INTRODUCTION
Circadian rhythms orchestrate daily fluctuations in physiology and behavior [
1]. These rhythms extend far beyond the basic sleep–wake cycle, exerting wide-ranging effects on metabolic, cardiovascular, and immune function that have been extensively studied in both animals and microorganisms [
1]. At an environmental level, light is the primary input, exerting its influence as it interacts with photoreceptors in the retina [
2]. At the central neural level, the suprachiasmatic nucleus receives input from the retinohypothalamic tract and is the master controller of a cascade of pathways [
2]. At the cellular level, the clock is maintained by transcriptional–translational feedback loops involving core genes such as
CLOCK and
BMAL1, which activate the expression of PER and CRY [
3]. These, in turn, inhibit their own transcription, creating sophisticated rhythmic oscillations with a 24-hour periodicity [
2].
The immune system is particularly sensitive to circadian control. Murine models show that clock components regulate adaptive immune modulators, including cytokines, chemokines, and antigen receptors [
3]. For example,
BMAL1 influences monocyte trafficking and pro-inflammatory cytokine production, while
CLOCK regulates NF-κB signaling pathways [
3]. Toll-like receptor 9, which is critical for innate and adaptive immunity, also demonstrates circadian expression in mice, influencing vaccine responses [
4]. This influence extends to B-cell and T-cell control, facilitating expression of key migratory regulators such as C-C chemokine receptor type 7, C-X-C motif chemokine receptor 4 (CXCR4), and interleukin-7 receptor (IL-7R) [
5]. Expression of these proteins is of particular importance when considering priming and antibody generation, where leukocyte migration between the bloodstream and secondary lymphoid tissue is key.
Vaccine chronotherapy has recently emerged as an area of clinical interest. Several human studies have examined the timing of vaccination and have yielded mixed results [
6,
7]. Several studies have investigated the effect of temporal vaccination and demonstrate that morning vaccination can enhance antibody responses compared to afternoon administration [
6]. However, results remain mixed; vaccine-dependent, and prior work has treated time of day as a fixed variable without accounting for interindividual variability in circadian rhythms [
7]. It has been shown that environmental disruptions to clock phase synchrony (shift work, social jet lag, sleep restriction, and nocturnal light exposure) lead to negative health outcomes, including diabetes, cardiovascular disease, and immunosuppression [
8]. Because circadian phase differs among individuals and can be drastically influenced by environmental factors, the same objective time may reflect different physiological states, suggesting circadian phase, rather than time-of-ay, may better predict vaccine efficacy [
6-
8].
When considering the importance of vaccination effectiveness, the influenza vaccine stands out as a prime candidate. Influenza is a vaccine-preventable disease that can cause symptoms such as fever, aches, chills, discomfort, and headache. While the majority of subjects recover from the disease, at-risk groups may develop severe complications, leading to 290,000–650,000 deaths each year [
9]. Because influenza viruses undergo continuous genetic drift, they are constantly evolving. As a result, vaccines must be updated each year to better match the strains circulating in the population. In rapid succession, inactivated vaccines are produced from candidate vaccine viruses (CVVs) with the goal of providing subjects with active immunity in the form of antibodies targeting one of the major viral membrane glycoproteins: hemagglutinin [
10]. Despite this specific targeting, even if CVVs are matched appropriately, vaccine effectiveness remains suboptimal.
Seasonal influenza vaccination from 2009–2025 demonstrated large variability, ranging from 19%–60% effectiveness, and was estimated as 40%–60% effective when the vaccine strains are antigenically well-matched [
11]. This variance is historically defined as a reflection of several concepts: the match between the strain included in the vaccine and the circulating virus, and the subject’s history of previous vaccination or infection [
10]. However, other potential factors to vaccine-mediated immune responses exist, and this feasibility study was designed to explore the hypothesis that individualized circadian rhythms impact the antibody response induced by this vaccine platform.
To investigate the correlation between individual circadian rhythms and immune responses against vaccines, this pilot study examined the possible role of circadian phase in humoral immune responses against influenza and identified potential avenues to further investigate this relationship. Several methods currently exist to evaluate circadian phase in humans. Endogenous melatonin, core body temperature (CBT), and cortisol have been widely used to infer circadian rhythm in humans [
12]. However, these methods have historically been either invasive or subject to activity masking [
13]. With the introduction of wrist actigraphy, an alternative method of continuous data collection has emerged, enabling non-invasive determination of one’s circadian rhythm in daily life. This pilot study utilized non-invasive wrist actigraphy monitoring in subjects using the Corsano Cardiowatch 287-2B, a wearable biosensor that continuously estimates rhythms of many biometric measurements, namely: cortisol, heart rate (HR), respiratory rate (RR), CBT, skin body temperature (SBT), and stress. The aim of this study was to determine the feasibility of using this data set to explore the hypothesis that antibody titers would differ based on an individual’s biometric phase at the time of vaccination. Using the biosensor data, we examined individual circadian phases and then compared influenza virus antibody titers in participants vaccinated at different points relative to their various rhythms.
METHODS
The study was reviewed and approved by the Institutional Review Board at Touro University Nevada (TUNIRB000314). Participants were required to provide informed consent before participating, and no identifiable personal data was collected.
Recruitment, participants
For the study, 20 healthy young students (10 male, 10 female, ages 20–28) were recruited from the student body of Touro University Nevada. Participants were asked to complete 3 surveys (informed consent, Morningness-Eveningness Questionnaire [MEQ] [
14], and an intake form). Inclusion criteria included individuals who were graduate students at Touro University Nevada between the ages of 20 and 30 years old and who had not received an influenza vaccine within the past year.
Exclusion criteria included individuals with active hematological malignancy, individuals with an inherited immune deficiency, individuals with an infection of human immunodeficiency virus, individuals with bleeding diathesis or condition associated with prolonged bleeding that would, in the opinion of the investigators, contraindicate a peripheral blood draw, individuals who have had a diagnosed influenza infection within the past 6 months, individuals who have already received an influenza vaccine or immunization of some kind within the past year, individuals who have an allergy to any component of the influenza vaccine, women who are pregnant, nursing, or planning a pregnancy, and individuals who are unable or unwilling to participate in all trial visits and assessments.
Using wrist actigraphy for data collection
This study utilized the Corsano Cardiowatch 287-2B (Corsano Health B.V.) as the sole device to monitor and collect physiologic data from participants. This bracelet is an FDA and Conformité Européenne–Medical Device Regulation-certified device. Participants were instructed on the use and care of the instrument before administration. This included charging, cleaning, and monitoring instructions via the Corsano smartphone application. At this point, participants were de-identified and given randomly generated patient identification codes, linking survey results, physiological data collected via bracelet, antibody titers, and erythrocyte sedimentation rate results. The bracelet continuously collected and uploaded encrypted data to the Corsano Cloud (General Data Protection Regulation and Health Insurance Portability and Accountability Act compliant).
The Corsano bracelet contains a plethora of sensors, including a galvanic skin response sensor, a triaxial accelerometer, a photoplethysmography module (containing 7 LEDs; Green, Red, and Infrared wavelengths; 2 photodiodes), a surface probe thermometer, and a heatflux sensor (greenteg). These sensors generated approximations of the following biometrics: pulse rate, SpO2, respiration rate, skin temperature, CBT, steps, energy expenditure, heart rate variability, stress, estimated cortisol contributions, and sleep data.
Data processing pipeline
Raw biometric data were processed and analyzed using custom code and available packages in R (v4.4.1; R Foundation for Statistical Computing), including CircaCompare [
15] and GLMMcosinor [
16]. Comma-separated values files containing raw biometric data were processed using a custom R pipeline and filtered to exclude physiologically impossible values as artifacts. The pipeline generated two primary outputs: Longitudinal Datasets and a Circadian Profile. Longitudinal Datasets maintained a continuous timeline for the 2-week monitoring period for day-to-day visualization and analysis. Circadian Profile used a representative day, accounting for all information from the 2-week monitoring period and folding the data into a standardized 24-hour day representative of their personalized circadian rhythm. Each biometric was run through GLMMCosinor to generate a cosinor wave using a fundamental period of 24 hours and a second harmonic of 12 hours. Using this cosinor wave, key parameters were extracted, namely, the midline estimating statistic of rhythm (mesor), amplitude, acrophase (peak), and bathyphase (trough).
Intake, interventions, and timeline
Participants completed an intake survey and the MEQ on day 1. Subjects (n=20) were then randomly assigned to either a morning vaccination group (8:00–9:00 am; n=10) or an afternoon vaccination group (4:00–5:00 pm; n=10). This intake survey collected demographic information, including age, sex, ethnicity, location, university affiliation, academic program, and upcoming high-stakes examinations. Medical history screening included prior autoimmune disease, recent or current illnesses, prescription or over-the-counter medication use, antibiotic use, and common chronic conditions such as asthma, allergies, diabetes, hypertension, and sleep disorders. Participants were also assessed for factors relevant to influenza vaccination safety, including history of respiratory illness, prior influenza infection, egg allergy, previous adverse vaccine reactions, and Guillain-Barré Syndrome. Lifestyle factors were recorded, including sleep habits, partner sleep status, exercise patterns, eating routines, and substance use within the past six months. Additional items included pregnancy status and recent exposure to significant life stressors or traumatic events. This intake form was used to verify eligibility, ensure vaccine safety, and characterize baseline participant health behaviors.
The MEQ is a 19-question survey used to assess individual daily rhythm patterns. The validated survey evaluates preferred sleep–wake timing, ease of awakening, alertness, and hunger in the morning, perceived peak performance times, and preferred scheduling of mentally or physically demanding tasks. Items also assess reliance on alarm clocks, likelihood of staying up late or waking early under unrestricted conditions, and self-identification as a “morning” or “evening” type. Scores from the MEQ were used to classify participants along the morningness-eveningness spectrum to characterize perceived chronotype.
Days 1–7, participants began continuous physiologic monitoring using the Corsano Cardiowatch. On day 8, a baseline blood sample was collected, followed by administration of the quadrivalent influenza vaccine. Monitoring with the Corsano Cardiowatch continued throughout days 9–16, during which participants also completed a wellness survey assessing interim symptoms and overall health.
The wellness survey confirmed that participants received the influenza vaccine and documented the date of administration. Participants reported post-vaccination symptoms, including fever, cough, fatigue, nausea, vomiting, other side effects, or no symptoms within the preceding seven days. Additionally, the survey asked whether participants sought medical care for these symptoms or required hospitalization.
No study-related interventions were required between days 17–35. On day 36, participants returned for the second intervention, which consisted of a follow-up blood sample to assess post-vaccination antibody responses.
Antibody quantification assays
Antibody titers were determined by hemagglutination (HA) inhibition assay using two influenza virus strains. Both H1N1 (BEI Resources, NIAID, NIH: influenza A virus, A/Connecticut/11/2023 [H1N1]pdm09 with NA-I223V and NA-S247N mutations, NR-59881) and H3N2 (BEI Resources, NIAID, NIH: influenza A virus, A/Baltimore/JH-335/2022 [H3N2], NR-59460) influenza strains were propagated in Madin-Darby canine kidney cells and tested in quantitative HA inhibition assays. Turkey red blood cells (RBC) (TURKEY RBCS 5%, Cat# 50-415-691, LAMPIRE Biological Laboratories) were prepared at 0.75% in 1x phosphate-buffered saline (PBS) and used in quantification assays. All sera were treated with cholera filtrate (Cat# C8772, Sigma-Aldrich), serving as a receptor-destroying enzyme (RDE) for nonspecific binders in sera.
Positive and negative controls were standardized and used in each Hemagglutination Inhibition Assay (HAI). A portion of all 19 participants’ post-vaccination sera was pooled, and aliquots were made for each assay. This undiluted serum then served as the positive control. Negative controls were standardized as PBS.
Direct HA assay
The HA assay was utilized to determine the volume of virus stock required. Specifically, it was used to generate HA units needed to agglutinate an equal volume of RBC solution and standardize the amount of virus used in each HAI. Target HA=4HA/25uL per WHO standard [
17].
To prepare the plate, turkey RBCs were diluted to 0.75% in PBS. For H1N1, 25 μL of undiluted H1N1 antigen stock was added to the first wells of rows 1 and 2, followed by 50 μL of PBS to achieve a 1:2 dilution. The contents were then serially diluted across each row using 50 μL transfers to generate two-fold dilutions (1:2, 1:4, 1:8, etc.). After dilution, 50 μL was discarded from the last well in each row. The steps were repeated for H3N2 in rows 4 and 5. Then, 50 μL of the diluted RBC solution was added to all used wells to reach a total volume of 100 μL per well. The plate was incubated at room temperature for 30 minutes and then tilted at 90° for 25 seconds prior to reading the HA pattern.
HA was assessed following the incubation period by tilting the plate at 90° to distinguish agglutinated from non-agglutinated patterns. Non-agglutinated wells displayed a compact “button” or “halo” of RBCs that settled to the bottom, whereas agglutinated wells showed a diffuse, uniform reddish suspension without a defined button. The HA titer was defined as the highest virus dilution demonstrating complete hemagglutination. This dilution corresponds to 1 HA unit. To determine the volume containing 4 HA units for subsequent assays, two dilution steps preceding the endpoint were chosen for the hemagglutination inhibition assay (e.g., for an HA endpoint of 1:32, the 1:8 dilution would be chosen for the HAI assay).
HAI
RDE preparation was completed following the manufacturer’s instructions. The lyophilized cholera filtrate was reconstituted in 5.0 mL of sterile, deionized water (DI H2O). The resulting solution contained 0.01% as a preservative.
Serum aliquots (pre-vaccination and post-vaccination) were thawed on ice, and individual polymerase chain reaction (PCR) tube strips were prepared for each participant. For each participant, 10 μL of pre-vaccination serum was dispensed into four wells of a PCR strip, and 10 μL of post-vaccination serum was dispensed into the remaining four wells, yielding a total of twenty labeled strips. To generate a positive control, 10 μL from each post-vaccination sample was pooled into a 1-mL cuvette to obtain 200 μL, which was then allotted into four additional PCR tube strips (10 μL per well). In total, 24 PCR tube strips were prepared, each containing 10 μL of its designated serum. RDE was added to each tube (30 μL per reaction), briefly vortexed, and the samples were incubated at 37°C overnight to remove nonspecific inhibitors. The following day, samples were heat-inactivated at 56°C for 30 minutes to terminate RDE activity, then stored at 4°C for short-term use.
The working antigen solution was prepared based on the predetermined 4 HA-unit titer obtained from the HA. For twenty 96-well plates, each requiring 25 μL of antigen per well (2,400 μL per plate), a total volume of approximately 50 mL was prepared. Appropriate dilution calculations were performed to adjust the antigen stock to the final working concentration corresponding to 4 HA units.
A 0.75% RBC suspension was freshly prepared for use in the HAI assay. For twenty 96-well plates, each requiring 50 μL of RBC solution per well (4,800 μL per plate), a total volume of approximately 100 mL was prepared at the final working concentration.
A 1:2 serial dilution was performed beginning with an effective starting dilution of 1:8 (generated from a 1:4 serum–RDE mixture combined with 25 μL PBS). Each well of the 96-well plate was pre-filled with 25 μL PBS. Column 1 received 50 μL of the serum–RDE mixture, after which 25 μL was serially transferred across each successive well, discarding 25 μL from the final well. Subsequently, 25 μL of the 4-HA antigen solution was added to every well. Plates were tapped ten times on each side and incubated at room temperature for 30 minutes. Following the antigen-serum incubation, 50 μL of the 0.75% RBC suspension was added to all wells. The plates were then tapped ten times on all sides and incubated for an additional 30 minutes at room temperature.
HAI titers were recorded as the reciprocal of the last well demonstrating complete inhibition of HA (e.g., inhibition through a 1:32 dilution corresponds to a titer of 32). Plates were tilted at a 90° angle for 25 seconds to visualize RBC flow patterns. Complete inhibition appeared as a compact dot, partial inhibition as a dot with a partial streak, and no inhibition as a straight line extending to the bottom of the well.
RESULTS
Participants
A total of 20 individuals volunteered, met the inclusion criteria, consented, and were randomly assigned anonymous patient IDs. Of the 20 participants, 19 completed the full protocol and were included in the final per-protocol analysis. A single participant failed to complete the second blood draw at one month post-vaccination and was removed from the data pool.
Multi-system ambulatory circadian rhythm capture
Physiologic data captured via biometric wearables during the two-week monitoring period were processed through a custom analytical pipeline. Daily biological rhythms distinct from random activity noise were observed in all individuals for the measured variables: cortisol, breath, energy, stress, HR, heart rate variability (HRV), RR, CBT, and SBT. The custom analytical pipeline applied to the data collected generated representative rhythms for each individual (
Figure 1).
Circadian biometric phase at vaccination predicts antibody response to both viral strains
We examined whether the phase of physiologic circadian rhythms at the time of vaccination was associated with differences in antibody titers against H1N1 and H3N2 after 1 month. Cortisol and HRV both showed significant differences in antibody response by vaccination phase. The following sections discuss trends in antibody magnitude for both viral strains, based on the peak/trough phase difference at the time of vaccination for cortisol and HRV. Negative phase differences indicate individuals vaccinated prior to their peak/trough, while positive phase differences indicate participants vaccinated after their peak/trough of the given rhythm.
Cortisol peak phase difference and antibody magnitude
Cortisol peak phase difference at the time of vaccination was significantly associated with antibody magnitude change one month post-vaccination. Phase data from participants (n=19) were compared to post-vaccination antibody magnitude difference for both viral strains. Linear (R
2=0.516, p=0.001) and quadratic (p=0.011) models demonstrated a significant positive association, indicating that more positive peak phase differences at the time of vaccination were associated with increased titers to H1N1 (
Figure 2). A similar result was observed for the H3N2 viral strain, where a linear regression (R
2=0.344, p=0.012) (
Figure 3) demonstrated a significant positive association.
HRV trough phase difference and antibody magnitude
HRV trough phase difference at the time of vaccination demonstrated a strong, consistent association with antibody response across both viral strains. Data from participants (n=19) were compared, where a linear regression (R
2=0.636, p=0.001) for H1N1 titers and a quadratic model (p=0.006) demonstrated a significant negative association (
Figure 4). For the H3N2 strain, linear (R
2=0.616, p=0.001) and quadratic (p=0.009) models demonstrated a similar result and significant negative association (
Figure 5).
Steps peak phase difference and antibody magnitude
Steps peak phase difference at the time of vaccination demonstrated a significant positive association with the magnitude of antibody response to one viral strain. Analysis of participant data (n=19) and linear regression (R
2=0.251, p=0.036) indicated that vaccination during the post-steps peak phase is associated with higher H1N1 antibody titers one month post-vaccination (
Figure 6). No significant association was found with the H3N2 viral strain.
Additional physiologic markers show no significant association
Investigations into breath, CBT, energy, HR, RR, and stress yielded no significant correlations.
Time of day vaccination was not associated with antibody response
Morning (n=10) and after (n=9) vaccination groups were used for secondary analysis of clock time effect on antibody titers. There was no significant difference in raw antibody titers or magnitude of response between morning and afternoon vaccination groups for either viral strain.
Subjective chronotype was inversely associated with antibody magnitude
Secondary analysis of predicted chronotype demonstrated a significant inverse relationship with antibody titer for both viral strains. Results of the MEQ yielded 6 participants classified as morning chronotype and 13 participants classified as intermediate chronotype (neither morning or evening). Linear regression showed that MEQ score was significantly inversely associated with magnitude change in H1N1 antibody titer (R2=0.279, p=0.009), and H3N2 antibody titer (R2=0.312, p=0.004), indicating that more morning chronotypes were associated with lower antibody responses.
Post-vaccination survey results
Results from the one-week post-vaccination survey revealed that seven individuals reported symptoms post-vaccination. In the morning group, 6 individuals reported symptoms, including fatigue, headache, and muscle aches. In the after group, 3 individuals reported symptoms including headache and fatigue. No individuals from either group reported illness, needed antibiotics, or were hospitalized.