The first reported case of filovirus persistence was in 1967 when viral RNA was detected in the semen of a Marburg virus (MARV) survivor and resulted in the infection of a sexual partner [1]. Similar small, sporadic outbreaks of filoviruses were recorded over the following decades, however, little insight was gained into the seemingly rare phenomenon of filovirus persistence. This was changed by the unprecedented scale of the 2013–2016 West Africa Ebola virus (EBOV) epidemic; > 28,000 cases and > 11,000 deaths were recorded [2]. The large number of EBOV survivors provided the unprecedented opportunity for studying filovirus persistence, sparking momentum in the field.

Here, we define persistence as the presence of viral RNA within an isolated tissue but not peripheral blood of the host. Persistence in immune privileged niches (IPNs), such as the brain, eyes, and testes, can lead to disease recrudescence and transmission months-to-years after clearance of initial infection which pose risks to both the individual and to public health. Additionally, EBOV persistence contributes to stigmatisation facing survivors within their communities which in turn contributes to transmission due to individuals hiding their infection [3]. Immune responses are dampened in IPNs by active immunosuppression and the presence of tight blood-tissue barriers [4], slowing the clearance of EBOV. Persistent EBOV may be shed into bodily fluids from IPNs, potentially resulting in sexual or vertical viral transmission. Additionally, the persistent virus may reactivate into its pathogenic life cycle leading to recrudescence with associated transmission risks like those in acute EBOV disease (EVD). EBOV pathogenesis, lifecycle and host-pathogen interactions are beyond the scope of this review, however, understanding of these core principles is incomplete, hampering the study of EBOV persistence which builds on this foundation.

This review will explore the evidence-base for the different sites of EBOV persistence in the body and the associated epidemiological implications, with a particular focus on the testes. We additionally present a meta-analysis of longitudinal semen sampling studies from EVD survivors to further evaluate the threat posed by viral shedding. Understanding of the molecular mechanisms underpinning the establishment and maintenance of the persistent reservoir is lacking and so we outline the evidence for three possible theories. Lastly, we examine the solutions to tackling EBOV persistence and propose future research priorities in the field to inform researchers and policymakers.

EBOV (aka Zaire) is one of five species in the Orthoebolavirus genus and, like MARV, belongs to the Filoviridae family. It is assumed filoviruses share the same sites and mechanism of persistence. Filoviruses have a non-segmented negative sense single-stranded RNA genome [5]. Their infection in humans is usually highly lethal; EBOV is no exception with a case-fatality rate up to 90% [6]. Zoonotic spillover from the EBOV animal reservoir, likely bats, has led to numerous outbreaks in Africa with the largest two in the past decade; never in recent history have there been more survivors of EBOV [7]. EBOV incubation period can last 2–21 days, often resulting in non-specific febrile illness followed by more severe symptoms that may include haemorrhagic fever [8]. EBOV is usually spread by the mucosal route and initially infects dendritic cells and macrophages which migrate to lymph nodes, facilitating the dissemination of EBOV throughout the body. The pathogenesis underlying acute EVD is a combination of direct cytopathic effects of viral replication plus immune dysregulation that can lead to fatal hypovolaemia [9].

Figure 1 summarises the immune privileged sites that EBOV has been reported to reside and Table 1 provides the supporting literature. The strongest clinical evidence of persistence in the eye and brain come from just two, but well-documented, cases of recrudescence in foreign healthcare workers volunteering during the 2013–2016 EBOV epidemic. Both received aggressive supportive care and experimental EBOV therapies. The first report of late severe EBOV recrudescence involved a Scottish nurse with meningoencephalitis 10 months after initial EBOV diagnosis, despite prior clearance of viraemia. EBOV RNA was detected in cerebrospinal fluid, strongly indicating persistence within the brain [10]. Recently two further case reports have been published regarding fatal meningoencephalitis at day 137 following initial EBOV infection in an older man and day 342 in a young woman. However, the young woman became pregnant during convalescence which questions whether the brain was the true site of the persistent reservoir [11].

The second healthcare worker involved a US doctor who developed vision-threatening unilateral uveitis nine weeks after viraemia clearance. qRT-PCR showed presence of EBOV RNA in an aqueous humour sample with viral culture confirming presence of viable EBOV [18]. Additional supporting evidence of the eye as a possible site of persistence include a case report of MARV ocular-related recrudescence and non-human primate (NHP) EBOV studies [37, 42, 47]. However, numerous prospective studies of EBOV survivors did not detect EBOV RNA in the aqueous humour [48–50]. One such prospective study investigated 22 EVD survivors undergoing cataract surgery and did not detect EBOV RNA in any aqueous humour samples. The study suggests cataract surgery is both effective and safe in EVD survivors who are many months past their initial infection and who show no signs of active inflammation at the time of surgery [48].

The sites of EBOV persistence identified in clinical studies during the 2013–2016 epidemic were later validated in the NHP model, the gold standard for filovirus research. The landmark study by Zeng et al. [37] was the first to investigate EBOV persistence in IPNs in an animal model. The retrospective nature of using archived tissue facilitated a large sample of 112 rhesus macaques. 11.5% of animals were EBOV RT-PCR-positive in the eye, 1.3% in the epididymides and 1.25% in the brain 28–43 days after viral challenge. Consistent with clinical studies, these findings validate NHPs as a suitable model for studying EBOV persistence although the timeframe is much shorter than possible with human studies.

It is even more unclear whether the female reproductive tract and mammary glands are sites of persistence. Limited evidence suggests viable EBOV is shed into vaginal secretions up to several weeks following acute EVD [28]. However, the largest cohort study of pregnancy in EBOV survivors revealed all 367 vaginal swabs from 79 women were EBOV negative [36]. Although not an IPN, EBOV RNA has been detected in breastmilk of lactating mothers who have recovered from EVD [29, 33, 36]. Whether this means the mammary gland acts as a dormant EBOV reservoir is uncertain. Recently, adipose tissue has been identified as an additional possible site of EBOV persistence. Adipose tissue supported high levels of viral infection for 28 days without cytopathic effects during in vitro EBOV infection of human and bat-derived cell lines [51]. The study illustrates the ongoing evolution of knowledge about the sites of EBOV persistence.

Animal models have been valuable to study EBOV persistence but have also proved challenging. EVD in NHPs is almost uniformly lethal, enabling the modelling of acute disease but less so long-term persistence. Exploration of mucosal routes of inoculation and treatment with mAbs may prove useful routes to improve the NHP persistence model [44, 52]. In contrast, mice are resistant to EBOV, but mouse-adapted EBOV (MA-EBOV) strains have been developed to enhance susceptibility to infection. Ferrets have been explored more recently as an intermediary between the high lethality of EBOV in NHPs and the resistance in mice. However, given the timeframe required to study persistence, there are significant ethical and financial challenges facing animal research in this field.

Clinical studies of EBOV persistence have particularly focused on the testes, largely due to the ease and relative non-invasiveness of seminal fluid sampling and the consequential risk of sexual transmission. Longitudinal cohort studies of male survivors from the 2013–2016 EBOV epidemic used qRT-PCR to detect EBOV RNA in semen samples, the results of which can be used to generate mathematical models for survival analyses of persistence [22]. Estimates of duration and frequency of persistence vary, but overall show an initial rapid drop in number of persistent survivors with 50% survivors persisting at ~ 4 months followed by a levelling off such that a small number of individuals experience persistence for 12+ months (< 10%) [21, 29]. The longest EBOV RNA detection in semen is 988 days after EVD discharge [32]. However, the longest detection time of any virus in semen is Hantavirus with six years [13]. The conflicting estimates between longitudinal studies could be addressed using meta-analyses, however, the variation in experimental designs makes comparison of studies challenging. Similar future investigations of filovirus outbreaks should therefore aim to adhere to standardised protocols, incorporating an international standard, so all studies can be collated.

Though longitudinal studies on the detection of EBOV RNA in the semen of male survivors provide insight into the potential of virus shedding, they are all limited by the number of individuals in each study. Published studies have captured longitudinal shedding in approximately 500 male survivors, however, it is estimated that there are > 5,000 men in West Africa that survived EVD. Added to the issue of a < 10% sampling size is the paucity of repeat sampling timepoints of individuals which may miss the relatively infrequent shedding of EBOV in semen. Furthermore, we know from the 2016 and 2021 sexual transmission flare ups [21, 53], that are linked to virus testicular persistence for ~ 500 days and ~ 5 years respectively, longitudinal EBOV RNA shedding studies may underestimate the true threat of virus persistence and humans as a source of future EBOV outbreaks.

Using prior published studies on EBOV persistence in male survivors, we conducted a meta-analysis of the probability of detecting EBOV in seminal fluid [21, 24, 29]. We included all cohort studies reporting longitudinal data on Ebola RNA carriage in seminal fluids from male survivors with dates from EBOV diagnosis and used a similar cycle threshold (Ct) criteria for definition of positive, in addition to providing full access to meta data on individuals within study. We excluded one study [32] as this was a case control study preferentially sampling from Ebola survivors with known EBOV persistence. For all included studies, we classified EBOV RNA presence as a Ct value of less than 41 from at least one RT-PCR assay. From these binary classifications, we fitted a binomial model for the probability of detecting EBOV RNA in seminal fluid by days since diagnosis. To adjust for repeat measurements on individuals, we included a random intercept for each participant. Models were implemented in a Bayesian framework using Integrated Nested Laplace Approximation in R statistical software (version 4.4.0) [54]. Predictions were made using 1,000 posterior samples.

Our final analysis included 287 observations from 66 unique individuals surveyed by three studies [21, 24, 29]. The number of observations for each individual ranged from 1 to 12 (mean 4.3) from 6 to 1,557 days post diagnosis (mean 504 days). No positive individuals were detected after 925 days since diagnosis. Results showed a strong decrease over time, however, with considerable uncertainty (Figure 2). For 5,000 male EBOV survivors, this estimates a median of 7 individuals [95% confidence interval (CI): 0–43] would have detectable EBOV RNA in semen at 5 years post infection.

The cellular reservoir at sites of persistence is currently unknown. Direct tissue sampling in NHPs identified CD68⁺ macrophages as a cellular reservoir of persistence in the eye [37]. Subsequent NHP studies by the same group confirmed the same cellular EBOV reservoir was found in the brain [44] and Sertoli cells as a site of MARV persistence in the testes [38]. Furthermore, epididymal epithelial cells have been identified as a reservoir for MA-EBOV in an immunocompetent mouse model [46]. Whether the same cellular reservoir exists in humans remains to be determined.

Figure 3 shows the Sertoli cells forming the walls of seminiferous tubules that make up the testis. This creates the blood-testis barrier which prevents developing auto-immunogenic spermatocytes from being recognised as “foreign”. Sertoli cells are susceptible to Zika [55] and can sustain viral replication in vitro, suggesting these cells as a possible persistent reservoir [56]. Theoretically virus may be shed from the reservoir in the Sertoli cells and infect the developing spermatocytes as they migrate towards the lumen of the seminiferous tubule. Interestingly, the shedding of Zika virus in semen coincides with the 74 days required for renewal of spermatogonia, the stem cell population responsible for spermatocyte production, suggesting these cells may also be a contender for the site of persistence [47]. Other potential contenders include Leydig cells which produce testosterone and peritubular myoid cells which contract the tubules, although both these cells lie outside the blood-testis barrier and therefore outside the IPN. Investigations of the EBOV persistence reservoir in human testicular tissue is absent in the literature and warrants further study.

Overall, there is a reasonable evidence base for the risk and duration of persistence especially in seminal fluid. However, study of the mechanisms resulting in the establishment and maintenance of the persistent reservoir are absent. The lack of immune response in IPNs does not explain why the cytopathic effects of EBOV replication are not observed during asymptomatic persistence. The detection of antigenomic EBOV RNA using fluorescence in situ hybridisation in semen from human survivors and in tissue in NHPs indicates ongoing viral replication during persistence [23, 37]. Phylogenetic analyses of EBOV spanning years over the course of transmission events suggest a low viral evolutionary (mutation) rate [11, 53, 71]. For instance, just two new nucleotide substitutions were identified in the viral RNA sequence from a case of CSF recrudescence compared to the sequence 10 months prior during initial infection [10]. Given the infidelity of the RNA-dependent RNA polymerase of EBOV, 22 substitutions would be expected in the same period [19]. Overall, the evidence strongly suggests a very low rate of EBOV replication during persistence. The underlying mechanism for this low viral replication rate has received little attention in the literature, but there are three theories supported by conflicting bodies of evidence.

During this EBOV-free period, research is shifting away from clinical studies towards use of animal studies and in vitro models. Human tissue studies of persistent reservoirs combined with modern omics technologies could provide a valuable approach to uncover the molecular mechanisms underpinning testicular persistence. Such models could enable the screening of small molecule antivirals at a scale currently not possible and would be particularly useful during this inter-epidemic period while there are too few survivors to perform clinical trials.

The absence of reports of male-to-male EBOV sexual transmission is worrying. Emphasis on strict confidentiality and impartiality is vital in contract tracing. Animal challenge studies comparing rectal vs. vaginal EBOV exposure could prove useful in determining if anal sex has a higher risk of transmission in the same way HIV behaves. Although the limited evidence has thus far indicated the low risk of persistence and shedding from the female reproductive tract, further longitudinal studies of EBOV shedding into vaginal secretions are important. Additionally, further molecular epidemiological investigations into the origin of transmission events will help uncover the likelihood of female-to-male sexual transmission.

The testicular EBOV reservoir holds the greatest threat for human-to-human transmission and the sparking of new outbreaks. Indeed, this reservoir is already responsible for a flare-up 5 years after an initial outbreak was declared over. Maternal transmission via breastfeeding is additionally concerning. Particularly the brain, but also the eyes, are likely further sites of persistence. Although these IPNs pose less of a risk of EBOV transmission due to the contained nature of the organs, they are associated with EVD recrudescence and post-Ebola sequelae. The incidence of recrudescence may increase in future outbreaks with the distribution of life-saving EBOV vaccinations and monoclonals. To address this concern long term follow-up of survivors is key and will also help elude the connection between persistence and post-Ebola sequalae, revealing the complete long-term burden of EBOV persistence in the survivor population.

Longitudinal semen studies have provided valuable indications for the length of persistence; however, they do likely provide an underestimate. The meta-analysis presented in this review aims to address some of the variability between studies but highlights small sample sizes increase the uncertainty of predictions. Based on our estimates, a median of 7 (95% CI 0–43) male survivors from a cohort of 5,000 will be shedding EBOV in semen 5 years following infection. Although long-term persistence is uncommon, it only takes a single case to start an entirely new outbreak, emphasising the importance of long-term follow-up of survivors. Identification of the molecular mechanisms sustaining the persistent reservoirs, facilitated by in vitro models, will help inform the use of broad-spectrum antivirals to clear these sites whilst opening the door to create novel reservoir-specific antivirals. Success of such therapeutics is the ultimate goal, reducing both the risk of future outbreaks and reducing stigmatisation facing survivors in communities.

EBOV and MARV are just two of a multitude of viruses that can establish persistence, therefore benefit from understanding of this phenomenon is not limited to the filovirus field. Filovirus persistence represents a paradigm shift regarding the origin of outbreaks. However, the recent 2023 and 2024 MARV outbreaks [97, 98] and 2025 Sudan virus outbreak (belonging to the same family as EBOV) [99] illustrate this human reservoir should not overshadow the threat imposed by the vast zoonotic reservoir. In a world of increasing human-animal conflicts a One Health approach must be taken. Mitigation of spillover events and initial containment of filovirus outbreaks must maintain priorities and in fact would prevent the establishment of human persistence in the first place.