Dynamic changes in cardiac autonomic function persist in the postacute phase after SARS-CoV-2 infection in a hamster model of COVID-19 Authors: Ting Liu, Deepthi Ashok, Andrew Pekosz, Jason Villano, Brian O’Rourke Published: 2026 ABSTRACT AbstractBackgroundAutonomic nervous system (ANS) dysfunction is a central feature of long coronavirus disease syndrome, yet little is known about how it develops during and after severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection.ObjectiveThis study aimed to characterize the time course of dynamic changes in linear and nonlinear heart rate variability measurements as indices of ANS function for up to 2 months after infection.MethodsHamsters implanted with radiotelemetry devices were inoculated with SARS-CoV-2. Electrocardiograms and subpleural pressures were recorded before infection, and for 56 days postinfection (DPI). Heart rate variability and respiratory sinus arrhythmia were analyzed and compared with the preinfection baseline or mock-infected controls. Ruxolitinib or mitochondrially targeted 2,2,6,6-tetramethylpiperidine-1-oxyl was administered from 4 days before infection to 10 DPI to assess the effects of Jak/Stat inhibition or antioxidant treatment, respectively.ResultsANS activity follows 3 phases after infection: (1) a pronounced increase in parasympathetic activity relative to sympathetic activity during the acute phase of infection (1–4 DPI), (2) a decrease in overall ANS activity in the subacute phase (7–14 DPI), and (3) re-emergence of elevated sympathetic and parasympathetic inputs that continue to increase during the postacute phase (28–56 DPI). The conclusions are corroborated by changes in respiratory sinus arrhythmia analysis and acute pharmacologic inhibition of cholinergic and β-adrenergic signaling. Postacute ANS remodeling was suppressed by early inhibition of innate immune signaling or mitochondrial oxidative stress.ConclusionCardiac dysautonomia after SARS-CoV-2 infection is phase dependent. Inhibiting early immune activation or mitochondrial oxidative stress during the acute phase offers a strategy to suppress postacute ANS remodeling that could contribute to long coronavirus disease syndrome.Keywords: Bradycardia, Arrhythmia, Long COVID syndrome, Interferon, Innate immunity, Heart rate variability, Autonomic dysfunction Key Findings ▪SARS-CoV-2 infection induces a triphasic pattern of cardiac autonomic dysfunction in hamsters: parasympathetic dominance in the acute phase (1–4 days postinfection [DPI]), autonomic exhaustion in the subacute phase (7–14 DPI), and sustained parasympathetic hyperactivity in the postacute phase (28–56 DPI), as shown by heart rate variability (HRV) and respiratory sinus arrhythmia analyses. SARS-CoV-2 infection induces a triphasic pattern of cardiac autonomic dysfunction in hamsters: parasympathetic dominance in the acute phase (1–4 days postinfection [DPI]), autonomic exhaustion in the subacute phase (7–14 DPI), and sustained parasympathetic hyperactivity in the postacute phase (28–56 DPI), as shown by heart rate variability (HRV) and respiratory sinus arrhythmia analyses. ▪Early treatment (−4 to +10 DPI) with mTEMPO, a mitochondrially-targeted 2-(2,2,6,6-tetramethylpiperidin-1-oxyl-4-ylamino)-2-oxoethyl)triphenylphosphonium chloride reactive oxygen species scavenger, or ruxolitinib (JAK inhibitor) prevents postacute HRV elevation without altering acute/subacute phases, suggesting that innate immune activation and oxidative stress drive long-term autonomic nervous system remodeling. Early treatment (−4 to +10 DPI) with mTEMPO, a mitochondrially-targeted 2-(2,2,6,6-tetramethylpiperidin-1-oxyl-4-ylamino)-2-oxoethyl)triphenylphosphonium chloride reactive oxygen species scavenger, or ruxolitinib (JAK inhibitor) prevents postacute HRV elevation without altering acute/subacute phases, suggesting that innate immune activation and oxidative stress drive long-term autonomic nervous system remodeling. ▪These findings provide a mechanistic basis for heterogeneous dysautonomia in the postacute phase of COVID-19, highlighting a critical window for intervention to mitigate persistent cardiac autonomic effects. These findings provide a mechanistic basis for heterogeneous dysautonomia in the postacute phase of COVID-19, highlighting a critical window for intervention to mitigate persistent cardiac autonomic effects. Introduction Although the coronavirus disease 2019 (COVID-19) pandemic ended in 2023, COVID-19 and its long-term effects remain a major health concern. After recovery from the acute phase of COVID-19 infection, more than 18 million patients in the United States continue to experience a wide range of signs and symptoms of disease,1,2 collectively known as long COVID or postacute sequelae of COVID syndromes. Long COVID syndrome is a multiorgan disorder that can include chronic fatigue, brain fog, dizziness, heart palpitations, sleep problems, shortness of breath, cough, headache, irregular heartbeat, digestion problems, orthostatic intolerance, and dysautonomia.1,3,4 Some of these symptoms can last for years or are expected to be lifelong. However, little is known about the mechanisms underlying long COVID syndrome, and effective treatments are inadequate. Among the various manifestations of both acute COVID-19 and long COVID, autonomic dysfunction has emerged as a prevalent and concerning issue. The autonomic nervous system (ANS) is responsible for regulating involuntary bodily functions such as heart rate, blood pressure, digestion, and temperature regulation. A balanced activity of the ANS is essential for maintaining overall bodily function and homeostasis and for adapting to environmental changes. Autonomic dysfunction is usually recognized as an altered ANS activity that adversely affects health. The most common manifestation is orthostatic intolerance, which includes orthostatic hypotension, vasovagal syncope, or postural orthostatic tachycardia syndrome (POTS). Other symptoms include chronic fatigue, lightheadedness, nausea, brain fog, palpitations, shortness of breath, bladder and bowel dysfunction, and cardiac arrhythmias. All of these symptoms have been documented in COVID-19 and long COVID. Studies have shown that 30%–35% of patients with long COVID developed POTS5, 6, 7 and various other forms of autonomic complications that persist and significantly affect quality of life and work ability.8, 9, 10, 11, 12 Heart rate variability (HRV) is a simple, validated measure of ANS activity. It reflects fluctuations in RR intervals between heartbeats, influenced by the opposing actions of the sympathetic (SNS) and parasympathetic (PNS) systems. HRV is assessed using time-domain (standard deviation of RR intervals [SDRR], root mean square of successive RR interval difference [rMSSD], percentage of successive RR intervals that differ by more than 100 ms [pRR100]), frequency-domain (low frequency [LF], high frequency [HF], and LF/HF ratio), and nonlinear measures (SD1 and SD2 from Poincaré plots). HRV analysis has been applied to assess autonomic dysfunction in patients with acute COVID13,14 and long COVID,15, 16, 17, 18, 19 yet the direction and magnitude of the changes in HRV parameters had a wide variation. Using a hamster model, we previously showed that, even though no viral protein or mRNA was detectable in the heart, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection activated a cardiac innate immune response and induced cardiac arrhythmias, tachypnea, cardiac conduction disorders, and autonomic dysfunction.20 We assessed the roles of mitochondrial oxidative stress and interferon signaling in the phenotype and demonstrated that early treatment with the mitochondrially-targeted reactive oxygen species scavenger 2-(2,2,6,6-Tetramethylpiperidin-1-oxyl-4-ylamino)-2-oxoethyl)triphenylphosphonium chloride 2,2,6,6-tetramethylpiperidine-1-oxyl (mTEMPO) or ruxolitinib (Ruxo), a Janus kinase (JAK1/JAK2) inhibitor, attenuated the electrophysiological alterations.20 However, it remains unclear whether the increase in HRV observed at 4 weeks after infection represents a transient condition or the onset of sustained ANS dysfunction. In the present study, we extended the time period after infection to 8 weeks to span the early postacute phase of COVID-19, which is defined as symptoms persisting for more than 4 weeks after infection,21 and characterized the ANS activity in different phases of SARS-CoV-2 infection with HRV analysis. We further examined the effects of mTEMPO and Ruxo on ANS dysfunction in the hamster COVID-19 model. The results demonstrate that autonomic dysfunction induced by SARS-CoV-2 infection is phase dependent and persistent. Inhibition of oxidative stress or interferon signaling during the acute phase demonstrated protective effects on ANS dysfunction in the postacute phase. The results provide experimental evidence for a temporal mechanism underlying the inconsistent observations of autonomic dysfunction in the acute, subacute, and postacute phases of COVID-19 and offer insights into ANS defects that may set the stage for subsequent long COVID syndrome. Methods The animal procedures in this study were approved by the Institutional Animal Care and Use Committee of Johns Hopkins University, in accordance with the National Research Council’s Guide for the Care and Use of Laboratory Animals (eighth edition). The Johns Hopkins animal care and use program is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International. SARS-CoV-2 infection SARS-CoV-2 infection experiments with Golden Syrian hamsters were performed in a biosafety level 3 facility at Johns Hopkins Research Animal Resources, in compliance with the established ethical guidelines. 50% tissue culture infectious dose/mL of SARS-CoV-2/delta variant (SARS-CoV-2/USA/MD-HP05660/2021; GISAID accession number EPI_ISL_2331507) in 100 μL Dulbecco’s modified Eagle medium was intranasally administered (50 μL per nare) to male hamsters (Envigo, Haslett, MI) as described previously.22 Mock control animals received Dulbecco’s modified Eagle medium alone. Radiotelemetry device and osmotic pump implantation The procedures were described previously.20 Briefly, device implantation was performed 4 days before the SARS-CoV-2 infection. Hamsters were anesthetized with 4% isoflurane in a closed chamber for 4 minutes and then maintained with oxygen and 2% isoflurane during surgery. To monitor the electrocardiogram (ECG), the biopotential leads (model: HD-X11 or ETA-F10, DSI, St. Paul, MN) were secured with suture on the muscle layer at the lead II position. The body of the device was placed on the side of the abdominal cavity. To measure subpleural pressure, the esophagus was isolated approximately 1.5 cm from its junction with the diaphragm, and a 25-gauge needle was inserted between the serosal and muscularis layers. The needle was tunneled past the junction with the diaphragm into the thoracic cavity. The needle was then removed, and the transmitter catheter of the dual pressure/biopotential device (model HD-X11, DSI) was threaded through the tunnel and secured in place with suture on the serosal layer. For treatment with mTEMPO or Ruxo, Alzet osmotic pumps (model 2ML2) filled with drug or vehicle were implanted into the abdominal cavity to deliver the drug for 14 days. Both mTEMPO and Ruxo were delivered at 1 mg/kg/d. After 3 days of recovery, baseline ECG and subpleural pressure were recorded for 24 hours and hamsters were then mock or SARS-CoV-2 infected. Radiotelemetry data analysis ECG and subpleural pressure data were collected with Ponemah 3.0 (DSI). Subpleural pressure and HRV were analyzed with LabChart 8 (ADInstruments, Colorado Springs, CO) with modules of pressure and HRV. The resting ECG recorded overnight was analyzed by a 6-hour segment. Time-domain indices are SDRR, rMSSD, and pRR100. Frequency-domain indices include LF and HF bands in power spectral analysis and the ratio of LF to HF (LF/HF). In the nonlinear domain, SD1 and SD2 are derived from the Poincaré plot, which is a plot of every RR interval against the previous RR interval. SD1 and SD2 are the short and long axes of an ellipse fitting the Poincaré plot. Generally, SDRR, rMSSD, pRR100, and HF are closely correlated with PNS activity, whereas LF reflects both SNS and PNS activities.23, 24, 25 LF/HF reflects sympathovagal balance, and SD1 and SD2 correlate with HF and LF, respectively.25 Statistical analysis Values are expressed as mean ± standard error of the mean. All data were analyzed using JMP Pro 18.0.1. The longitudinal analysis of body weight and activity in response to SARS-CoV-2 infection (Figure 1A and 1B) was performed with a mixed model within each group followed by Dunnett’s test to compare each time point with baseline (0 DPI). The time-course data of RR (Figure 1C) and HRV indices (Figure 2) were analyzed in the same model with a nested effect. Data from each phase (Figure 3, Figure 4, Figure 5, Figure 6) were analyzed with nested 2-way analysis of variance followed by the Tukey honestly significant difference test for pairwise comparison. Data in Figures 7 and 8 were analyzed with 2-way analysis of variance followed by the Tukey honestly significant difference test for pairwise comparison. Sample size is 5 in the mock group and 6 in the SARS-CoV-2 group in Figure 1, Figure 2, Figure 3, Figure 4, Figure 5, Figure 6 and Figure 7E and 7F, 3 in the mock and SARS-CoV-2 groups in Figure 7C and 7D, and 4 in all groups in Figure 8. SARS-CoV-2 infection induces weight loss, reduced activity, and bradycardia in the hamster model. A: Body weight of the mock and SARS-CoV-2 groups was recorded from 0 to 56 DPI and indicates a rapid decrease in body weight after SARS-CoV-2 infection, which then gradually recovered to baseline level after 42 DPI. B: The activity of the hamsters, recorded by radiotelemetry, was significantly reduced at 1, 4, and 7 DPI and recovered after 14 DPI in the SARS-CoV-2 group, whereas there was no significant change in the mock group. C: RR interval showed a triphasic response to SARS-CoV-2 infection. ∗P < .05; †P < .01; ‡P < .001; ‡‡P < .0001 compared with baseline. DPI = days postinfection; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2. Time-series plot of HRV indices. HRV indices from 0 to 56 DPI: SDRR (A), rMSSD (B), pRR100 (C), power of LF (D), power of HF (E), ratio of LF to HF (F), SD1 (G), and SD2 (H). ∗P < .05; †P < .01; ‡P < .001; ‡‡P < .0001 compared with baseline. Mock-infected group (black, n = 5); SARS-CoV-2–infected group (red, n = 6). DPI = days postinfection; HF = high frequency; HRV = heart rate variability; LF = low frequency; pRR100 = percentage of successive RR intervals that differ by more than 100 ms; rMSSD = root mean square of successive RR interval difference; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SDRR = standard deviation of RR intervals. HRV analysis at baseline. HRV analysis at 0 DPI shows no difference in time-domain indices (A–C), frequency-domain indices (D–F), and nonlinear indices (I and J) between the mock and SARS-CoV-2 groups. n = 5 in the mock group and n = 6 in the SARS-CoV-2 group. G and H: Representative Poincaré plot of the mock (G) and SARS-CoV-2 groups (H). DPI = days postinfection; HF = high frequency; HRV = heart rate variability; LF = low frequency; pRR100 = percentage of successive RR intervals that differ by more than 100 ms; rMSSD = root mean square of successive RR interval difference; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SDRR = standard deviation of RR intervals. HRV increased during the acute phase. Group means of HRV indices at 1 and 4 DPI: (A–C) time-domain indices, (D–F) frequency-domain indices, and (I and J) nonlinear indices. n = 5 in the mock group and n = 6 in the SARS-CoV-2 group. G and H: Representative Poincaré plot of the mock (G) and SARS-CoV-2 groups (H). ∗P < .05; †P < .01; ‡P < .001; ‡‡P < .0001. DPI = days postinfection; HF = high frequency; HRV = heart rate variability; LF = low frequency; pRR100 = percentage of successive RR intervals that differ by more than 100 ms; rMSSD = root mean square of successive RR interval difference; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SDRR = standard deviation of RR intervals. HRV analysis during the subacute phase. Group means of HRV indices at 7 and 14 DPI: (A–C) time-domain indices, (D–F) frequency-domain indices, and (I and J) nonlinear indices. n = 5 in the mock group and n = 6 in the SARS-CoV-2 group. G and H: Representative Poincaré plot of the mock (G) and SARS-CoV-2 groups (H). ∗P < .05; †P < .01; ‡‡P < .0001. DPI = days postinfection; HF = high frequency; HRV = heart rate variability; LF = low frequency; pRR100 = percentage of successive RR intervals that differ by more than 100 ms; rMSSD = root mean square of successive RR interval difference; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SDRR = standard deviation of RR intervals. HRV increased again during the postacute phase. Group means of HRV indices at 28 and 56 DPI: (A–C) time-domain indices, (D–F) frequency-domain indices, and (I and J) nonlinear indices. n = 5 in the mock group and n = 6 in the SARS-CoV-2 group. G and H: Representative Poincaré plot of the mock (G) and SARS-CoV-2 groups (H). ∗P < .05; ‡P < .001. DPI = days postinfection; HF = high frequency; HRV = heart rate variability; LF = low frequency; pRR100 = percentage of successive RR intervals that differ by more than 100 ms; rMSSD = root mean square of successive RR interval difference; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SDRR = standard deviation of RR intervals. Autonomic nervous system activity evaluation with respiratory sinus arrhythmia and MChR and β-AdR blockade. Respiration was monitored with subpleural pressure, and respiratory sinus arrhythmia was evaluated as ΔHR, which is HRmax-HRmin within each respiratory cycle. A and B: Representative traces of subpleural pressure (upper) and HR (lower) at 0, 1, 14, and 28 DPI in the mock (A) and SARS-CoV-2 groups (B). C and D: Group mean of ΔHR at different phases in the mock (n = 3) (C) and SARS-CoV-2 groups (n = 3) (D). Mean RR interval measured before treatment (Pre-tx), followed by MChR blockade with methylatropine and by β-AdR blockade with propranolol in the mock (n = 5) (E) and SARS-CoV-2 groups (n = 6) (F) during acute and postacute phases. Horizontal bar, 2 seconds; vertical black bar, 10 mm Hg; vertical gray bar, 100 beats/min. ∗∗P < .0001 compared with 0 DPI; †P < .01; ‡P < .001; ‡‡P < .0001. β-AdR = β-adrenergic receptor; DPI = days postinfection; HF = high frequency; HR = heart rate; LF = low frequency; MChR = muscarinic cholinergic receptor; pRR100 = percentage of successive RR intervals that differ by more than 100 ms; rMSSD = root mean square of successive RR interval difference; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SDRR = standard deviation of RR intervals. Treatment with mTEMPO or Ruxo prevents the increase in HRV during the postacute phase. HRV indices (A–H) at 0, 1, 7, and 28 DPI in the SARS-CoV-2, SARS + mTEMPO, and SARS + Ruxo groups. n = 4 in each group. ∗P < .05; †P < .01; ‡P < .001; ‡‡P < .0001 compared with 0 DPI. DPI = days postinfection; HF = high frequency; HRV = heart rate variability; LF = low frequency; mTEMPO = mitochondrially targeted 2,2,6,6-tetramethylpiperidine-1-oxyl; pRR100 = percentage of successive RR intervals that differ by more than 100 ms; rMSSD = root mean square of successive RR interval difference; Ruxo = ruxolitinib; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SDRR = standard deviation of RR intervals. Results SARS-CoV-2 induced weight loss, reduced activity, and bradycardia 4-month-old male hamsters were inoculated intranasally with SARS-CoV-2 (delta variant) or vehicle (mock-infected controls). Body weight was monitored for 8 weeks and compared with the baseline at 0 days postinfection (DPI). Consistent with our previous observations in this model,20 SARS-CoV-2 infection resulted in weight loss that peaked at 6 DPI, with a 14% ± 1% decrease compared with baseline (Figure 1A). Body weight then recovered gradually toward baseline levels (Figure 1A) but remained slightly lower than age-matched mock-infected controls at 56 DPI (Figure 1A). The ECG and activity of the hamsters were monitored with biopotential radiotelemetry devices in freely moving animals. SARS-CoV-2 infection significantly decreased activity during the first week after infection (average counts/min was 0.19 ± 0.02 at 0 DPI vs 0.07 ± 0.002 at 1 DPI, P < .001; 0.08 ± 0.01 at 4 DPI, P < .001; 0.06 ± 0.01 at 7 DPI, P < .0001), with recovery at 14 DPI, whereas the activity of the mock group was not significantly changed (Figure 1B). Our previous study showed that SARS-CoV-2 infection induced multiple arrhythmias, including bradycardia, sinus pauses, and atrioventricular block in this model.20 The mean RR interval measured during 0–56 DPI showed a triphasic pattern of RR response to SARS-CoV-2 infection. The mean RR interval of the SARS-CoV-2 group immediately increased after infection at 1 and 4 DPI (262.5 ± 7.3 ms at 1 DPI and 249.6 ± 4.5 ms at 4 DPI compared with 201.3 ± 2.6 ms at 0 DPI), which then declined significantly, returning to baseline level by 14 DPI (Figure 1C). After the recovery, the mean RR interval again increased monotonically to levels significantly higher than that of 0 DPI at 28 and 56 DPI (215.0 ± 3.1 ms and 227.5 ± 3.1 ms at 28 and 56 DPI, respectively) (Figure 1C). In contrast, RR did not significantly change compared with baseline in the mock group (Figure 1C). It should be noted that, in this model, the lungs are negative for SARS-CoV-2 viral RNA by 7 DPI, as measured by in situ hybridization, although declining levels of viral RNA can still be detected by reverse-transcriptase polymerase chain reaction until 28 DPI.22 Cardiac autonomic function changes dynamically in a triphasic manner after SARS-CoV-2 infection To investigate whether autonomic dysfunction contributes to the triphasic response, we evaluated the autonomic nervous activity using HRV analysis. The results indicated that nearly all parameters from the time domain, frequency domain, and nonlinear domain exhibited a triphasic response pattern after SARS-CoV-2 infection (Figure 2). HRV transiently increased immediately after infection but returned to baseline or lower levels by 7 DPI. After remaining depressed during the second week after infection, HRV gradually increased again, reaching values significantly higher than baseline at 28 and 56 DPI (Figure 2). These dynamic changes in HRV suggest that SARS-CoV-2 infection induces cardiac autonomic dysfunctions that vary depending on the postinfection phases, which we refer to as baseline, acute, subacute, and postacute phases, respectively. We then characterized the alterations of cardiac autonomic function in detail for each phase. Except for the baseline at 0 DPI, we analyzed 2 time points in each phase, based on ANS activity: 1 and 4 DPI for the acute phase, 7 and 14 DPI for the subacute phase, and 28 and 56 DPI for the postacute phase. It should be noted that HRV is influenced by heart rate, particularly at lower heart rates, and correction for heart rate is often recommended26; however, this effect is expected to be minimal within the hamster heart rate range of 250–350 beats/min (bpm), and therefore, HRV was not corrected for heart rate in this study. HRV analysis of the baseline before infection (0 DPI) showed no significant difference in time-domain, frequency-domain, and nonlinear indices between the SARS-CoV-2– and mock-infected groups (Figure 3). After infection, HRV indices increased immediately in the SARS-CoV-2 group but not in the mock group. Time-domain analysis revealed that SDRR, rMSSD, and pRR100 at 1 DPI in the SARS-CoV-2 group were significantly higher than baseline by 51%, 69%, and 264%, respectively (Figure 2A–2C), and the corresponding mock group by 72%, 89%, and 348%, respectively (Figure 4A–4C). At 1 DPI, in the frequency domain, the power of both the LF and HF components increased significantly in the SARS-CoV-2–infected group compared with baseline (fold increase 1.6 for LF and 3.1 for HF) (Figure 2D–2F) or compared with the mock group (fold increase 3.7 for LF and 3.6 for HF) (Figure 4D and 4E), whereas the ratio of LF to HF was not significantly different (Figure 4F). Nonlinear analysis also showed that SD1 and SD2 in the SARS-CoV-2 group increased by 62% and 37%, respectively, compared with baseline (Figure 2G and 2H) and by 89% and 53%, respectively, compared with the mock group (Figure 4I and 4J). The increase in HRV induced by SARS-CoV-2 infection was sustained at 4 DPI but was affected to a lesser extent than at 1 DPI. SDRR, rMSSD, pRR100, and SD1 at 4 DPI in the SARS-CoV-2 group were significantly higher than those in the mock group (Figure 4A–4C) or their baseline (Figure 2A–2C) but were significantly lower than those at 1 DPI in the SARS-CoV-2 group, except pRR100, which was similar at 1 and 4 DPI (Figure 4A–4C). In the frequency domain, the power of LF decreased significantly at 4 DPI and was lower than baseline (Figure 2D–2F) and 1 DPI, whereas the power of HF decreased slightly compared with that at 1 DPI (Figure 4D and 4E). As a result, LF/HF significantly decreased at 4 DPI (0.08 ± 0.01 at 4 DPI vs 0.53 ± 0.05 at baseline, P < .0001, and 0.32 ± 0.04 at 1 DPI, P < .0001) (Figure 4F). SD1 and SD2 at 4 DPI remained higher than their baselines (Figure 2G and 2H), whereas SD2 was not different from mock controls (Figure 4J). The increase in HRV during the acute phase diminished with a decreasing trend in the subacute phase. Compared with mock controls, all indices returned to normal levels at 7 and 14 DPI except LF and LF/HF, which remained significantly lower than mock controls at 7 DPI, and SD2, which further declined and was lower than mock control at 14 DPI (Figure 5D, 5F, and 5J). The decreasing trend of HRV during the subacute phase was more pronounced than baseline within the SARS-CoV-2 group. At 14 DPI, SDRR, rMSSD, SD1, and SD2 were significantly lower by 21%, 22%, 25%, and 17%, respectively, than those at baseline (Figure 2). LF power remained lower at 7 DPI and was not significantly different at 14 DPI compared with baseline (901 ± 112 ms2 at 0 DPI vs 175 ± 19 ms2, P < .01, at 7 DPI; 369 ± 61 ms2, P = .052, at 14 DPI) (Figure 2D). The power of HF at 7 and 14 DPI was slightly lower than baseline without statistical significance, whereas LF/HF was significantly lower at 7 and 14 DPI than baseline (0.53 ± 0.05 at 0 DPI vs 0.12 ± 0.01, P < .0001, and 0.24 ± 0.05, P < .0001, at 7 and 14 DPI, respectively) (Figure 2E and 2F). HRV increased again after the subacute phase. At 56 DPI, SDRR, rMSSD, and HF were significantly elevated compared with the mock group (Figure 6). Compared with the baseline of the SARS-CoV-2 group, time-domain and nonlinear parameters were significantly higher at 28 and 56 DPI (Figure 2). In the frequency domain, LF was higher at 56 DPI and HF was higher at both 28 and 56 DPI, whereas LF/HF was lower at both 28 and 56 DPI than baseline (Figure 2). HRV analysis suggested hyperactive PNS, impaired PNS, and increased PNS in the acute, subacute, and postacute phases, respectively, in this hamster model. To confirm the PNS activity estimated with HRV analysis, we further evaluated PNS activity with the analysis of respiratory sinus arrhythmia (RSA). RSA primarily arises from changes in PNS output during inspiration and exhalation. During inspiration, impulses from stretch receptors in the lung cause feedback inhibition of central vagal output, resulting in decreased cardiac vagal tone and heart rate acceleration. Heart rate then slows during exhalation as vagal tone increases. Respiration was monitored with subpleural pressure measurements, recorded together with ECG using a dual pressure/biopotential device. As described previously, SARS-CoV-2 infection induced tachypnea during the acute and early subacute phases (4 and 7 DPI), which we found impeded the reliability of the RSA analysis. Therefore, we performed RSA analysis at 0, 1, 14, and 28 DPI when respiration rates were at similar levels. The amplitude of RSA (HRmax-HRmin), reflecting the range of vagus nerve modulation, was significantly higher than baseline at 1 and 28 DPI and below baseline at 14 DPI (63.3 ± 4.3 bpm at 0 DPI vs 97.33 ± 4.6 bpm at 1 DPI, P < .0001; 3.6 ± 2.3 bpm at 14 DPI, P < .0001; 91.3 ± 5.8 bpm at 28 DPI, P < .0001), whereas no differences were observed in the mock group (Figure 7A–4D). The RSA analysis confirmed that PNS activity was high during acute and postacute phases and was below normal during the subacute phase. We further estimated the PNS and SNS activity by blockade of muscarinic cholinergic receptors and β-adrenergic receptors with methylatropine (mAtrp) and propranolol (Prop), respectively, during the acute and postacute phases. mAtrp (2 mg/kg) and Prop (1 mg/kg) were administered sequentially via intraperitoneal injection to assess PNS and SNS contributions to heart rate: as expected, RR interval was decreased by mAtrp and increased by Prop (Figure 7E and 7F). The change of RR induced by mAtrp was significantly higher in the SARS-CoV-2 group than in the mock group during both the acute and postacute phases (ΔRR by mAtrp during the acute phase, 55.9 ± 11.3 ms in mock vs 146.4 ± 12.4 ms in the SARS-CoV-2 group, P < .0005; postacute phase, 44.1 ± 7.9 ms in mock vs 75.8 ± 5.6 in SARS-CoV-2, P < .01) confirming an increase in PNS activity the during acute and postacute phases after SARS-CoV-2 infection. However, the increase in RR induced by Prop did not differ between the mock and SARS-CoV-2 groups during the acute or postacute phase. The similar RR intervals after inhibition of both ANS branches with mAtrp and Prop in the mock- and SARS-CoV-2–infected groups suggest that HRV changes are minimally attributable to alterations in intrinsic sinoatrial node function. Changes of ANS activity by SARS-CoV-2 infection were suppressed by inhibition of innate immune signaling or mitochondrial oxidative stress Our previous study showed that the effects of SARS-CoV-2 infection on pulmonary function and cardiac electrophysiology were attenuated by a mitochondrially targeted antioxidant or a JAK/STAT signaling inhibitor.20 Here we analyzed HRV of SARS-CoV-2–infected hamsters with or without treatment of Ruxo or mTEMPO at 0, 1, 7, and 28 DPI. Although the increase in HRV induced by SARS-CoV-2 infection during the acute phase at 1 DPI was not affected by mTEMPO or Ruxo, the treatment with mTEMPO or Ruxo prevented the chronic increase in HRV in the postacute phase at 28 DPI (Figure 8). Compared with baseline at 0 DPI, time-domain (Figure 8A–8C) and nonlinear indices (Figure 8G and 8H) were significantly higher at 28 DPI in the SARS-CoV-2 group, but were restored to baseline levels with either the mTEMPO or Ruxo treatments. Discussion In the present study, longitudinal HRV analysis demonstrated that SARS-CoV-2 infection induces a triphasic response of ANS activity extending to up to 8 weeks after infection. The findings provide an unprecedented view of the dynamic changes in SNS and PNS activity during an active response to SARS-CoV-2 infection and reveal persistent ANS dysfunction that develops in the postacute phase of recovery. Acute effects of SARS-CoV-2 infection on ANS The results indicated that both SNS and PNS were activated early in the acute phase by SARS-CoV-2 infection, but with PNS dominance, reflected by increased RR and reduced LF/HF. As part of the “fight or flight” response under acute stress, SNS activation can be induced by SARS-CoV-2 infection with multiple mechanisms. Previous studies on this model showed that SARS-CoV-2 infection induced innate immune activation with the release of cytokines, infiltration of immune cells, and respiratory dysfunction.20,22 The proinflammatory cytokines, such as interleukin-1β, interleukin-6, and tumor necrosis factor α, released during innate immune activation can stimulate the hypothalamic-pituitary-adrenal axis and promote the release of catecholamines.27 Hypoxia, owing to respiratory distress, can further aggravate SNS overactivity by stimulating the carotid body chemosensory input to the brainstem. The SNS surge mobilizes energy resources, increases cardiovascular output, and modulates immune cell activity. The PNS plays a crucial role in modulating the body’s inflammatory response during acute infections through the cholinergic anti-inflammatory pathway (CAP). Primarily mediated by the vagus nerve, CAP functions by stimulating acetylcholine release from the vagus nerve, which interacts with nicotinic receptors on immune cells, particularly macrophages. This interaction suppresses cytokines and other inflammatory mediators to limit the extent and duration of the inflammatory response. In addition, this pathway enhances the body’s ability to clear pathogens while minimizing collateral damage to host tissues. ANS dysfunction has been reported in patients with COVID-19 from asymptomatic to severe cases with divergent outcomes.28, 29, 30 Hyperactive SNS with PNS withdrawal is commonly observed in patients with COVID-19 with preexisting conditions, such as hypertension, heart failure, diabetes, or pulmonary disease, and is believed to be a risk factor for the higher morbidity and mortality after SARS-CoV-2 infection in this population.31, 32, 33 The mechanisms underlying PNS withdrawal associated with hyperactive SNS are unclear, but could be attributed to vagus nerve infection by SARS-CoV-234 or reciprocal inhibition of PNS by hyperactive SNS.35 PNS withdrawal may decrease the CAP to exacerbate the inflammatory response. Indeed, HRV analysis of patients with COVID-19 showed that lower HRV indices (RR, SDNN, and rMSSD) were associated with myocardial injury.36 PNS dominance can also contribute to autonomic dysfunction in patients with COVID-19, reflected by increased RR and/or HRV.13,37, 38, 39, 40 Approximately 22.6% of COVID-19–associated arrhythmias are bradyarrhythmias.41 HRV analysis of patients with COVID-19 indicated higher PNS tone than uninfected controls, but no difference between subjects with symptomatic and asymptomatic COVID-19.13 Dumargne et al39 also showed increased HRV, in both time and frequency domains, in patients with COVID-19 with acute respiratory distress syndrome compared with patients with other respiratory etiologies. Patients with COVID-19 with bradycardia were also shown to have increased rates of mortality.42,43 Autonomic exhaustion after hyperactivation The defined subacute phase in our model showed decreases in HRV indices to levels at or below baseline and recovery of body weight and activity. Although LF is affected by both SNS and PNS activity,44 the rapid drop of LF from the 1 DPI peak to a level lower than baseline at 4 DPI, when HF remained higher than baseline, suggests that the early LF decline is the result of reduced SNS activity. The decreases in HF and time-domain indices at 7 and 14 DPI indicated reduced PNS activity, which was confirmed with RSA analysis. Therefore, it is plausible to believe that the activities of both SNS and PNS were low in the subacute phase. The restoration of HRV after the subacute phase suggested that the loss of HRV is a transient effect after hyperactivation of ANS during the acute phase in this model. A global reduction in HRV indices, including SDNN, pRR100, rMSSD, LF, and HF, was also reported in subjects recovered from COVID-19.45 However, whether HRV returned to normal later was not determined. ANS reactivation during postacute phase HRV indices gradually increased and heightened RR, rMSSD, HF, and RSA at 28 and/or 56 DPI confirmed autonomic dysfunction with PNS dominance during the postacute phase. In patients with long COVID, both PNS dominance and SNS dominance with PNS withdrawal have been reported,15,46 but the mechanism for these diverse responses remains unknown. The dynamic changes in the ANS observed in our study suggest that the timing of the post-COVID study is crucial and could contribute to the inconsistency. Very few reports are available about the longitudinal effect of SARS-CoV-2 infection on autonomic function in patients with long COVID. A follow-up HRV analysis in health care workers who contracted SARS-CoV-2 showed that LF/HF increased at 1 month and decreased at 6 and 13 months after a negative swab test.47 Another study with HRV analysis in young adults at 3–22 weeks after COVID diagnosis showed that rMSSD and HF were positively correlated and LF/HF was negatively correlated with time since diagnosis.48 Santos-de-Araujo et al49 investigated COVID-19’s impact on HRV by time after infection and found that rMSSD was low in groups less than 6 weeks and 2–6 months after infection, but higher in the 7–12-month group. These studies support dynamic changes of ANS activity in long COVID. Effects of mTEMPO and Ruxo on ANS dysfunction Based on our previous work implicating mitochondrial oxidative stress and interferon signaling in the pathophysiology of COVID-19,20 we examined the effects of mTEMPO and Ruxo treatments on ANS dysfunction induced by SARS-CoV-2 infection during the acute, subacute, and postacute phases. Although mTEMPO or Ruxo administration in the early postinfection period blunted the acute arrhythmias and tachypnea in our previous study,20 surprisingly, these treatments did not significantly attenuate the elevated HRV parameters in the acute phase, nor did the treatments affect the decline in HRV parameters in the subacute phase. Nevertheless, the redevelopment of ANS dysfunction in the postacute phase was prevented by mTEMPO or Ruxo. This indicates that the delayed increase in HRV long after the virus has been cleared likely results from a latent injury or remodeling of the cardiac ANS occurring in the early stages of SARS-CoV-2 infection, when the innate immune response is at its peak, emphasizing a critical time window for intervention. At the postsynaptic level, we previously showed that infiltrating monocyte-derived macrophages increased and resident macrophages decreased in the cardiac conduction system during the acute phase of SARS-CoV-2 infection.20 The early loss of resident macrophages was prevented by mTEMPO or Ruxo, without significant effects on the infiltration of monocyte-derived macrophages.20 Thus, the loss of important supporting functions of resident macrophages50 in the conduction system could contribute to altered neuron-pacemaker interactions. Activation of the innate immune system has been shown to induce chronic inflammation or neuropathy via neuroimmune cross talk,51 and there is increasing evidence that SARS-CoV-2 infection is associated with neuroinflammation and disorders of the cranial and peripheral nervous system,34,52 which could contribute to ANS dysfunction in patients with long COVID. Clinical implications Our results showing dynamic ANS remodeling after SAR-CoV-2 infection suggest that the timing of clinical evaluation may influence the assessment of ANS function in patients with COVID-19. Therefore, longitudinal monitoring may be necessary to distinguish transient from persistent ANS remodeling. Increased RR interval and sustained PNS dominance in this model align with clinical reports linking relative bradycardia and increased HRV to adverse outcomes in subsets of patients with COVID-19,13,39 underscoring the need to consider both SNS excess and excessive vagal drive as potentially maladaptive responses. Early hyperactivation of the immune system and mitochondrial oxidative stress in the acute phase seem to be major contributors to delayed autonomic dysfunction in the postacute phase. Both the cytokine storm and mitochondrial dysfunction have been identified as key aspects of COVID-19 underlying multiorgan damage and long COVID pathogenesis.53, 54, 55, 56, 57, 58 These results support further exploration of early-phase immunomodulatory or mitochondria-directed therapies as strategies to mitigate long COVID dysautonomia. Limitations The use of the SARS-CoV-2 hamster model in the BSL-3 laboratory imposed certain limitations on sample size, including a limited capacity for telemetry recording, housing availability, and limitations on ex vivo studies. In addition, because male hamsters displayed more severe symptoms of SARS-CoV-2 infection,22 we excluded females from our study; thus, we did not assess sex-, age-, or preexisting pathology-dependent differences that are prominent in patients with long COVID.59,60 Although abnormal HRV was observed in humans, the translation of hamster findings to humans requires caution in light of the high heart and respiratory rates in small rodents, as well as species differences in ANS balance, stress and environmental sensitivity, and responses to infection. Although HRV analysis is a useful noninvasive measure of autonomic regulation, its interpretation in terms of distinct SNS and PNS contributions is the subject of debate. Both branches can influence multiple HRV indices in nonlinear and context-dependent ways. In addition, HRV is modulated by factors beyond autonomic activity, including sinoatrial node properties, respiration, baroreflex function, and inflammation. Therefore, indices such as LF power and LF/HF ratio should be interpreted with caution. The present study focuses primarily on cardiac autonomic indices and RSA and does not directly evaluate autonomic regulation of other organ systems commonly affected in long COVID, such as cerebrovascular, gastrointestinal, or sudomotor function. As a result, the findings may not encompass the full spectrum of systemic dysautonomia experienced by patients, particularly manifestations such as orthostatic intolerance and POTS that depend on integrated cardiovascular, vascular, and baroreflex responses often assessed by tilt-table or standing tests in humans. Conclusion SARS-CoV-2 infection in the hamster model induces cardiac autonomic dysfunction, which displays a triphasic pattern: acute hyperactivation with PNS predominance, subacute autonomic exhaustion, and postacute reactivation with sustained PNS dominance. Inhibition of innate immune activation or mitochondrial oxidative stress during the acute phase prevents the autonomic dysfunction in the postacute phase, offering a potential strategy to mitigate long COVID syndromes. Declaration of generative AI and AI-assisted technologies in the writing process Artificial intelligence was not used in the preparation of this manuscript. Disclosures The authors have no conflicts of interest to disclose. Acknowledgments Funding Sources This study was supported by American Heart Association grants 965158 and 25TPA1478354 (BO’R). Authorship All authors attest they meet the current ICMJE criteria for authorship. Ethics Statement The animal procedures in this study were approved by the Institutional Animal Care and Use Committee of Johns Hopkins University, in accordance with the National Research Council’s Guide for the Care and Use of Laboratory Animals (eighth edition). References