Although conventional methods such as peripheral blood or nasopharyngeal swab sampling may fail to detect any ongoing presence of SARS-CoV-2, the virus might nonetheless establish a persistent infection or leave noninfectious remnants in deep tissues. In fact, viral replication might occur for several months after the initial infection as evinced by the detection of subgenomic RNA, a marker of recent virus replication; the isolation of replication-competent SARS-CoV-2 from respiratory and nonrespiratory tissues^{164, 165 and 166}; and the existence of viral reservoirs for SARS-CoV-2.^{68,150,167, 168, 169, 170, 171, 172, 173, 174 and 175} More specifically, these studies have demonstrated the persistence of viral antigens, viral RNA, and whole virus in the brain, sinus, adrenal glands, kidneys, gut, lymph nodes, spleen, lungs, and heart, which can underlie symptoms through direct viral cytopathic effects; regional inflammation; triggering an immune response causing an elevated and prolonged state of generalized inflammation; and prompting autoimmunity in pediatric or adult populations.^168,176,177

Evaluation in a Brazilian study for the presence of subgenomic nucleoprotein gene (N) mRNA in rectal swab samples and cytopathic effects in Vero cell culture demonstrated the replication capacity of SARS-CoV-2 in the gastrointestinal tract and infectious viruses in one rectal swab sample,¹⁷⁸ further supporting previous studies that detected SARS-CoV-2 in feces.^{179, 180, 181, 182, 183, 184 and 185} Immunofluorescence and PCR analyses of intestinal biopsies obtained from asymptomatic individuals at 4 months after the onset of COVID-19 revealed the persistence of SARS-CoV-2 nucleic acids and immunoreactivity
in the small bowel of 7 out of 14 individuals.¹⁵⁰ Persistent viral protein and RNA infection of enterocytes¹⁸⁶ in intestinal biopsies for several months after infection renders the small intestine a reservoir of long-term viral replication.¹⁵⁰

An endoscopy study reported expression of SARS-CoV-2 RNA in the gut mucosa in 32 of 46 patients with inflammatory bowel disease, and viral nucleocapsid protein persistence in gut epithelium and CD8⁺ T cells in 24 of 46 patients at approximately 7 months after mild acute COVID-19 in patients with long COVID but not in those without it; expression of SARS-CoV-2 antigens was not detectable in stool, and viral antigen persistence was unrelated to severity of acute COVID-19, immunosuppressive therapy, and gut inflammation.¹⁵¹

Although usually lacking whole-genome sequencing and phylogenetic analyses to distinguish recurrent infection from reinfection,¹⁸⁷ other studies also have documented that some patients infected with SARS-CoV-2 do not rapidly clear the virus.^172,188,189 In one study,¹⁸⁹ the estimated time to loss of SARS-CoV-2 RNA detection ranged from 45.6 days for nasopharyngeal swab samples to 46.3 days for feces samples in mild cases and from 48.9 days for nasopharyngeal swab samples to 49.4 days for feces samples in severe cases, which was longer than those documented for SARS-CoV-1¹⁹⁰ and MERS-CoV.¹⁹¹ Consistently, SARS-CoV-2 can be detected in anal swabs, a signal that remains positive long after nasopharyngeal swabs are negative.¹⁹²

SARS-CoV-2 antigen persistence has been identified in long COVID cases in various tissues.^{68,150,168, 169, 170, 171, 172 and 173,193, 194, 195, 196 and 197} For instance, studies have documented associations between persistent antigenemia and long COVID manifestations, including in the olfactory bulb perhaps explaining persistent anosmia.^196,197

Spike protein persists in the plasma of patients with long COVID-19 for up to 12 months postdiagnosis.¹⁷¹ Spike protein or its S1 fragment is released from cells during infection, reaching different tissues, including the central nervous system, irrespective of the presence of the viral RNA.^198,199 Cells expressing the spike protein release extracellular vesicles containing the full-length protein,²⁰⁰ which would be another means of its circulation in the body. Free S1 has been shown to cross the blood-brain barrier in mouse models, reaching different memory-related regions of the brain, suggesting that the protein itself, independent of the viral particles, would affect brain functions.²⁰¹ Moreover, high levels of circulating spike protein were detected several months after SARS-CoV-2 infection in patients diagnosed with long COVID but not in individuals who did not present long-term sequelae.¹⁷¹

In a comprehensive imaging and omics study of mice and humans, the skull, brain, and meninges, along with the vertebra and pelvic bones, were assessed from 20 patients who had died from non-COVID causes but had previously documented COVID-19. Long after their acute COVID-19 episode, most (12 of 20) of these individuals had marked accumulation of the SARS-CoV-2 spike protein in the skull-meninges and brain tissue, not found in controls. Only the spike protein, not other parts of the virus, was found in brain parenchyma. When the SARS-CoV-2 spike protein was injected in a mouse model, there was brain cell injury, death, and persistent inflammation in the skull, reflecting the importance of this immunologic niche reservoir.²⁰²

Persistent pathogen reservoir or remnants will generate pathogen-associated molecular patterns (PAMPs), such as viral RNA or proteins, and these can engage various host pattern recognition receptors (PRRs) to trigger innate immune activation. Persistent pathogen or remnant antigens can also trigger T and B lymphocytes.¹⁵⁰ If the effector functions of T cells and antibodies are insufficient to clear the pathogen, chronic stimulation of these lymphocytes can cause inflammatory conditions.

Studies predicted that SARS-CoV-2 spike protein binds to Toll-like receptor (TLR)4, a host PRR,²⁰³ with higher affinity than to angiotensin-converting enzyme 2 (ACE2),^{204, 205 and 206} and its aberrant signaling is involved in the hyperinflammatory response of patients with COVID-19.²⁰⁷ In vitro studies demonstrated that SARS-CoV-2 spike protein activates TLR4 in cultured phagocytic cells, stimulating production of proinflammatory mediators.^{208, 209, 210 and 211} TLR4 has been implicated in microglial activation and cognitive dysfunction of Alzheimer disease,²¹² and a study using a mouse model of intracerebroventricular infusion of the trimeric SARS-CoV-2 spike protein demonstrated late cognitive impairment, synapse loss, and microglial engulfment of presynaptic terminals, which were prevented by early TLR4 blockage.²⁰² The latter model recapitulated not only long-term cognitive impairment but also recovery of memory function seen in long COVID, expanding previous studies that were focused on the short-term effects of exposure to the spike S1 fragment after direct infusion into hippocampal tissue in aged mice.^208,213,214 The TLR4 single-nucleotide polymorphism
rs10759931 was also associated with long-term cognitive impairment in patients who recovered from mild COVID-19.²⁰²

Viral persistence has also been documented for RNA viruses such as West Nile virus, with patients with chronic symptoms still testing positive in urine reverse transcription (RT)-PCR tests 1.6 to 6.7 years after recovery from acute illness.²¹⁵ The longest bouts of SARS-CoV-2 infection documented have been in immunocompromised patients, which is also believed to be an important source of variation for the virus, including the emergence of its Alpha and Omicron variants, with rapid accumulation of many mutations within a short period of time and development of resistance to available preventive and therapeutic interventions.^{216, 217, 218, 219, 220, 221 and 222} A patient in the United Kingdom, for instance, tested positive for COVID-19 for 505 days until her death.²²³

Several cases of prolonged SARS-CoV-2 infection have been reported in people living with human immunodeficiency virus (HIV) with severe T-cell depletion and/or acquired immunodeficiency syndrome (AIDS), with subsequent emergence of a multitude of mutations conferring extensive escape from antibody neutralization elicited by both ancestral SARS-CoV-2 infection and vaccines.^{224, 225, 226 and 227} It should be noted that HIV infection per se is not a risk factor for SARS-CoV-2 infection,^{228, 229, 230, 231 and 232} and people living with HIV with well-controlled HIV infection have similar immune responses to both SARS-CoV-2 natural infection and vaccination as those of people without HIV.²³³ However, people living with HIV appear to be at increased risk of developing long COVID.^{126,233, 234, 235 and 236} To this end, a cohort study of 280 adults reported that underlying HIV infection was independently associated with neurocognitive long COVID (odds ratio, 2.5) at a median 4 months post-SARS-CoV-2 infection.¹²⁵ Increased risk of long COVID in people living with HIV may also be possibly driven by sociodemographic factors such as age and clinical factors such as other comorbidities.²³⁰ Although lack of virologic-immunologic control may lead to long COVID in people living with HIV, a study found no relationship between long COVID and SARS-CoV-2–specific humoral and T-cell responses or immune exhaustion.²³⁵

It remains unknown if SARS-CoV-2 can reactivate and retransmit after several years of dormancy, as has been documented for Ebola virus, which reemerged in 2021 in people who previously had Ebola virus disease in 2016.²³⁷

In terms of viruses other than SARS-CoV-2, reactivation of EBV has been indirectly inferred to correlate with long COVID through antibody titer measurements.^68,235,238 A cohort study of 280 adults documented that fatigue and neurocognitive dysfunction at a median of 4 months following initial diagnosis of SARS-CoV-2 infection were independently associated with serologic evidence, suggesting recent EBV reactivation (early antigen-diffuse immunoglobulin G [IgG] positivity) or high nuclear antigen (EBNA) IgG levels but not with ongoing EBV viremia. Serologic evidence suggesting recent EBV reactivation (early antigen-diffuse IgG positivity) was most strongly associated with fatigue (odds ratio, 2.12).¹²⁵ However, study participants who had serologic evidence of prior CMV infection were less likely to develop neurocognitive long COVID (odds ratio, 0.52), which suggests differential effects of chronic viral coinfections, even from the same family of herpesviruses, on the likelihood of developing long COVID and association with distinct syndromic patterns.²³⁵ Studies such as the latter are usually limited by incomplete assessments during the acute phase of COVID-19 to better generate covariate-adjusted binary logistic regression models to identify independent associations between variables and long COVID symptoms.

Beyond reactivation of EBV and CMV, that of human herpesvirus 6 has also been proposed as playing a role in long COVID based on previous evidence of their effects on viral superantigen activation of the immune system, disturbance in the gut microbiome, and multiple tissue damage and autoimmunity.²³⁹ However, an association with pathogenesis of long COVID, chronic fatigue syndrome, and post-acute infection syndrome remains to be robustly substantiated. Maybe reactivation of herpesviruses is one in a multitude of cofactors in disease pathogenesis.

Persistent immune activation is regarded as a major pathogenic mechanism of long COVID.^{105,240, 241 and 242} Inflammation during early SARS-CoV-2 infection has been associated with adverse outcomes, particularly in those who were hospitalized with COVID-19^{241, 242, 243, 244, 245, 246, 247, 248 and 249} and can persist for weeks to months.^250,251 For example, individuals well enough to donate convalescent plasma have elevations
in certain markers of inflammation, such as interferon gamma, monocyte chemoattractant protein 1, and several interleukins, as compared to prepandemic plasma donors.²⁵⁰ Although various immune-activating cytokines can have protective or homeostatic effects,^{252, 253 and 254} their dysregulation may lead to detrimental clinical conditions.

Studies have identified immune dysregulation and inflammation, including pathways associated with tumor necrosis factor-α, and interleukins-6 and -1β, among others, that could contribute to systemic symptoms such as fever and fatigue in individuals with long COVID relative to individuals with full recovery.^{68,152,153,235,255,256} Persistent immune activation may be associated with ongoing symptoms following COVID-19. In the Long-term Impact of Infection with Novel Coronavirus (LIINC) study¹⁵² of 121 participants (>10% with lung problems, diabetes, and obesity) at San Francisco General Hospital, United States, during early recovery (ie, <90 days) post-SARS-CoV-2 infection, those who went on to develop long COVID generally had significantly higher levels of cytokines including tumor necrosis factor-α (1.14-fold) and interferon-γ-induced protein 10 (1.28-fold). Among those with long COVID, there was a trend toward higher interleukin-6 levels during early recovery (1.29-fold), which became more pronounced in late recovery (1.44-fold). These differences were more pronounced among those with a greater number of long COVID symptoms.¹⁵²

Other forms of immune dysregulation may be associated with long COVID.^170,257 Persistent adaptive immune system dysregulation has been documented into convalescence and up to 8 months, notably changes in cell populations toward a more inflammatory, reactive state with increases in proinflammatory cytokines from various cells and decreases in naïve T-cell and B-cell populations, which might contribute to long COVID development.^{105,258, 259, 260, 261, 262, 263, 264 and 265}

Persistent alterations in the myeloid cell compartment, including monocytes and dendritic cells, after SARS-CoV-2 infection have also been reported,^{266, 267 and 268} and unique monocyte signatures were found among subgroups of patients with long COVID.²⁶⁸ A multiomics study of a few hundred patients reported immune cell phenotypes associated with subtypes of prominent sequelae within 2 to 3 months after acute COVID-19, including myeloid-derived suppressor cells in cough subtype, a memory-like natural killer cell subtype in sputum subtype, and activated T_reg in several subtypes.⁶⁸ However, these findings have not been more widely confirmed and the study did not discern which patients later developed long COVID.

A study of 46 individuals with evidence of interstitial lung changes at 3 to 6 months after recovery from severe COVID-19 documented that they had an upregulated neutrophil-associated immune signature including increased chemokines, proteases, and markers of neutrophil extracellular traps that were detectable in the blood.²⁶⁹ Similar pathways were enriched in the upper airway with a concomitant increase in antiviral type I interferon signaling. Interaction analysis of the peripheral phosphoproteome identified enriched kinases critical for neutrophil inflammatory pathways. Evaluation of these individuals at 12 months after recovery indicated that a subset of the individuals had not yet achieved full normalization of radiologic and functional changes.²⁶⁹ Whether these findings apply to individuals who did not suffer severe COVID-19 remains to be determined.

In parasite and viral infections, aberrant B-cell responses can suppress germinal center reactions, thereby blunting long-lived memory and may provoke immunopathology including autoimmunity. Likewise, adaptive immunity to SARS-CoV-2 may shift toward autoreactivity.^255,270 Although autoreactive immunity has been demonstrated during and just after acute SARS-CoV-2 infection,^{68,271, 272, 273, 274, 275 and 276} the data for long COVID are mixed.^{68,94,170,273,274,277, 278 and 279}

Autoantibodies, especially those that neutralize type I interferons, have been reported to be associated with immune dysfunction and COVID-19 mortality^280,281 and have been speculated to be associated with long COVID.¹⁷⁰ Diverse autoantibody classes are produced in acute COVID-19 as well as post-COVID-19 multisystem inflammatory syndrome in children.^{254,280,281, 282, 283, 284, 285 and 286} Moreover, many reports suggest that patients may experience de novo or worsening of preexisting autoimmune conditions, such as autoimmune cytopenias, Guillain-Barré syndrome, or systemic lupus erythematosus.^{287, 288, 289 and 290}

It remains elusive if autoantibodies represent an inflammatory epiphenomenon or pathophysiologically contribute to long COVID.²⁹¹ However, new diagnoses or exacerbations of clinical entities with autoimmune mechanisms, such as diabetes²⁹² and thyroiditis,²⁹³ have been reported following SARS-CoV-2 infection.

Microvascular dysfunction, platelet activation, and clotting might be potential drivers of long COVID via organ dysfunction manifesting as inflammation,^{154,294, 295, 296, 297, 298, 299, 300 and 301} even after mild COVID-19.³⁰² Endothelial cells are infected by SARS-CoV-2 via the ACE2 receptor expressed on their surface,^{303, 304, 305 and 306} which is downregulated by the SARS-CoV-2 spike protein.³⁰⁷ Viral-mediated endothelial injury results in contraction of cell margins, shedding of glycocalyx, extension of phosphatidylserine-rich filopods, and barrier damage.^{302,308, 309 and 310} During acute SARS-CoV-2 infection, damaged endothelial cells promote diffuse microthrombi and destroy the endothelial (including blood-air, blood-brain, glomerular filtration, and intestinal-blood) barriers, leading to multiple organ dysfunction.^{297,311, 312, 313 and 314} Even viral debris, whether nucleic acids or proteins, may continue to damage endothelial cells, induce cytokines, and activate immune responses.³¹⁵ During the convalescence period, a subset of patients is unable to fully recover because of persistent endothelial dysfunction, contributing to long COVID.³⁰⁸ These patients continue to show elevated levels of D-dimer, factor VIII, thrombin, von Willebrand factor, intercellular adhesion molecule 1 (ICAM-1), and interleukin-6 even 1 year after recovery,^297,312 and persistent endothelial glycocalyx damage, higher levels of circulating endothelial cells, and impaired vascular and myocardial function.^{316, 317, 318 and 319} Sustained microcirculation damage in patients with long COVID has also been observed using sublingual and retinal microscopy.^320,321

A study showed that plasma samples from patients with long COVID or with acute COVID-19 contain large amyloid deposits or microclots, which are resistant to fibrinolysis (compared with plasma from controls and patients with type 2 diabetes mellitus), even after trypsinization.¹⁵⁴ Microclots were solubilized after a second trypsinization, and various inflammatory molecules were substantially increased in both the supernatant fluid and trapped in the solubilized pellet deposits of acute COVID-19 and long COVID, versus the equivalent volume of fully digested fluid of the control samples and patients with type 2 diabetes mellitus. Substantially greater amounts of α(2)-antiplasmin, various fibrinogen chains, as well as Serum Amyloid A were trapped in the solubilized fibrinolytic-resistant pellet deposits.¹⁵⁴

In cerebral small vessels of patients with COVID-19 as well as in mouse and hamster models, a study documented an increase in empty vascular basement membrane tubes, so-called string vessels that have been implicated in regulating cerebrovascular coupling, endothelial cell death, and blood-brain barrier disruption,^322,323 reflecting microvascular pathology.¹⁵⁵ The study showed that endothelial cells were infected, and that the main protease of SARS-CoV-2 (NSP5 or M^pro) cleaves NEMO, the essential modulator of nuclear factor-κB. By ablating NEMO, M^pro induces the death of human brain endothelial cells and the occurrence of string vessels in mice. Deletion of receptor-interacting protein kinase (RIPK) 3, a mediator of regulated cell death, blocks the vessel rarefaction and disruption of the blood-brain barrier because of NEMO ablation.¹⁵⁵

In a study in Spain of patients with no history of cardiovascular disease who experienced chest pain and were referred to a long COVID outpatient clinic, adenosine stress perfusion cardiac magnetic resonance imaging (MRI) showed significant circumferential subendocardial perfusion defect in 50% of cases over 8 months after SARS-CoV-2 infection, despite the absence of any ischemic patterns and myocardial T1 and T2 mapping changes, which is highly suggestive of microvascular dysfunction.³²⁴ Similarly, another study on patients without a history of coronary heart disease who suffered from cardiac injury during the acute phase of COVID-19 reported that patients returned to normal cardiac function within 6 months follow-up.³⁰⁸ However, cardiac MRI revealed a significantly higher proportion in positive late gadolinium enhancement in the cardiac injury group compared with controls, suggestive of myocardial fibrosis. Although the latter fibrosis may be secondary to myocardial damage during acute illness, a study on SARS-CoV-1 had shown that fibrosis might be associated with changes in transforming growth factor-β signaling.³²⁵

Microvascular thrombi are also involved in respiratory system–associated morbidity in long COVID.³²⁶

Microbial translocation, including fungal translocation, because of reduced integrity of the blood-gut barrier may underlie long COVID,^{327, 328, 329, 330 and 331} as might dysbiosis,^{328,332, 333, 334 and 335} with perturbations in the microbiome at over 6 months postinfection, which may explain a variety of gastrointestinal-related symptoms.^{335, 336, 337 and 338} Translocation of gut microbes and their products across the gut epithelium can exacerbate initial systemic inflammation.^186,339

In an observational study, persistence of respiratory symptoms correlated with opportunistic gut pathogens; neuropsychiatric symptoms and fatigue with nosocomial gut pathogens; and hair loss with butyrate-producing species.³³⁵ There was also a positive correlation between 6-minute walking distances and some microflora products (such as several short-chain fatty acids). Another study documented an association between intestinal microbial alterations and persistent low-grade inflammation and respiratory dysfunction (lower diffusing capacity) at 3-month follow-up after severe COVID-19.³⁴⁰ In a case report, modulation of intestinal microbiota (a high-fiber formula) alleviated severe long COVID symptoms, including severe appetite loss, nausea, palpitations, and anxiety.³⁴¹