Diurnal rhythm of the human plasma proteome

Abstract

Background

Plasma is the most used clinical specimen, yet diurnal variation in plasma proteins remains largely unexplored. We aimed to identify diurnally-regulated proteins in healthy individuals and assess their potential diagnostic implications, and highlight how diurnal awareness can advance future biomarker research.

Methods

Twenty-four healthy young individuals were studied under highly controlled conditions. Venous blood was drawn every three hours over a 24-h period, yielding 216 samples, of which 208 high-quality plasma samples were analyzed via high-throughput mass spectrometry. The missing data were filtered and imputed, and rhythmicity was assessed using Cosinor-based modeling with Benjamini–Hochberg correction. Tissue and pathway enrichment analyses were performed using the DAVID functional annotation tool.

Findings

Of 523 proteins that passed quality thresholds, 138 (~ 26%) exhibited significant diurnal oscillations. Tissue enrichment analysis revealed that most rhythmic proteins originated from the liver and platelets, with additional enrichment in a variety of tissue types. Pathway enrichment showed diurnal regulation of hemostasis, immune signaling, integrin-mediated processes, glucose metabolism, and protein synthesis. Notably, 36 clinically utilized biomarkers, including albumin, amylase, and cystatin C exhibited diurnal variation, suggesting that failing to account for temporal fluctuations may reduce diagnostic precision.

Interpretation

These findings demonstrate that over one-quarter of the human plasma proteome is under diurnal control. Such oscillations might have direct clinical implications, as the time-of-day may alter biomarker accuracy. Incorporating diurnal timing into diagnostic and research protocols, through standardized sampling or time-sensitive reference intervals, could improve patient care and inform future biomarker discoveries. Further research in larger, more diverse populations is needed to generalize these results and streamline practices in a way that takes diurnal variation into account.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12014-025-09551-7.

Introduction

Blood is the most used diagnostic specimen. Despite significant advancements in biomarker discovery, most clinical diagnostics do not account for diurnal fluctuations in circulating proteins. This gap could lead to misinterpretation of results and suboptimal treatment timing, affecting patient care and research outcomes. Understanding these temporal variations could improve diagnostic accuracy, optimize clock-based therapeutic strategies, and enhance biomarker research.

The circadian rhythm (Latin: circa meaning ‘approximately’ and diem meaning ‘day’) governs 24 h oscillations in biological processes. It is regulated by a central biological clock in the hypothalamic suprachiasmatic nucleus (SCN), also known as the master clock, and by peripheral clocks in tissues throughout the body. While the functions of central and peripheral clocks are well understood, their interactions remain unclear. At the cellular level, a transcriptional-translational feedback loop drives rhythmic gene expression, leading to oscillations in biological functions throughout the day [1]. These rhythms are modulated by zeitgebers (German: ‘time givers’), which are external and internal cues, signals that synchronize (entrain) the biological rhythm to the 24 h sleep and wake cycle. Light is the primary zeitgeber, but feeding, exercise, and temperature also play crucial roles in circadian entrainment [2].

In this study, we use the term diurnal variation to denote the 24-h fluctuations observed under normal, entrained living conditions, arising from the interplay between the endogenous circadian clock and external zeitgebers such as light, feeding, and physical activity. Accordingly, the protein rhythms observed here should be viewed as diurnally regulated, potentially arising from endogenous circadian control, external cues present (e.g., light, feeding, physical activity), or both. Given the real-life, entrained setting of the experiment, we cannot determine the exact contribution of each factor.

Proteins are the most frequently used biomarkers [3]. However, little is known about the circulating proteome during diurnal rhythm in healthy humans. High-throughput proteomic analysis involves the study of proteins, including their expression pattern and function, and this is typically obtained by mass spectrometry-based approaches. This technology is increasingly being used, not only for biomarker research, and is increasingly implemented in clinical diagnostics [4].

To investigate the diurnal regulation of plasma proteins, we analyzed plasma samples from 24 healthy individuals. Blood samples were collected nine times over a 24 h-period while the participants remained in a standardized environment with controlled light exposure, food intake, movement, and sleep conditions. Mass spectrometry (MS)-based proteomics was applied to analyze plasma protein dynamics. By identifying diurnally-regulated plasma proteins, this study aims to provide insight into potential diagnostic implications, and highlight how diurnal awareness can inform optimal timing for blood-based diagnostics and therapeutic interventions.

Research in context

Evidence before this study Blood sampling is fundamental to diagnosis and treatment in hospital settings. However, research on circadian and diurnal rhythms in the plasma proteome, and their implications for diagnostic accuracy, remains limited. We searched Google Scholar up to February 26, 2025, using the terms “Plasma Proteins and Circadian Rhythm”, “Plasma Proteins and Diurnal Variations”, “Circadian Rhythm and Plasma Proteome” and “Mass Spectrometry, Plasma Proteomics and Circadian Rhythm”. We identified fewer than five relevant studies, most of which relied on SomaScan aptamer-based assays that provide only partial protein coverage and may introduce bias in specificity and quantitation. Consequently, unbiased mass-spectrometry analyses of circadian and diurnal rhythm protein variation in healthy individuals are lacking. This gap hinders our understanding of how time-of-day protein fluctuations affect patient testing, protein research and ultimately diagnostic accuracy Added value of this study

Using a high-throughput, unbiased mass spectrometry approach with multiple time point sampling in 24 healthy individuals, we found that 26% of plasma proteins exhibited significant diurnal rhythms. Our study offers a comprehensive overview of how the plasma proteome varies by time-of-day, providing critical insights that can guide future research and help refine diagnostic protocols Implications of all the available evidence These findings suggest that time-of-day fluctuations in plasma proteins may be relevant to the interpretation of clinical blood tests. However, additional research in larger and more diverse cohorts is needed to determine whether standardized sampling times or time-sensitive reference ranges would improve diagnostic accuracy and patient outcomes. Moreover, our results imply that previous proteomic investigations may have been influenced by the lack of circadian and diurnal consideration, emphasizing the potential value of carefully timed sampling and diurnal considerations in future diagnostics, treatment and research

Methods

Study design and approvals

The Bispebjerg Study of Diurnal Variation has been described previously [5] and is summarized here. This prospective time-series analysis involved a 24-h hospital stay under standardized conditions. Participants spent 15 h awake in ordinary daylight or room light (mean light intensity 219 lx) and were provided with a 9 h opportunity to sleep in the dark (mean 0.04 lx), from 11:00 PM to 8:00 AM. During the day the, the participants were prohibited from napping but permitted low-intensity activities such as walking, television watching and reading [5].

Standardized isocaloric, low-fat, sugar-free meals were provided at 9:30 AM, 1:00 PM and 7:00 PM. Water intake was not restricted. Meals were identical for all participants, without individual caloric adjustments or measurement of food intake. Meal composition has been described previously [6] and is presented in Fig. 1. The participants fasted for 11 h before the start of the study [5].

Fig. 1.

Overview of the Experimental Workflow. a Study design detailing blood sampling every 3 h over 24 h under standardized conditions, including controlled meal composition, caloric content, and a scheduled sleep opportunity during the night. Participants consumed standardized mixed meals: breakfast (846 kcal, carbohydrate 47.8 energy percent (E%), protein 16.3 E%, fat 35.1 E%), lunch (636 kcal, carbohydrate 58.8 E%, protein 19.9 E%, fat 21.1 E%), and dinner (841 kcal, carbohydrate 45.3 E%, protein 19.6 E%, fat 35.1 E%). b Workflow schematic illustrating plasma isolation, enzymatic digestions, peptide separation via liquid chromatography, mass spectrometry analysis (LC–MS/MS), and statistical analysis for identification of proteins exhibiting significant diurnal rhythmicity

The study was conducted in 2008 at the Department of Clinical Biochemistry, Bispebjerg Hospital, Copenhagen, Denmark. It was approved by the local independent ethics committee (protocol number H-B-2008-011) and the Danish Data Protection Agency (journal number 2008–41-1821). It was conducted according to the Helsinki Declaration, with all participants signing a written informed consent. The trial has been retrospectively registered at clinicaltrials.gov with the identifier NCT06166368.

Participants

Eligible participants were healthy men aged 18–45 years with a regular sleep–wake cycle and hemoglobin concentration greater than 8.0 mmol/L. Exclusion criteria included an acute or chronic illness, use of medication within the past 30 days, regular tobacco use, night-shift work, recent time zone shift (travel), or increased alcohol consumption or smoking within the last 14 days prior to the study. Strict inclusion/exclusion criteria were implemented to ensure homogeneity and minimize participant variation.

Participants were recruited through advertisements at the Faculty of Health Science, University of Copenhagen. A total of twenty-four healthy Caucasian male volunteers aged 20–40 (mean age 26 years) were included [5].

Procedures

Blood sampling was performed every three hours over a 24-h period, beginning at 9:00 AM, for a total of nine collections (Fig. 1). Samples were drawn from the cubital vein in alternating arms at each time point using minimal tourniquet application and collected into serum clot activator tubes coated with microscopic silica particles (Greiner Bio-one, Frickenhausen, Germany). Tubes were centrifuged, plasma was isolated, and immediately stored at − 80 °C until analysis.

During the wake period, blood was drawn following a 10-min rest, with participants seated at a 45-degree angle in a hospital bed, legs extended and positioned horizontally. During the sleep opportunity period, blood samples were collected with minimal disturbance using low-intensity red light while participants remained in a supine position.

Additional measurements included blood pressure, pulse and self-reported height and weight. The body mass index (BMI, kg/m²) was calculated from height and weight. Light intensity was measured using the RS 180–7133 lx meter (RS Components, Corby, United Kingdom) [5].

Proteomic analysis

Each plasma sample was aliquoted into a 96-well plate and prepared on an Agilent Bravo Liquid Handling Platform. Samples were diluted 1:10 with lysis buffer (1 M Tris, 0.5 M Tris(2-carboxyethyl)phosphine (TCEP) and 0.5 M Chloroacetamide (CAA) in H₂O) and incubated at 95 °C for 10 min. After cooling to RT, a Trypsin/LysC mixture (1 µg to 100 µg protein) was added and incubated for 4 h at 37 °C at 1000 rpm. The enzymatic reaction was quenched by adding 64 µl of 0.2% TFA. The samples were loaded onto Evotips according to the manufacturer’s recommendations (Evosep Biosystem, Denmark). Briefly, Evotips were prepared by washing with buffer B (100% acetonitrile (ACN), 1% formic acid (FA)), activated by soaking in isopropanol, and equilibrated with 20 µl of buffer A (1% FA) before loading the samples. Evotips were then loaded with 250 ng of peptides per sample, followed by a wash with 20 µl of buffer A. Each step was followed by a 1-min centrifugation at 700 g to facilitate liquid passage. Finally, the Evotips were stored in buffer A to prevent drying.

LC–MS/MS analysis was performed using an Orbitrap Astral mass spectrometer (Thermo Scientific) coupled to an Evosep One system (Evosep Biosystem, Denmark). Peptide separation was carried out on a commercial 8 cm analytical ‘Performance’ column (EV1109, Evosep Biosystem, Denmark) using the predefined 60 samples per day method (21-min gradient) and analyzed in data-independent acquisition mode. The mass spectrometer operated in positive mode. Full MS spectra (380–980 m/z) were acquired using the Orbitrap analyzer with a resolution of 240,000 at 200 m/z. Precursor ions were isolated with an automatic gain control (AGC) target of 500% (5e6 charges) and a maximum injection time (maxIT) of 3 ms. In parallel to the full MS scan, fragment spectra of 200 consecutive windows (3-Th width) within the 380–980 m/z precursor mass range were recorded using the Astral analyzer operating at the resolution of 80,000. Precursor ions were isolated with an AGC target of 500% (5e4 charges) and a maxIT of 5 ms, then fragmented at 25% normalized collision energy.

Statistical analysis

Data processing

Raw mass spectrometry data were initially processed using DIA-NN (version 1.9) in a data-independent acquisition (DIA) search [7]. Further data processing was conducted using python, where a stringent filtering approach was applied to address missing data: (1) samples with a low protein count—defined as values below 1.5 times the interquartile range (IQR) from the 25th percentile of the combined distribution—were excluded, and (2) proteins missing in more than 40% of samples were removed. Log₂ transformation was applied to the data, and remaining missing values were imputed using the variational autoencoder implemented in PIMMS [8].

Data analysis

Differential protein abundance according to diurnal rhythm was assessed using the CosinorPy python package [9] with multiple hypothesis correction applied using the Benjamini–Hochberg method. Adjusted p-values less than 0.05 were considered statistically significant. Protein abundances and p-values were visualized using heatmaps. Agglomerative hierarchical clustering with single linkage was performed to identify clusters of proteins with similar diurnal patterns. Heatmaps, combined with clustering dendrograms and sankey plots of Reactome Pathways, were used to highlight changes in significant proteins, protein clusters, and their associated biological relevance [10]. Protein abundances in heatmaps were z-scored.

To externally validate our findings, we retrieved routine clinical laboratory data from the laboratory information database LABKA, encompassing timestamped measurements of albumin, activated partial thromboplastin time (aPTT), and prothrombin time (PT/INR). Data extraction was restricted to samples collected across hospitals within the Capital Region of Denmark from April 1 to April 7, 2025. Diurnal rhythmicity was subsequently assessed via Cosinor analysis, as described previously for the proteomic data.

Enrichment analysis

Tissue and pathway enrichment analyses were conducted for identified protein clusters individually using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) version, 2024 [11, 12]. We provided a list of significant proteins under their official gene symbols (e.g., B2M, ALB) and used Homo sapiens (9606) as the background species. Tissue enrichment was performed with the UP_TISSUE database, excluding pathological, fetal, fluid-based (e.g., serum, plasma, cerebrospinal fluid), and placental enrichments. Pathway enrichment used the REACTOME_PATHWAY database, and we further supported these findings with Gene Ontology Biological Processes (GOTERM_BP_FAT) analysis. Enrichments with Benjamini–Hochberg–adjusted p-values below 0.05 were deemed significant.

Outcome

The primary outcome was to assess diurnal variation in plasma protein levels through blood samples collected every 3 h over 24 h from all participants. The secondary outcome was to determine whether diurnal variations in plasma proteins have diagnostic relevance and potential implications for clinical decision-making.

The study design, controlled environment, and strict inclusion/exclusion criteria minimized the influence of zeitgebers and confounders. Steps to reduce bias included standardized sample collection protocols to minimize measurement and performance bias.

Role of the funding source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

From the 24 healthy individuals (age 20–40 years, mean 26.0 ± 5.2 years), a total of 216 plasma samples were collected over 24 h. Eight samples (3.7%) were excluded due to insufficient quality, defined by low protein count (below 1.5 × IQR from the 25th percentile), leaving 208 high-quality samples available for final proteomic analysis. Baseline characteristics of participants are presented in Table 1 [1].

Table 1.

Baseline Characteristics of Study Participants (n = 24 healthy men)

Variable	Mean (SD)	Range (95% CI)	Clinically Accepted Normal Ranges
Age (Years)	26 (5.2)	23.8–28.2
Weight (kg)	76.6 (6.6)	73.8–79.4
Height (m)	1.83 (0.05)	1.82–1.85
Body mass index, BMI (kg/m2)	22.9 (1.6)	22.2–23.6	18.5–24.9
Pulse (beats per minute)	66 (9.9)	61.8–70.2	60–100
Systolic blood pressure (mmHg)	128 (10.2)	123.7–132.3	≤ 120
Diastolic blood pressure (mmHg)	68 (8.2)	64.5–71.5	≤ 80
Hemoglobin concentration (mmol/L)	9.2 (0.6)	8.95–9.45	8.3–10.5 (men)

Number of participants: 24 healthy men

Participant characteristics and measurements. SD, standard deviation; CI, confidence interval

Data are presented as mean (standard deviation, SD) with corresponding 95% confidence interval (95% CI). BMI (body mass index) was calculated from self-reported weight and height [5]

A total of 832 unique plasma proteins were identified, with a median of 533 proteins per sample (Fig. 2a). After stringent quality filtering, 523 proteins (62.9%) remained for diurnal rhythm analysis. Using CosinorPy Rhythmometry with z-scoring and stringent Benjamini–Hochberg correction, we identified 138 proteins (26.4%) exhibiting significant diurnal rhythmicity (as shown in Supplementary Table 5) (mean adjusted p-value: 0.0093, SD: 0.015; 95% CI: 0.0068–0.012). Proteins with adjusted p-values below 0.05 following Benjamini–Hochberg correction were considered significantly rhythmic. The variation in protein levels across the day and across individuals (coefficient of variation) correlated significantly both with the residual standard error from the CosinorPy analysis and with the amplitude of the rhythmic peak (Pearson’s r = 0.79 (p-value = 8.121e-113) and r = 0.76 (p-value = 9.23e-98), respectively) (Fig. 2b). Additionally, the proteins exhibiting diurnal rhythmicity were overrepresented among proteins with high coefficient of variation (Odds Ratio = 1.755, p-value = 0.006) and high amplitude (Odds Ratio = 4.67, p-value = 3.91e-13).

Fig. 2.

Key Metrics of Proteomic Data Quality. a Distribution of identified plasma proteins per sample. The boxplot indicates median proteomic coverage (533 proteins per sample) and individual data points reflecting variability among 216 samples collected from 24 healthy male participants. Eight of the 216 samples were excluded due to a low protein count (protein count < 457). b Bubble plots depicting the relationship between the coefficient of variation (%) and residual standard error or amplitude from the CosinorPy analysis for plasma proteins exhibiting significant diurnal rhythmicity. Each dot represents a quantified protein (n = 523), with significantly rhythmic proteins (n = 138) highlighted. Significance is showed as point sizes (-log10 adjusted p-values (Benjamini–Hochberg correction; significance threshold indicated at adjusted p < 0.05)). Coefficient of variation represents the variation across samples and timepoints, residual standard errors represent the distance from the fitted rhythmic regression to the actual protein measurements. The amplitude represents the magnitude of diurnal oscillation of the fitted rhythmic regression

Hierarchical clustering analysis revealed two distinct diurnal clusters, an afternoon cluster (Cluster A; 91 proteins) and an early morning cluster (Cluster B; 47 proteins) (Fig. 3).

Fig. 3.

Protein Abundance Heatmap with Tissue and Pathway Enrichments. Hierarchical clustering of diurnal‐regulated plasma proteins (z‐scored abundance) sampled every three hours over 24 h revealed two distinct clusters based on peak time (acrophase). Cluster A (91 proteins) peaks mainly in the afternoon, with a median acrophase at 16.5 h (interquartile range [IQR] 12.9–17.1), whereas Cluster B (47 proteins) peaked earlier, with a median acrophase at 3.3 h (IQR 2.0–5.8). Tissue enrichment is shown on the right, highlighting liver and platelet origins, among others, while Reactome pathway enrichment show 19 functional clusters of protein groups related to processes such as hemostasis, immune regulation, and metabolic pathways. Note: Acrophase values are absolute 24-h clock times (0–24 h), not times relative to sampling start

Tissue enrichment analysis (DAVID) annotated 129 of 138 rhythmic proteins (93.5%), and was performed separately for the two distinct clusters: afternoon and early morning. Cluster A predominantly originated from the liver (66 proteins, adjusted p = 5.8 × 10⁻¹⁷) and platelets (34 proteins, adjusted p = 2.3 × 10⁻²⁶), but additional enriched tissues included lymphoblasts (16 proteins), T-cells (10 proteins), skeletal muscle (14 proteins), keratinocytes (7 proteins), skin (21 proteins), tongue (13 proteins), fibroblasts (6 proteins), erythrocytes (3 proteins), and lungs (23 proteins), as shown in Supplementary Table 3. Cluster B was significantly enriched only in the liver (27 proteins; adjusted p = 1.4 × 10⁻⁵) (Fig. 3 and Fig. 4a).

Fig. 4.

Protein Count in Tissue and Pathway Enrichments. a Bar plot illustrating the distribution of diurnal proteins across enriched tissue types, underscoring the predominance of liver‐ and platelet‐associated proteins. b Bar plot quantifying the number of diurnal proteins represented in each of the 19 pathway clusters

Tissue enrichment analysis provides insight into potential sources or tissues of origin for the identified rhythmic plasma proteins. Enriched tissues likely indicate either primary sites of synthesis and secretion into circulation (e.g., liver-derived proteins) or tissues known to release proteins upon activation or damage (e.g., platelet-derived proteins). Intracellular or structural tissues identified may reflect turnover, leakage, or shedding processes contributing to plasma composition.

Reactome pathway enrichment (DAVID) annotated 79 proteins (86.8%) in Cluster A, and 32 proteins (68.1%) in Cluster B, and revealed extensive diurnal rhythmicity across multiple clinically and biologically critical pathways. Among the most significant enrichments were pathways associated with platelet function and coagulation, such as “Platelet degranulation”, “Response to elevated platelet cytosolic Ca²⁺”, “Platelet activation, signaling, and aggregation”, and the broader"Hemostasis"pathway. Platelet activation and hemostasis was the only common denominator between Cluster A and Cluster B, although it was more significantly enriched in Cluster A (adjusted p = 5.2 × 10⁻¹⁴ to 2.7 × 10⁻¹⁹) compared to Cluster B (adjusted p = 9.4 × 10⁻³ to 1.1 × 10⁻⁴) (Fig. 3 and Fig. 4b).

Cluster A was significantly enriched in pathways involving immune system function and response, including “Neutrophil degranulation” and the broader “Innate immune system” processes. Additionally, enrichments in pathways associated with “Rho GTPase Signal Transduction”, “Cell–Matrix, Cell–Cell Adhesion, and Cytoskeletal Remodeling”, “Axon Guidance and Neural development”, “Oncogenic RAS/RAF-MAPK Signaling and Integrin Signaling”, “Interleukin-12 and JAK/STAT Cytokine Signaling”, “Glucose Metabolism” including “Glunoneogenesis” and “Glycolysis”, and more (Supplementary Table 4).

Importantly, Cluster B enrichment extended beyond coagulation-related pathways to include significant diurnal regulation in protein metabolism and growth signaling processes, such as “Regulation of Insulin-like Growth Factor transport and uptake by IGFBPs”, “Post-translational protein phosphorylation”, and “Metabolism of proteins”. Notably, pathways involving lipoprotein metabolism and transport also displayed significant rhythmicity, including “Plasma lipoprotein assembly, remodeling, and clearance”, “Chylomicron assembly”, and “HDL remodeling”. Additionally, enrichments in pathways associated with the coagulation cascade and specifically, fibrin clot formation were significant in the early morning peak cluster of proteins.

The Gene Ontology Biological Processes (GOBP) enrichment analysis (DAVID) matched 89 proteins from Cluster A (97.8%) and 46 proteins from Cluster B (97.9%), and corroborated the findings described earlier, providing further biological insight into diurnal regulation across diverse physiological systems. Both clusters showed particularly robust enrichment in processes related to coagulation and wound healing, although it was again stronger in Cluster A (Data not shown).

Collectively, these detailed enrichment analyses using Reactome and GOBP databases emphasize the wide-ranging impact of diurnal biology on fundamental clinical and biological functions.

The acrophase—defined as the actual peak time in clock hours (0–24 h), representing the time of day at which each protein reaches its maximum abundance—was used to characterize the temporal distribution of diurnally-regulated plasma proteins. Cluster A proteins exhibited a median acrophase of 16.5 h (IQR: 12.9–17.1 h), corresponding approximately to late afternoon. Cluster B proteins peaked earlier in the day, with a median acrophase of 3.3 h (IQR 2.0–5.8 h), representing early morning hours (Fig. 5).

Fig. 5.

Temporal Dynamics of Diurnal Plasma Protein Clusters. a Boxplot representation illustrating the median protein abundance (amplitude, normalized z-score) with interquartile ranges over 24 h, separated into two distinct clusters. Cluster A exhibits peak abundance during the afternoon hours, whereas Cluster B peaks in the early morning. b Line graph displaying individual protein trajectories across 24 h, grouped into the same two clusters. This visualization emphasizes the consistent but opposite diurnal fluctuations in protein levels, reinforcing distinct temporal dynamics for Cluster A and Cluster B. Selected representative proteins within each cluster are indicated. c) Polar histogram illustrating the distribution of peak times (acrophases) for diurnal plasma proteins grouped into two clusters. Cluster A (Afternoon, orange) shows a concentration of peak expression in the late afternoon, while Cluster B (Early morning, purple) peaks earlier in the day. The radial axis indicates the number of proteins peaking at each time point, and the circular layout represents the 24-h cycle. Note: Acrophase values reflect actual peak clock time in hours (0–24 h), not relative to a reference point in the diurnal cycle

Diurnally-regulated proteins showed a mean amplitude of 0.172 (SD: 0.0107; 95% CI: 0.151–0.194), but interestingly, Cluster A had a significantly higher amplitude with a mean of 0.215 (SD: 0.132; 95% CI: 0.19–0.24) versus 0.091 (SD: 0.05; 95% CI: 0.076–0.11) in Cluster B. No significant changes in protein abundance were observed in relation to meal intake (Supplementary Table 1).

Through cross-referencing our significant rhythmic proteins against the standard clinical diagnostic assay preference list from the Danish regional health authority (Region H, Sundhedsplatformen)—which reflects clinical practice guidelines and diagnostic standards currently implemented across hospitals in the Capital Region of Denmark—we identified 36 proteins (26.1% of all significantly rhythmic proteins) currently used as biomarkers in routine clinical practice (Table 2). These clinically important biomarkers span multiple essential diagnostic categories, including coagulation factors (e.g., fibrinogen alpha chain [FGA], coagulation factor V [F5], and protein C [PROC]), markers of liver and kidney function (albumin [ALB], cystatin C [CST3]), inflammatory biomarkers (calprotectin subunits S100A8 and S100A9), endocrine-related proteins (insulin-like growth factor 1 [IGF1]), proteins indicative of cardiac and skeletal muscle injury (creatine kinase M-type [CKM], lactate dehydrogenase A [LDHA]), and a marker for acute pancreatitis (amylase [AMY2A]). Additionally, critical immune-related diagnostic markers such as complement proteins, human leukocyte antigen B (HLA-B), and immunoglobulins also demonstrated clear diurnal rhythmicity.

Table 2.

Clinical Relevance of Diurnally-Regulated Proteins in Human Plasma

Clinical Test	Diurnally-Regulated Protein(s)	Clinical Significance	Mean Acrophase (h) ± SD
aPTT	F11; F5; F9; PROC; SERPINC1; FGA	Coagulation status; Bleeding disorders; Liver function monitoring	4.6 ± 2.3
PT (INR)	F5; FGA	Coagulation status; Liver synthetic function assessment	5.2 ± 1.1
D-Dimer	FGA	Thrombosis diagnosis; Disseminated intravascular coagulation (DIC)	4.5 ± –
Fibrinogen	FGA	Coagulation status; Acute-phase inflammatory response	4.5 ± –
Albumin	ALB	Nutritional status; Liver function; Chronic illness monitoring	0.1 ± –
Cystatin C	CST3	Kidney function; Glomerular filtration rate estimation (eGFR)	1 ± –
Calprotectin	S100A8/S100A9 (complex)	Systemic inflammation marker; Leukocyte activation	12.9 ± 0.1
Beta-2-Microglobulin	B2M	Tumor marker (Myeloma, CLL, lymphoma); Kidney function assessment	3.6 ± –
Complement Activity tests	(CH50 test): C3; C1r (AH50 test): C3; CFD; CFHR1	Complement system evaluation; Immunodeficiency diagnosis	(CH50) 4.6 ± 2.2 (AH50) 6 ± 5.6
Complement C3	C3	Complement activation; Inflammatory/autoimmune disorders	3.1 ± –
Complement C1r	C1r	Classical complement pathway activity	6.1 ± –
Prostaglandin D-Synthase	PTGDS	Inflammation marker; Sleep regulation (research)	2.7 ± –
Insulin-Like Growth Factor I	IGF1	Growth disorders; Nutritional/endocrine assessmnet	6.3 ± –
Creatine Kinase	CKM	Muscle injury diagnosis; Rhabdomyolysis; Myopathy	13.8 ± –
Glyceraldehyde-3-phosphate dehydrogenase	GAPDH	Cellular damage; Hemolysis	17.2 ± –
Heparin-PF4-IGG (HIT)	PF4	Heparin-induced thrombocytopenia diagnosis	18.7 ± —
Erythrocyte Transketolase Activity Test	TK	Vitamin B1 (Thiamine) deficiency diagnosis; Nutritional assessment	18.9 ± —
Protein C	PROC	Thrombophilia evaluation; Coagulation disorders diagnosis	5.7 ± –
HLA-AB	HLA-B	Transplant compatibility; Immunogenetic profiling	7.1 ± –
Myoglobin	MB	Acute myocardial infarction (AMI); Rhabdomyolysis	9.5 ± –
LDH	LDHA	Tissue injury; Hemolysis; Prognostic monitoring	9.9 ± –
Total IgA	IGHA2	Immunodeficiency diagnosis; Immune status monitoring	19.5 ± –
Free Kappa Chains (Ig)	IGKV2D-29; IGKV2-28; IGKV2-29	Plasma cell disorder diagnosis (Myeloma, MGUS); Therapeutic monitoring	8.5 ± 8.1
Free Lambda Chains (Ig)	IGLV3-25; IGLV3-19; IGLV2-11; IGLV3-1	Plasma cell disorder diagnosis (Myeloma, MGUS); Therapeutic monitoring	2.1 ± 1
Free Kappa/Lambda Chains (Ig) Ratio	IGKV2D-29; IGKV2-28; IGKV2-29; IGLV3-25; IGLV3-19; IGLV2-11; IGLV3-1	Plasma cell disorders diagnosis; Disease monitoring; Clonality analysis	4.9 ± 5.9
CD56 Flow Cytometry (whole blood)	NCAM1	NK/T-cell malignancy diagnosis; Immunophenotyping	22.1 ± –
Amylase	AMY2A	Acute pancreatitis diagnosis; Pancreatic function assessment	19.4 ± –
Antithrombin	SERPINC1	Thrombosis risk; Coagulation inhibitor deficiency	2.2 ± –
Coagulation Factor XI	F11	Hemophilia C; Bleeding disorder evaluation	7.5 ± –
Coagulation Factor V	F5	Factor V deficiency diagnosis; Thrombophilia evaluation	6 ± –
Transferrin-Receptor Fragment	TFRC	Iron deficiency diagnosis; Iron metabolism assessment	18.5 ± –
Coagulation Factor IX	F9	Hemophilia B; Factor IX Deficiency; Vitamin K deficiency	1.8 ± –

The listed blood tests correspond to clinical routine assays according to the standardized regional clinical laboratory test preference list in Denmark (EPIC system, Region H, Denmark). Diurnally-regulated proteins linked to each test are indicated alongside descriptions of their diagnostic, prognostic, or monitoring significance in clinical practice. Note: Acrophase values are absolute 24-h clock times (0–24 h), not times relative to sampling start

aPTT, activated partial thromboplastin time; PT, prothrombin time; AMI, acute myocardial infarction; MGUS, monoclonal gammopathy of undetermined significance; eGFR, estimated glomerular filtration rate; HIT, heparin-induced thrombocytopenia; CLL, chronic lymphocytic leukemia; LDH, lactate dehydrogenase; Ig, immunoglobulin

Discussion

In this study, we demonstrate that one fourth of the human plasma proteome display diurnal rhythmicity. Using a rigorous, unbiased mass spectrometry-based proteomics approach in a cohort of healthy young individuals under highly controlled conditions, we identified 138 proteins with distinct diurnal patterns.

In line with our final analyses, we identified two distinct diurnal clusters of plasma proteins: one peaking in the afternoon (Cluster A; 91 proteins, median acrophase of 16.5 h) and another in the early morning (Cluster B; 47 proteins, median acrophase of 3.3 h). While Cluster A showed a more diverse tissue origin—including platelets, liver, skeletal muscle, immune cells, and several others—Cluster B was significantly enriched only in the liver. Functionally, both clusters shared platelet activation and hemostasis, although these processes were more strongly enriched in Cluster A. Notably, Cluster B alone encompassed insulin‐like growth factor regulation, protein metabolism, and lipoprotein metabolism, suggesting these pathways peak during the early morning or night hours. By contrast, Cluster A included prominent pathways in innate immunity, Rho GTPase signaling, integrin signaling, oncogenic RAS/RAF‐MAPK signaling, and carbohydrate metabolism (gluconeogenesis and glycolysis). Intriguingly, the afternoon‐peaking proteins (Cluster A) also exhibited a higher rhythmic amplitude (~ 0.22) compared to Cluster B (~ 0.09), implying more pronounced day‐time oscillations.

The tissue enrichment analysis highlights liver and platelet origins for many of the identified rhythmic plasma proteins, suggesting these tissues are prominent contributors to the diurnal proteome. These observations likely reflect a combination of rhythmic synthesis, regulated secretion, and possibly diurnal variation in cellular turnover or shedding processes within these tissues. Collectively, these mechanisms underscore the complex biological regulation of plasma protein rhythms and emphasize tissue-specific temporal control.

These rhythmic fluctuations were observed under standardized meal conditions and did not consistently align with postprandial periods, suggesting that food intake was not the primary driver of the observed variation. Critically, 36 of these diurnal‐regulated proteins (26% of the total rhythmic proteins) are already measured in routine clinical practice, as reflected in the Danish guidelines, underscoring the clinical importance of recognizing diurnal patterns in widely used biomarkers and highlights a potential role for diurnal‐informed testing protocols.

Our findings are consistent with prior studies but also extend the understanding of diurnal plasma proteomics significantly. A recent study (2024), employing a SomaScan aptamer-based methodology, observed circadian rhythmicity of 15% of proteins [13], fewer than our observed 26.4%. Methodological differences, specificially the broader, unbiased coverage provided by mass spectrometry, likely explain the higher proportion of rhythmic proteins identified in our study. Notably, similar to our findings, they also reported protein oscillations primarily peaking in the early morning and afternoon hours. Additionally, circadian regulation of hemostatic proteins has previously been observed in ELISA-based studies [14] and is further supported by studies using constant routine protocols demonstrating circadian variation in platelet activation [15]. Our comprehensive proteomic analysis confirms that many proteins involved with the hemostasis system have a diurnal rhythm and further expands the understanding by identifying rhythmic proteins in diverse biological processes, emphasizing the broader relevance of diurnal variation in clinical diagnostics and protein research.

Our unbiased MS-based proteomics method significantly extends beyond previous targeted studies, providing an innovative and robust framework for biomarker discovery and research. Unlike targeted proteomic techniques such as ELISA or aptamer-based assays (e.g. SomaScan), MS-based proteomics offers direct identification and quantification of peptide sequences, thereby increasing analytical specificity by eliminating cross-reactivity and reducing specificity bias. Recent technological advancements in MS-based proteomics have markedly enhanced sensitivity, robustness, and throughput, enabling the comprehensive and precise characterization of complex biological systems. Consequently, our methodological approach provides a more complete and unbiased exploration of diurnal rhythmicity in plasma proteins, facilitating discoveries that may have been overlooked using targeted approaches, and setting new standards for diurnal biomarker identification and clinical diagnostics [16].

These rhythmic fluctuations of biological processes support essential physiological functions such as hemostasis, energy metabolism, and immune regulation. Dysregulation of these rhythms can predispose individuals to adverse health effects [17] such as metabolic [18], inflammatory [19], psychiatric [1], and cardiovascular diseases [20], underscoring the clinical importance of accurately characterizing temporal protein dynamics.

Clinically, our results suggest significant implications for biomarker discovery and diagnostic accuracy. Ignoring diurnal rhythms could lead to inconsistent biomarker validation, inaccurate reference intervals, and suboptimal therapeutic monitoring. For instance, diurnal variation in hemostatic proteins may influence the optimal timing of diagnostic tests or medication administration for conditions like atrial fibrillation if the risk of thrombotic events are found to be varying throughout the day. A prime illustration of diurnal‐aware testing is serum cortisol measurement. Because cortisol normally peaks in the early morning and reaches a nadir at night, clinicians measure morning serum cortisol to screen for adrenal insufficiency, whereas late‐night salivary cortisol helps diagnose hypercortisolism (Cushing’s syndrome). This timing ensures more accurate assessment of hypothalamic–pituitary–adrenal axis function [21]. Building upon these established principles, our findings suggest that many other clinically used biomarkers with significant diurnal oscillations may also benefit from standardized sampling times or time‐adjusted reference intervals to improve diagnostic precision and therapeutic outcomes. Proteins such as amylase, found in our study to have a diurnal rhythm and is commonly used as a biomarker to assess acute and chronic pancreatitis [22], and cystatin C, which can be used to estimate the glomerular filtration rate and monitor kidney function [23], may benefit from the use of time-sensitive reference ranges to improve diagnostic precision and patient management.

Despite efforts to standardize light exposure, meal timing and composition, and participant activity, some zeitgeber influences inevitably remained. We therefore interpret the observed protein rhythms as diurnally regulated, potentially reflecting intrinsic circadian signals, external cues such as feeding, physical activity and sleep, or both. Future constant-routine or behavioral-modelled protocols will be necessary to separate these componentrs more precisely.

Although integrating diurnal-aware sampling into clinical diagnostics introduces practical and logistical challenges, such as more complex scheduling and adjustments to reference intervals, the potential improvements in diagnostic accuracy and treatment optimization are substantial. Thus, future guidelines should consider incorporating time-of-day recommendations for biomarkers with pronounced rhythmicity to enable more precise and individualized medical care.

Strengths and limitations

Our study’s strengths include a thoroughly controlled experimental environment, standardized sampling conditions, and advanced mass spectrometry methods minimizing potential biases inherent in antibody- or aptamer-based assays. Furthermore, robust statistical analysis with the CosinorPy rhythmometry, including z-scoring and correction for multiple testing using the Benjamini–Hochberg procedure, strengthened confidence in the rhythmicity estimates and reduced the risk of false discovery.

However, several limitations exist. Firstly, generalizability is limited by our homogeneous study population of healthy young men with regular lifestyles, which may not reflect the diurnal variability seen in clinical populations with broader demographic and behavioral diversity. Secondly, the exclusion of women is a notable limitation, particularly given that sex-specific differences in circadian regulation of liver gene expression have been demonstrated in mice [24], and that female mice have shown greater circadian resilience under disruptive conditions [25]. While these findings may not fully translate to humans, they underscore the need for future studies to include both sexes to evaluate potential sex-based differences in plasma diurnal proteomics. Additionally, out of a total of 216 plasma samples collected, eight (3.7%) were excluded due to low protein count, potentially introducing minor bias. Nevertheless, the stringent filtering and robust statistical methodology mitigate the impact of these missing data points, preserving the overall validity of our findings.

Further limitations include the absence of objective measurements of actual sleep duration; the provided interval represents a sleep opportunity rather than verified sleep length. Furthermore, because we did not statistically model post-prandial effects, we cannot exclude that feeding, together with other diurnal factors such as light exposure or physical activity, may influence the timing or amplitude of of a subset of proteins. Importantly, the true circadian variation may be masked by the influence of external zeitgebers present in our study protocol, meaning that some observed diurnal rhythms might differ under strictly controlled circadian conditions. Finally, the observed rhythmicity patterns identified here may not be present in a clinical routine protocol setting, and tissue enrichment analyses reflect likely but not definitive proven origins of rhythmic proteins.

Future research

Our findings open several important avenues for future research. Further studies should expand investigations into populations beyond healthy young men, including diverse age groups, females, and individuals with varying chronotypes or circadian disruptions (e.g., shift work, metabolic disorders). Exploring the impact of sleep quality and actual sleep duration, as well as detailed postprandial effects on protein rhythmicity, would enhance understanding of the mechanisms underlying diurnal regulation. Additionally, prospective clinical trials evaluating the practical integration of diurnal-aware blood sampling protocols into routine diagnostics could clarify the impact on diagnostic accuracy, therapeutic effectiveness, and patient outcomes. Finally, mechanistic studies are needed to reveal the molecular pathways underpinning circadian and diurnal oscillations of plasma proteins, potentially revealing novel therapeutic targets and biomarkers. As our findings were obtained under tightly controlled experimental conditions, future studies should examine how real-world clinical rhythms—such as exposure to artificial lighting, irregular dietary habits, and shift-work—affect plasma protein rhythmicity in more variable environments.

Conclusion

In conclusion, our comprehensive proteomic analysis reveals extensive diurnal regulation within the human plasma proteome, identifying rhythmic fluctuations in 26% of the human plasma proteome measured under controlled conditions. By using an unbiased mass spectrometry approach, we accurately characterize these fluctuations, overcoming analytical bias inherent in previous targeted studies and highlighting oscillatory proteins with potential clinical relevance. Recognizing these diurnal variations through standardized sampling protocols or the establishment of time-sensitive reference intervals could significantly improve diagnostic accuracy, therapeutic monitoring, and the design of future biomarker studies.

Future research should further investigate diurnal variation in broader and more diverse clinical populations, unravel underlying molecular mechanisms, and evaluate how time-aware practices can be pragmatically implemented in routine care. Integrating diurnal biology into clinical diagnostics and biomarker development represents a pivotal step toward more precise, personalized, and time-informed medicine.

Author contributions

NJWA conceived the study and initiated the investigation of circadian rhythms in the human plasma proteome. HPS and HLJ conducted the clinical experiments and HPS provided the plasma samples. CR prepared the samples for analysis, and GZP performed the mass spectrometry analysis. CYCY conducted the preliminary data analysis. ABN carried out the statistical analysis, and together with EMSJ, performed the final data analysis and interpretation. NJWA, CYCY, HPS, JH, and ABN contributed with intellectual input and critical discussion. EMSJ wrote the manuscript. All authors reviewed and approved the final version of the manuscript.

Funding

Open access funding provided by Copenhagen University. Open access funding provided by Copenhagen University. NJWA is supported by the European Foundation for the Study of Diabetes Future Leader Award (NNF21SA0072746), Independent Research Fund Denmark (1052-00003B, 10.46540/4308-00056B, 10.46540/4285-00131B), Novo Nordisk Foundation (NNF23OC0084970, NNF19OC0055001 and NNF24OC0088402). The Novo Nordisk Foundation Center for Protein Research is supported financially by the Novo Nordisk Foundation (NNF14CC0001).

Data availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [26] partner repository with the dataset identifier PXD066727.

Code availability

The jupyter notebooks and interactive plots are available at https://github.com/annelauralab/DiurnalPlasmaProteome.git.

Declarations

Competing interests

NJWA has received funding from and served on scientific advisory panels and/or speakers’ bureaus for Boehringer Ingelheim, MSD/MERCK, Roche, Novo Nordisk and Mercodia. The remaining authors have no conflicts of interest to declare.