whether memory durability is host-regulated. The mechanistic rationale for a central lever is that IL-2 — a cytokine rTMS demonstrably elevates in humans — is the same cytokine that maintains memory T-cell survival, so a host-side intervention targets the decay mechanism rather than the initial peak. If our sham and bilateral arms show indistinguishable decay curves, this objection wins and the schedule/adjuvant approach is the right tool; the trial is powered to show that cleanly.

Why not just model immunity computationally from existing datasets, without any brain stimulation? Harmonized repositories like HR-VILAGE-3K3M are valuable — its benchmark of F1 = 0.822, AUC = 0.748 for predicting vaccine responsiveness is the reference our predictive sub-study builds on — but they are observational and transcriptomics-only, lacking serum proteomics, antibody titers for many cohorts, and any mechanistic perturbation, limitations its own authors acknowledge HR-VILAGE-3K3M: A Human Respiratory Viral Immunization Longitudinal Gene Expression Dataset for Systems Immunity. Prediction and causal intervention are different goals; only a controlled trial can say whether rebalancing cortical circuitry changes a decay curve. The dataset still earns a role in the proposal as the source of the responder baseline our fNIRS-augmented model must beat.

Why not just use bilateral physical training instead of electrodes? Bilateral robot-assisted training does improve outcomes and connectivity in stroke patients Bilateral upper limb robot-assisted rehabilitation improves upper limb motor function in stroke patients: a study based on quantitative EEG, and combining robotic training with targeted stimulation adds further benefit Effects of high-definition tDCS targeting individual motor hotspot with EMG-driven robotic hand training on upper extremity motor function: a pilot randomized controlled trial. But physical training cannot target the prefrontal cortex, cannot reverse polarity (excite one side, inhibit the other), and cannot be sham-blinded in the same way, so it cannot test the interhemispheric set-point hypothesis this proposal is about. It would be a reasonable adjunct in a larger effectiveness trial, but not a substitute for the mechanism question here.

Proposed Experiments and Decisive Checks

The first experiment is the primary trial readout: the mixed-effects decay model above, with the pre-registered success criterion that the bilateral arm shows a significantly shallower log-titer slope than both sham and unilateral arms (one-sided α = 0.05 on $\beta_3$), with month-12 geometric mean titers and memory T-cell plateau height as co-primary endpoints. The second experiment is the mechanistic mediation analysis: session-14 interhemispheric connectivity change must (a) replicate the delayed-emergence pattern Yang et al. observed and (b) mediate a statistically meaningful fraction of the arm effect on decay slope; failure of (b) would falsify the central set-point account even if a total effect exists. The third experiment is the assay-discrepancy check: any apparent loss of cellular memory under stimulation must be confirmed by the in-vitro expansion assay before being believed, per the PREVAC lesson that ex-vivo staining alone can report waning where none exists Long-term cellular immunity of vaccines for Zaire Ebola Virus Diseases. The fourth is the predictive sub-study on HR-VILAGE-3K3M features plus baseline hemispheric-balance measures, with the published PCA-logistic-regression benchmark of F1 = 0.822 as the number to beat. Trial hygiene follows the known failure modes of this literature directly: active sham, blinded outcome assessors, exclusion of immunomodulatory medications (the confound the depression trial carried), and a pre-specified interim futility analysis, because the honest prior on effect size — correlations of r ≈ 0.27–0.34 in the cytokine predecessor — says a null result is a live possibility the design must be able to detect and report.