Persistent Borrelia burgdorferi Sensu Lato Infection after Antibiotic Treatment: Systematic Overview and Appraisal of the Current Evidence from Experimental Animal Models

SUMMARY

Lyme borreliosis is caused by spirochetes belonging to the Borrelia burgdorferi sensu lato group, which are transmitted by Ixodes tick species living in the temperate climate zones of the Northern Hemisphere. The clinical manifestations of Lyme borreliosis are diverse and treated with oral or intravenous antibiotics. In some patients, long-lasting and debilitating symptoms can persist after the recommended antibiotic treatment. The etiology of such persisting symptoms is under debate, and one hypothesis entails persistent infection by a subset of spirochetes after antibiotic therapy. Here, we review and appraise the experimental evidence from in vivo animal studies on the persistence of B. burgdorferi sensu lato infection after antibiotic treatment, focusing on the antimicrobial agents doxycycline and ceftriaxone. Our review indicates that some in vivo animal studies found sporadic positive cultures after antibiotic treatment. However, this culture positivity often seemed to be related to inadequate antibiotic treatment, and the few positive cultures in some studies could not be reproduced in other studies. Overall, current results from animal studies provide insufficient evidence for the persistence of viable and infectious spirochetes after adequate antibiotic treatment. Borrelial nucleic acids, on the contrary, were frequently detected in these animal studies and may thus persist after antibiotic treatment. We put forward that research into the pathogenesis of persisting complaints after antibiotic treatment for Lyme borreliosis in humans should be a top priority, but future studies should most definitely also focus on explanations other than persistent B. burgdorferi sensu lato infection after antibiotic treatment.

INTRODUCTION

Lyme borreliosis is the most common vector-borne disease in the Northern Hemisphere (1). Between 300,000 and 470,000 human cases are estimated to occur annually in the United States alone, and the number and distribution of cases are steadily increasing worldwide (2, 3). Lyme borreliosis is caused by the spirochete Borrelia burgdorferi sensu lato, which constitutes a heterogeneous group of multiple species, among which B. afzelii, B. garinii, B. bavariensis, and B. burgdorferi sensu stricto are mainly responsible for human infections in Europe (1, 2). In North America, the dominant causative agent is B. burgdorferi sensu stricto (1, 2). These spirochetes are maintained in nature by an enzootic cycle involving transmission by hard-bodied Ixodes tick species to several vertebrate hosts. Although humans are not part of the natural life cycle of B. burgdorferi sensu lato, these bacteria can be transmitted to humans when bitten by an infected tick (1, 2, 4).

The clinical signs and symptoms of human Lyme borreliosis depend on the stage and duration of infection (5, 6). Asymptomatic seroconversion for IgG has been described (7–9), but symptomatic infection usually manifests at the site of the tick bite, with a gradually expanding, erythematous skin lesion designated erythema migrans (EM) (10). If this is not recognized, is not properly treated, or does not occur, bacteria can disseminate throughout the body within days to weeks after localized infection, commonly affecting the nervous system, joints, heart, or other parts of the skin (5). After months to years, untreated early disseminated infection can progress to late disseminated infection and cause chronic disease, frequently manifesting as arthritis, although neurological or dermatological complications can also occur (1, 5).

The current internationally recommended treatment for Lyme borreliosis is a course of antibiotics, most often orally administered doxycycline or intravenously administered ceftriaxone, with a duration ranging from 10 days to 1 month depending on the stage of the disease and the affected organ (6, 10, 11). Antibiotics usually cure objective clinical manifestations of Lyme borreliosis. However, a small percentage of patients develop persistent Lyme borreliosis symptoms due to persistent infection, reinfection, an aberrant ongoing immune response, or residual damage (6). Therefore, a second course of antibiotics may be indicated in specific cases, for instance, in patients with Lyme arthritis with no or a minimal response to the initial treatment (12).

In humans, very limited evidence exists for definite persistent B. burgdorferi sensu lato infection after the recommended antibiotic treatment. Positive skin cultures have been described for a small number of patients after the recommended treatment for EM (12–19), while, for instance, a more recent study with 88 culture-positive EM patients showed no positive cultures after treatment with doxycycline (20). Culture-confirmed persistent infection in patients treated with the recommended antibiotic regimes for disseminated disease is even rarer and has been described in only a few cases (13, 21–24). In contrast, the detection of nucleic acids derived from B. burgdorferi sensu lato by PCR after antibiotic treatment has been described more regularly, predominantly in synovial fluid or synovial membranes from patients treated for Lyme arthritis (25–27). During an 11-month follow-up period, a study by Li et al. showed positive PCR results in synovial fluid in up to 67% of patients with antibiotic-refractory Lyme arthritis after the recommended antibiotic treatment (27). However, PCR copy numbers declined over time, culture and PCR on synovectomy samples were consistently negative, and the results at synovectomy and positive PCR results did not correlate with indications of relapse or ongoing arthritis. Li et al. concluded that the spirochetes in the synovial fluid were moribund or dead (27). Overall, detection of viable B. burgdorferi sensu lato in humans after the recommended antibiotic treatment is highly infrequent.

Despite the recommended antibiotic treatment, a subset of patients develops persistent symptoms without clinical or laboratory evidence of a persistent B. burgdorferi sensu lato infection. These symptoms often consist of fatigue, arthralgia, myalgia, and/or cognitive problems (28). In the current literature, the estimated prevalence of such symptoms following the treatment of Lyme borreliosis ranges from 0 to 48% (28, 29). A recent large Dutch prospective study with strict criteria for Lyme borreliosis and persistent symptoms found that the prevalence of persistent symptoms after physician-diagnosed and treated Lyme borreliosis was 27.2%, which was significantly higher than the percentage of such symptoms in the general population (21.2%). Furthermore, symptoms were more severe after treated Lyme borreliosis than in the available reference cohorts (29). These continuous or relapsing subjective symptoms are defined as post-treatment Lyme disease syndrome (PTLDS) by the Infectious Diseases Society of America (IDSA) in cases where they comprise fatigue, musculoskeletal pain, and/or cognitive complaints; occur within 6 months following a case of physician-diagnosed Lyme borreliosis that was treated with a generally accepted antibiotic regimen; persist for at least 6 months and are of such severity that daily functioning is impaired (12). However, it should also be noted that patients who might not be functionally impaired and thus would not meet the criteria for PTLDS could still experience disabling posttreatment Lyme disease symptoms. The exact mechanism underlying these persistent symptoms is unknown, and this continues to be an enigma and a topic of debate (6). Although persistent B. burgdorferi sensu lato infection has been suggested, multiple double-blind randomized controlled trials have shown no substantial and sustained effects of prolonged antibiotic treatment with either ceftriaxone intravenously or doxycycline or clarithromycin orally on patients with posttreatment Lyme disease symptoms (30–33). The only study that concluded that there is a positive effect of treatment, in this case with amoxicillin, on the health of patients with “recurrence of Lyme disease symptoms after previous successful treatment” suffers from several methodological flaws (34). Regardless, these symptoms impose a substantial burden on patients and society, as depicted by a considerable number of disability-adjusted life years and significant health care costs (35, 36), and also illustrated by the perspectives of a clinical microbiologist, a clinical infectious disease specialist, and a patient with chronic symptoms attributed to Lyme borreliosis (see Text S1 in the supplemental material). Further research into the pathophysiology of posttreatment Lyme disease syndrome and symptoms is therefore required and might reveal novel treatment modalities.

A highly discussed theory on the pathophysiology of persistent posttreatment Lyme disease symptoms postulates that these symptoms are due to antibiotic persistence by so-called persisters (37, 38). The term “persisters” thereby refers to a nongrowing bacterial subpopulation that is able to survive bactericidal antibiotic concentrations that kill the vast majority of the overall population (39, 40). In general, this is a transient, noninheritable quality that, in essence, all bacterial cells might possess. The surviving persisters are able to revert into replicating forms after relief from a stressor, such as antibiotic pressure, thereby leading to recurrent or chronic infection. The progeny of these cells is composed of both susceptible bacteria and persisters in the same proportion as in the original population. In vitro, this is demonstrated by a biphasic killing curve after the administration of antibiotics (40, 41). The existence of persisters has been described for multiple bacterial species such as Escherichia coli, Pseudomonas aeruginosa, Mycobacterium tuberculosis, Salmonella enterica subsp. enterica serovar Typhimurium, and Staphylococcus aureus (42–44). Sharma et al. described in vitro antibiotic experiments with Borrelia burgdorferi sensu lato demonstrating the above-mentioned pattern of antibiotic persistence (37). Importantly, at this moment, it is unclear whether there is a correlation between the existence of B. burgdorferi sensu lato persisters in vitro and the in vivo occurrence of borrelial persister formation, let alone the clinical relevance of such persisters in causing persistent B. burgdorferi sensu lato infection after antibiotic treatment.

The field of research on borrelial persisters as a potential explanation for persistent symptoms after antibiotic treatment for Lyme borreliosis was ignited when studies from the 1990s reported the detection of borrelial compounds by PCR in blood and urine from antibiotic-treated Lyme borreliosis patients (45, 46). Recently, a xenodiagnostic experiment in humans showed that ticks feeding on PTLDS patients did not acquire cultivable spirochetes, but borrelial DNA was acquired from one patient at two separate time points (47). However, apart from the above-mentioned treatment studies in patients with posttreatment Lyme disease symptoms (30–33) and the above-mentioned study by Marques et al. (47), clinical research assessing the correlation between persistent B. burgdorferi sensu lato infection after antibiotic treatment and posttreatment Lyme disease symptoms is scarce. Hence, in the ongoing and growing societal and scientific discussion on the role of persistent borrelial infection in patients with posttreatment Lyme disease symptoms, reference is often made to animal data, for example, as reviewed by Cabello et al. (48). However, these data have not yet been reviewed systematically.

To be able to systematically review animal data regarding persistent B. burgdorferi sensu lato infection after antibiotic treatment, unambiguous definitions should be used since in published experimental studies, the terms “antimicrobial persistence,” “persister,” and “persistent infection” are often used interchangeably. Recently, a consensus statement proposing distinguishing definitions for these terms has been published (49). Persistent infection in this consensus statement is defined as continuing infection due to, for instance, failure of the host immune system and does not include the use of antibiotics. Antibiotic persistence, on the other hand, focuses on the above-mentioned in vitro characteristics of persisters, thereby distinguishing this phenomenon from antimicrobial resistance or tolerance. Since these proposed definitions do not cover the entire spectrum of the possible causes of ongoing B. burgdorferi sensu lato infection after antibiotic treatment in vivo, we introduce the concept of “persistent B. burgdorferi sensu lato infection after antibiotic treatment” in this review. This definition includes all possible causes of recurrent or chronic infection after antibiotic treatment, including, among others, host factors (i.e., failure of the immune system), treatment factors (i.e., ineffective killing by drugs), or pathogen factors (including immune evasion and antibiotic resistance, tolerance, and/or persistence). Thereby, this review provides a comprehensive overview of current studies regarding persistent B. burgdorferi sensu lato infection after antibiotic treatment, focusing on the most clinically relevant antibiotics, in the most extensively utilized animal models for Lyme borreliosis.

METHODS

The PubMed database was searched for studies on persistent B. burgdorferi sensu lato infection after systemic antibiotic treatment in the most extensively utilized animal models, i.e., the murine, canine, and nonhuman primate models. Separate searches were performed for each of the different animal models (Table 1). The title and abstract were screened to identify relevant articles, followed by a full-text screening of the identified publications by two researchers (Y. L. Verschoor and A. Vrijlandt) independently. In this review, the primary outcome measure to evaluate persistent infection was culture positivity. Secondary outcome measures were results obtained by molecular diagnostics, xenodiagnosis, immunofluorescence assays (IFAs), direct fluorescence assays (DFAs), immunohistochemistry (IHC), serology, histopathology, allografting, and description of clinical symptoms. If available, results from in vitro analyses of positive cultures, including determinations of the MIC, were also reviewed. Studies were excluded in cases where only one secondary outcome measure was described (for instance, serology only), except when additional results from the same animals were reported separately. Other exclusion criteria were a wrong publication type, such as a review or case report; language other than English; or no full text. Data were extracted, appraised, and reported per animal model by two researchers (Y. L. Verschoor and A. Vrijlandt); discrepancies in interpretation were discussed with a third researcher (J. W. Hovius). The corresponding author was contacted in specific cases. Although papers were included irrespective of the type of antibiotic that was described, this review focuses on doxycycline and ceftriaxone, the antibiotics most clinically relevant for the treatment of human Lyme borreliosis. Results for all other antibiotics can be found in Tables S1 to S3 in the supplemental material.

TABLE 1.

Overview of the PubMed searches performed per animal model^a

Animal model	Search terms	No. of search results	No. of studies included from the search
Murine	(“mice” [MeSH] OR mus-musc* [tiab] OR “mice” [tiab] OR “mouse” [tiab] OR “murine” [tiab]) AND (“borrelia” [MeSH] OR borrel* [tiab] OR “Lyme disease” [MeSH] OR Lyme* [tiab]) AND ((“persistent infection” [MeSH] OR persister [tiab] OR persisten* [tiab]) OR (“anti-bacterial agents” [pharmacological action] OR “Spirochaetales/drug effects” [MeSH] OR anti-biotic* [tiab] OR anti-bacterial [tiab] OR doxycycline [tiab] OR ceftriaxone [tiab]))	326	30
Canine	(“dogs” [MeSH] OR dog [tiab] OR dogs [tiab] OR canine [tiab]) AND (“borrelia” [MeSH] OR borrel* [tiab] OR “Lyme disease” [MeSH] OR Lyme* [tiab]) AND ((“persistent infection” [MeSH] OR persister [tiab] OR persisten* [tiab]) OR (“anti-bacterial agents” [pharmacological action] OR “Spirochaetales/drug effects” [MeSH] OR anti-biotic* [tiab] OR anti-bacterial [tiab] OR doxycycline [tiab] OR ceftriaxone [tiab]))	80	5
Nonhuman primate	((“Haplorhini” [MeSH] NOT humans [MeSH]) OR ape [tiab] OR apes [tiab] OR monkey* [tiab] OR primate* [tiab] OR macaque* [tiab] OR macaca [tiab]) AND (“borrelia” [MeSH] OR borrel* [tiab] OR “Lyme disease” [MeSH] OR Lyme* [tiab]) AND ((“persistent infection” [MeSH] OR persister [tiab] OR persisten* [tiab]) OR (“anti-bacterial agents” [pharmacological action] OR “Spirochaetales/drug effects” [MeSH] OR anti-biotic* [tiab] or anti-bacterial [tiab] OR doxycycline [tiab] OR ceftriaxone [tiab]))	30	5

^aAll searches were performed on 6th of June 2022. * = truncation character; [tiab] = limits search to title and abstract; [MeSH] (Medical Subject Headings) = National Library of Medicine controlled vocabulary thesaurus used to index PubMed articles.

RESULTS

The Murine Model

General remarks and pharmacokinetics.

Mice have been extensively utilized as an experimental model for Lyme borreliosis. Upon inoculation, selected mouse strains are susceptible to infection and develop arthritis and carditis (50–53). Persistent borrelial infection was demonstrated by positive cultures for up to 1 year after inoculation in untreated mice, resembling human infection (54). Unlike humans, mice do not develop skin or brain lesions (50, 54). Moreover, wildlife mice are a natural reservoir host for B. burgdorferi sensu lato, implying that long-term persistent infection in mice is essential in the borrelial enzootic cycle. This should be considered when extrapolating data regarding persistent B. burgdorferi sensu lato infection, either with or without the use of antibiotics, from the murine model to humans, who are dead-end hosts rather than reservoir hosts.

On top of that, pharmacokinetics are different in mice than in humans (55). However, thorough studies on the pharmacokinetics of ceftriaxone and doxycycline in mice seem to be remarkably rare. For instance, only eight of the studies included in this review collected data on the pharmacokinetics of ceftriaxone and/or doxycycline in mice (56–63). All of these studies reported antibiotic serum levels but often solely peak and/or trough levels while lacking calculations of the primary pharmacokinetic parameters volume of distribution (Vd) and clearance (CI). In all but two studies (59, 61), serum levels were determined by the now obsolete agar well diffusion assay, possibly making pharmacokinetic results inaccurate. Besides, serum levels were collected after different modes of drug administration and at unstandardized time intervals. Additionally, protein binding was not assessed. Consequently, comparison to the human situation was often based on these limited pharmacokinetic data. Bearing the above-mentioned limitations in mind, pharmacokinetic differences in mice and humans that are apparent are discussed below.

First of all, ceftriaxone should likely be dosed more frequently in mice than in humans because of its short half-life in mice. A crucial pharmacodynamic characteristic for the efficacy of β-lactam antibiotics such as ceftriaxone is the time that the unbound serum concentrations exceed the MIC (fT_>MIC) (64–67). For ceftriaxone, an unbound serum concentration higher than the MIC is required for at least 50% of the dosing interval (68). Correspondingly, the ceftriaxone-mediated killing of B. burgdorferi sensu lato requires prolonged high serum levels (64). Early studies on the pharmacokinetics of ceftriaxone in a mouse model reported that the half-life in serum was approximately 1.5 h. Those studies furthermore stated that a subcutaneous dose of 50 mg/kg ceftriaxone in mice led to peak serum levels that are similar to peak serum levels after 1 g of intravenous ceftriaxone in humans (65, 69). More recently, Pavia et al. determined serial plasma levels of ceftriaxone by liquid chromatography and reported that these values were best fitted to a two-compartment model with a half-life of 0.5 h in the first phase and a terminal elimination half-life of 3.3 h (59). Moreover, a study by Hodzic et al. reported that ceftriaxone serum levels 2 and 4 h after the administration of a dose of 16 mg/kg exceeded the minimal bactericidal concentration (MBC) by 300- and 30-fold, respectively, while after 8 h, inhibition was undetectable (63). On the other hand, the reported half-life of ceftriaxone in humans is 5.8 to 8.7 h (70). This suggests that administration in mice should be more frequent than once daily (q.d.) to obtain the required 50% fT_>MIC during a dosing interval.

Second, doxycycline should likely be administered at a higher dose in mice than in humans. The efficacy of doxycycline is determined primarily by the ratio of the area under the concentration-time curve for the free, unbound fraction of the drug (fAUC) to the MIC (fAUC/MIC) (71). Limited data suggest that this principle also applies to Lyme borreliosis, but this requires further research (72). Nevertheless, the half-lives of doxycycline are 3 to 6 h in mice (73) and 15 to 25 h in humans (74). This might indicate that mice require a relatively high dosage of doxycycline per kg of body weight compared to humans to reach a similar fAUC value with equal dosing frequencies (twice a day [b.i.d.]).

An extensive overview of the murine studies included in this systematic overview is provided in Table S1 in the supplemental material.

B. burgdorferi sensu lato cultures in immunocompetent mice.

Multiple studies demonstrated that the most clinically used antibiotics, namely, ceftriaxone and doxycycline, effectively eliminated cultivable spirochetes from immunocompetent mice infected with B. burgdorferi sensu lato, while cultures from untreated or saline-treated control animals were usually positive (57–59, 61–63, 75–90).

(i) Ceftriaxone.

In total, 22 studies reported culture results of immunocompetent mice infected with B. burgdorferi sensu lato that were subsequently treated with ceftriaxone (38, 56–60, 63, 75–82, 84, 85, 87, 88, 90, 91). These studies reported cultures from a total of 454 mice, comprising different tissues obtained from mice of different genetic backgrounds after various treatment regimens and follow-up durations. Of note, experiments that did not report the cumulative number of mice or that used a Borrelia strain with genetically modified infectivity were not included in the total number (78, 85, 87). In 4 of these studies, positive cultures were obtained in a total of only 10/454 mice, from one tissue type per mouse, predominantly skin tissue.

First of all, in a study by Grillon et al., five mice had positive cultures after only 1 day of ceftriaxone treatment (16 mg/kg b.i.d.), while cultures were uniformly negative for all mice euthanized after 2 to 5 days of treatment (79). Apart from these positive cultures after a highly inadequate treatment duration, three other studies reported positive cultures after the completion of a course of at least 5 days of ceftriaxone: two cultures were positive in a study by Malawista et al. (91), two positive cultures were reported in a study by Hodzic et al. (63), and one culture was positive in a study by Feng et al. (38).

Culture positivity in the study by Feng et al. was reported in 1/5 ceftriaxone-treated mice infected with microcolony forms of B. burgdorferi sensu lato (38). The positive culture reported in this study was likely due to a dosing frequency that was too low, a single daily dose of 16 mg/kg q.d. for 30 days, as discussed in detail in the section above. Unfortunately, those researchers did not provide any pharmacokinetic data. Additionally, Feng et al. described results after several combination therapies (i.e., doxycycline plus ceftriaxone, vancomycin plus ceftriaxone, and doxycycline plus ceftriaxone plus daptomycin). Although the results after the administration of these combinations are described in Table S1, they are not discussed here since these combinations are not currently used in clinical practice and are beyond the scope of this review (38).

Furthermore, a study by Hodzic et al. reported the observation of spirochetes in cultures from 2/19 ceftriaxone-treated mice (16 mg/kg b.i.d. for 5 days, and subsequently q.d. for 25 days) that were euthanized 18 months after treatment (63). However, the relevance of these cultures was difficult to appreciate as it was unclear from that publication whether the described spirochetes were ever motile. Hodzic et al. describe that, after 4 weeks of incubation, these spirochetes were nonmotile with irregular curves and could not be subcultured (63). Therefore, it appears to us that the viability of these spirochetes is unclear, as well as whether these cultures should have been appreciated as positive.

Finally, Malawista et al. reported 2/15 culture-positive mice after ceftriaxone treatment (16 mg/kg b.i.d. for 5 days) (91). Bladder cultures were positive in mice euthanized at 60 days posttreatment, while bladder cultures from mice euthanized at earlier time points were consistently negative, as were all collected skin cultures. However, Malawista et al. did not describe the definition of a positive culture in this paper, for instance, regarding spirochetal motility or the ability to subculture these cells, nor do the authors describe any in vitro characteristics of the recultured B. burgdorferi sensu stricto N40 isolate, such as the MIC. Curiously, just one of the culture-positive bladders was also PCR positive (for the ospB gene only; the ospA gene PCR remained negative) (91). The culture positivity after ceftriaxone treatment in this study cannot be explained by inadequate antibiotic treatment since multiple other studies applied the same ceftriaxone dose of 16 mg/kg for 5 days (b.i.d.), after which all mice were consistently culture negative (56, 79). Malawista et al. report ex vivo contamination as a possible explanation for the positive cultures (91). Another explanation mentioned by those authors is a resurgence of spirochetes (91). While the validity of these cultures is not completely known, pathogen characteristics underlying possible persistent infection after adequate antibiotic treatment, such as antibiotic resistance, tolerance, or persistence, cannot be ruled out or confirmed due to a lack of information. In conclusion, positive cultures after a complete course of ceftriaxone treatment were highly infrequent and have not been obtained in studies other than those discussed in this section.

(ii) Doxycycline.

A total of nine studies that investigated the effectiveness of doxycycline in B. burgdorferi sensu lato-infected immunocompetent mice were identified (38, 56, 58, 61, 62, 82, 83, 85, 86). Despite most papers reporting negative posttreatment cultures, 2/9 studies described positive cultures from ear punch biopsy specimens or spleen tissue samples in a total of 28/122 doxycycline-treated mice. Again, experiments that did not report the cumulative number of mice were not included in the total number (85).

In a study by Moody et al., several antibiotics effectively eliminated cultivable spirochetes after both short and long follow-up periods (56). However, strikingly, 22/24 doxycycline-treated (1.3 or 13 mg/kg b.i.d. for 5 or 14 days) mice were culture positive. Also, a more recent study by Feng et al. obtained positive cultures in 3/5 log-phase and 3/5 stationary-phase B. burgdorferi sensu stricto-infected mice treated with doxycycline (5 mg/kg q.d. for 30 days) (38). In both studies, positive cultures can likely be explained by inadequate doxycycline dosing given that differences in pharmacokinetics between mice and humans were seemingly not accounted for. In the study by Feng et al., a low dose of 5 mg/kg was administered intraperitoneally only once daily, and serum concentrations were not measured (38). In contrast, another study reported consistently negative cultures after the administration of a 10-fold higher dose of 50 mg/kg doxycycline (q.d.) intraperitoneally (83). Therefore, the culture positivity in doxycycline-treated mice described in that study can very likely be explained by inadequate antibiotic treatment because cultivable spirochetes have not been detected in studies that apply higher doses.

Overall, culture positivity after antibiotic treatment occurs infrequently in immunocompetent mice and seems to be related predominantly to inadequate antibiotic treatment.

B. burgdorferi sensu lato cultures in immunocompromised mice.

(i) Ceftriaxone.

A total of seven studies examined persistent borrelial infection in ceftriaxone-treated immunocompromised mice (57, 59, 62, 78, 82, 88, 92). The investigated mice had genetic defects in the innate immune system (MyD88^−/− mice) or B- and/or T-cell responses (SCID mice, CD28^−/− mice, and major histocompatibility complex class II-deficient [MHC-II^−/−] mice), or the inflammatory response was inhibited by corticosteroids or anti-tumor necrosis factor alpha (TNF-α). Five of these studies showed negative cultures in mice that were treated for at least 5 days (57, 59, 62, 78, 82), while one study lacked culture results but reported negative PCR results for all treated mice (92).

In contrast, a study by Yrjänäinen et al. reported that 11 mice were positive by cultures of pinnae, bladder, and/or joint tissue out of a total of 39 immunosuppressed mice after combining ceftriaxone (50 mg/kg q.d. for 5 days) with anti-TNF-α (88). Naive mice that were subsequently infected with spirochetes recovered from a ceftriaxone- and anti-TNF-α-treated mouse developed joint swelling and were positive by culture, indicating that the recovered spirochetes had retained their virulence. Besides, the recovered spirochetes had an MIC similar to that of the spirochetes used to initially infect the mice, indicating that culture positivity was not due to bacterial resistance. It should be noted, however, that the culture positivity among immunosuppressed mice was insignificantly increased compared to the ceftriaxone-only-treated controls, for which no cultures were positive (66). The same research group later performed a similar experiment, which was reported by Salo et al., in which no positive cultures were obtained from ceftriaxone- and anti-TNF-α-treated mice (78). The authors of the latter study stated that the administration of ceftriaxone twice daily (25 mg/kg) instead of once daily (50 mg/kg), in combination with anti-TNF-α, might explain the differences in culture positivity between the studies. This commentary supports that ceftriaxone should be administered more frequently than once daily in mice.

(ii) Doxycycline.

Three papers have reported data on doxycycline-treated immunocompromised mice (62, 82, 86). A recent study by Sharma et al. found negative cultures for all doxycycline-treated C57BL/6J wild-type (WT) mice as well as several types of genetically engineered immunocompromised mice (deficient innate immune system: Toll-like receptor 2 deficient [TLR2^−/−], or MyD88^−/−; deficient adaptive immune system: muMt^−/−, T-cell receptor α deficient [TCRα^−/−], or SCID) (86). Bockenstedt et al. reported 1/12 culture-positive doxycycline-treated (0.2% in drinking water continuously for 30 days) immunocompromised B6 MyD88^−/− mice, which was significantly different from the untreated controls (3/3 positive) (62). Those authors stated that this culture positivity might have been caused by an adjusted drinking pattern of this particular mouse and, thus, “inconsistent” antibiotic levels despite seemingly adequate serum levels in a distinct cohort of mice after the administration of doxycycline in drinking water (62).

Besides the immune system, the genetic background of mice appears to play a role in the clearance of B. burgdorferi infection after antibiotic treatment, although the underlying pathophysiology is not yet clear. This could be relevant when interpreting these data, especially when bacteriostatic antibiotics such as doxycycline are used. Interestingly, the above-mentioned study by Sharma et al. compared mice of different genetic backgrounds and reported that 6/12 C3H-SCID mice remained culture positive after doxycycline treatment, while all 12 C57BL/6J-SCID mice became culture negative (86). In line with these findings, the third study reporting culture results in immunocompromised mice treated with doxycycline, conducted by Wu et al., found a comparable ratio of 4/6 culture-positive C3H-SCID mice after doxycycline treatment (82).

In summary, studies in immunocompromised B. burgdorferi sensu lato-infected antibiotic-treated mice show mainly negative cultures after ceftriaxone treatment, while culture results are more variable following doxycycline treatment, with several reported positive cultures in C3H-SCID mice. These differences in borrelial clearance between the types of administered antibiotics and the backgrounds of the mice suggest that after antibiotic clearance of the majority of spirochetes, supplementary actions of the immune system are required for complete elimination, especially in the case of bacteriostatic antibiotics.

B. burgdorferi sensu lato DNA detection.

Despite negative cultures in the majority of studies, borrelial DNA has been detected by PCR following various antibiotic regimens. In three studies, after a 30-day (50 mg/kg b.i.d. or 0.2% in drinking water, continuously) or a 5-day (50 mg/kg q.d.) course of doxycycline, cultures were negative, but borrelial DNA was detected in some mice (58, 62, 83). Additionally, DNA detection was also reported after ceftriaxone treatment of various durations (16 mg/kg b.i.d. for 5 days and then 23 or 25 days q.d., 25 mg/kg b.i.d. for 5 days, or 50 mg/kg q.d. for 5 or 18 days), although the cultures remained negative (58, 60, 63, 76, 78, 89, 90). Since the detection of DNA was consistently accompanied by negative cultures in the studies mentioned above, detectable DNA does not seem to reflect viable spirochetes and may be explained otherwise.

The quantification of DNA copy numbers generally demonstrated markedly smaller amounts of borrelial DNA in antibiotic-treated mice than in saline-treated controls (60, 63, 76, 83, 89). Additionally, studies by Hodzic et al. quantified B. burgdorferi sensu stricto DNA at consecutive time points after the completion of ceftriaxone treatment (60, 63, 76). In two of these studies, it was demonstrated that DNA levels decreased over time (60, 63). In another study, Hodzic et al. reported increasing DNA loads 12 months after treatment since flagellin (flaB) DNA became detectable again in all triplicate samples from multiple tissues in all ceftriaxone-treated mice, with copy numbers being almost equivalent to those in the saline-treated controls, suggesting spirochetal resurgence (76). Those authors stated that this borrelial upsurge was reproduced in a confirmatory experiment. However, in this confirmatory experiment, flaB DNA became detectable again in only a single tissue site in most mice, and DNA copy numbers were lower in antibiotic-treated mice than in the saline-treated controls, indicating that this upsurge was not completely reproduced (76). Moreover, antibody titers against B. burgdorferi sensu lato declined over time in two studies, including at the 12-month time point in the study reporting the alleged increased DNA load at that same moment (60, 76). These serological findings further call into question the reported borrelial resurgence since this likely would have elicited an antibody response.

B. burgdorferi sensu lato RNA detection.

Several researchers have implied that, irrespective of negative cultures, posttreatment detectable RNA suggests persistent B. burgdorferi sensu lato infection after antibiotic treatment (63, 76, 89). One study detected RNA transcripts in antibiotic-treated mice, but the transcription patterns were unlike those of the saline-treated controls (89). For example, flaB transcripts were absent in antibiotic-treated mice (89), while viable B. burgdorferi sensu lato spirochetes are known to constitutively express this gene (93). Hodzic et al. reported inconsistent results. An initial study noted that although cDNA amplification was attempted, RNA transcription could not be detected due to low levels of spirochetal nucleic acids (60). In a second study, multiple RNA transcripts were detected, but the scarcity of cDNA and the low sensitivity of detection prevented conclusions about specific RNA transcription (76). In a subsequent study, samples from treated mice constitutively contained 16S rRNA, while flaB transcripts declined over time after treatment (63).

Xenodiagnosis and allografting.

Borrelial DNA has also been detected in xenodiagnostic ticks feeding on experimentally infected mice after antibiotic treatment (58, 60, 62, 63, 76, 89). However, DNA copy numbers proved to be much lower after feeding on antibiotic-treated mice than after feeding on saline-treated controls (60, 76, 89). Interestingly, despite the transmission of borrelial DNA, none of the xenodiagnostic ticks that fed on antibiotic-treated mice were able to transmit motile spirochetes to naive SCID mice (60, 89). Moreover, none of the SCID mice fed upon by ticks from antibiotic-treated mice displayed histological evidence of inflammation of the joints or heart, while inflammatory lesions were observed in SCID mice after the feeding of ticks retrieved from saline-treated mice (89). This is of particular interest because SCID mice are known to be highly susceptible to B. burgdorferi sensu lato infection and develop severe signs of experimental Lyme borreliosis (67). Additionally, naive mice did not seroconvert to B. burgdorferi after the feeding of ticks retrieved from antibiotic-treated mice, providing further support that infectious spirochetes cannot be transmitted from antibiotic-treated mice by xenodiagnostic ticks (58).

Furthermore, the allografting of tissues or inoculation of tissue extracts from antibiotic-treated mice into naive mice generally did not transmit cultivable spirochetes (62, 77, 89). In agreement with the results of the xenodiagnostic experiments, borrelial DNA was detected at low numbers in several SCID mice after the allografting of tissues from antibiotic-treated mice but inflammatory lesions were absent in tissues from these immunocompromised animals (89). Interestingly, Hodzic et al. obtained a positive culture in one mouse after the allografting of ear tissue from a ceftriaxone-treated mouse (treated 4 months after infection and euthanized 3 months after treatment) into five naive C3H mice (60). Peculiarly, the ceftriaxone-treated mouse from which the allografted tissue that resulted in culture positivity was retrieved was negative by culture, PCR, xenodiagnosis, and immunohistochemistry. Moreover, none of the allografted tissues from saline-treated mice at the same time point (3 months after treatment) transmitted cultivable spirochetes to naive mice (60). Besides, information on this positive culture was very limited regarding amongst others: the type of tissue, the number of culture-positive tissue samples, and the motility of the cultured spirochete(s). For instance, Hodzic et al. considered a single nonmotile spirochete a “positive culture” in another table from the same paper (60). Therefore, the relevance of the finding of this single positive culture cannot be adequately appreciated.

Borrelial debris.

Intravital microscopy demonstrated the rapid elimination of B. burgdorferi sensu stricto after late treatment with ceftriaxone (4 months postinoculation) but showed the persistence of nonmotile, amorphous debris with spirochetal remnants within joint entheses and adjacent to cartilage in ear tissue, which was the tick bite site in this particular study (62). The allografting of such tissues could not transmit infection to naive mice but did induce IgG seroconversion against borrelial proteins (62).

These findings provide support for the “amber” hypothesis proposed by Wormser et al., stating that nonviable spirochetal debris could become entangled within fibrinous/collagenous tissues and stimulate immune responses by releasing antigens over time (94). Furthermore, the persistence of posttreatment DNA was shown to be dependent on the borrelial expression of decorin binding protein A (DbpA)/DbpB (78), adhesin molecules that mediate attachment to the extracellular matrix (ECM) (95). Thus, spirochetal attachment to the ECM appears to be required for detectable posttreatment DNA. Remarkably, rare nonmotile spirochetes were previously observed in cultures directly from antibiotic-treated mice or SCID mice after being fed upon by xenodiagnostic ticks retrieved from antibiotic-treated mice (60, 62, 63). Were these nonmotile organisms perhaps dead bacteria that were entangled within debris (62)? Of interest, antibiotic clearance of cultivable spirochetes could not always resolve arthritic symptoms in mice, indicating that factors other than viable B. burgdorferi sensu lato bacteria may cause ongoing inflammation or symptoms (87, 88, 92). Thus, persistent borrelial debris that contains antigens, but no viable spirochetes, may provide an explanation for posttreatment detectable nucleic acids without culture positivity.

Conclusion.

Overall, these studies in mice demonstrated the persistence of low levels of posttreatment borrelial nucleic acids, generally in the absence of cultivable spirochetes (58, 60, 62, 63, 76, 78, 83, 89, 90). Altogether, the persistence of B. burgdorferi sensu lato infection after adequately dosed antibiotic treatment with doxycycline or ceftriaxone is not supported by these findings. Detectable nucleic acids may be derived from posttreatment borrelial debris containing spirochetal remnants (62). This debris might release such antigens over time, thereby potentially stimulating immune responses (94). Whether or how such immune responses may be associated with persistent symptoms after antibiotic treatment remains to be further investigated.

The Canine Model

General remarks.

Since the 1990s, researchers have studied Lyme borreliosis in experimentally infected dogs. The most prominent symptoms manifest after 2 to 5 months as acute recurrent lameness and arthritis, which are commonly accompanied by fever and lymphadenopathy (96–100). Positive cultures and PCR results have been obtained for up to 581 days after infection, demonstrating long-term persistent B. burgdorferi sensu lato infection in untreated dogs (99). An elaborate overview of the included canine studies is given in Table S2 in the supplemental material.

Straubinger et al. performed three experiments regarding persistent B. burgdorferi sensu lato infection after antibiotic treatment in experimentally infected dogs, the results of which were published in four original papers (101–104) and a review (105). In each subsequent experiment, the follow-up time between infection and necropsy was extended to increase the opportunity of finding possible borrelial resurgence (R. K. Straubinger, personal communication). In general, scarce positive cultures were found after antibiotic treatment, a finding that could not be reproduced in the following experiments, while borrelial DNA was frequently detected by PCR (101–103, 105). Because of the few studies that have been performed in dogs, we also report on data for amoxicillin and azithromycin treatment in this section.

As for the pharmacokinetics in dogs, relevant pharmacokinetic data extracted from the included articles are discussed below in the section on culture. However, the reader should bear in mind that pharmacokinetic data in these studies were limited to serial serum levels without the reporting of the primary pharmacokinetic parameters, making it difficult to accurately compare the results from these studies with the human situation.

B. burgdorferi sensu lato cultures.

In the first experiment, skin punch biopsy specimens from all antibiotic-treated animals remained culture negative, while 2/4 postmortem cultures from treated dogs were positive (1 dog was administered doxycycline at 50 mg b.i.d. for 30 days and 1 was administered amoxicillin at 100 mg b.i.d. for 30 days) (101). These postmortem cultures were positive solely in axillary lymph nodes near the tick attachment site, while in all untreated dogs, multiple skin and postmortem cultures (25 tissue samples) were positive. Inadequate antibiotic dosing may have affected this culture positivity in the amoxicillin-treated dog. In experiment 2, serum levels were determined by bioassays after the administration of amoxicillin three times a day (t.i.d.). Straubinger et al. mention that serum amoxicillin levels dropped below the MIC 6 and 8 h after administration (101). Although the actual MIC value that was used here was not mentioned, the levels dropped below 1 mg/L after 6 h, therefore indicating that the dosing of amoxicillin twice daily in this first experiment might have resulted in insufficient time above the MIC.

In the second experiment, all cultures from skin punch biopsy specimens and postmortem tissue samples remained negative for all four amoxicillin-treated dogs (100 mg t.i.d. for 30 days). For four doxycycline-treated dogs (50 mg b.i.d. for 30 days), cultures were negative, apart from one positive culture from a single skin punch biopsy specimen taken 6 months after treatment (101). Prior and subsequent skin punch biopsy specimens as well as postmortem tissue samples from this dog were consistently negative. This pattern is unlike that for dogs with persistent infection, as shown by the consistently positive cultures obtained from untreated dogs in the same experiment (101), therefore raising the concern that this single positive culture may have been a stochastic event or perhaps even ex vivo contamination.

In the third experiment, the last and most elaborate experiment, all skin and postmortem cultures remained negative throughout the experiment in dogs treated with ceftriaxone (25 mg/kg q.d.), doxycycline (10 mg/kg b.i.d.), or azithromycin (25 mg/kg q.d.) (102, 105). Even after subsequent immunosuppressive corticosteroids, all antibiotic-treated dogs remained culture negative and asymptomatic, while untreated dogs developed severe lameness and arthritis (102).

Wormser and Schwartz suggested in their review that the differences in culture positivity in the first two experiments (101) versus the third experiment (102) could be due to stochasticity or may be explained by a higher bacterial load in the first experiments, which initiated treatment sooner after infection (60 versus 120 days) (67). Besides, as mentioned in the same review, it seems that the doxycycline plasma concentrations that were measured in the second and third experiments were lower in the second experiment than in the third experiment (despite similar dosages) (67, 101). However, quantitative interpretation of the data from the bioassay used should be done with caution. Finally, yet importantly, the fact that positive cultures were obtained only during initial experiments (101) and could not be reproduced in subsequent experiments (102) may be an indication that the authors fine-tuned the methodology of their experiment over time (R. K. Straubinger, personal communication), which is not uncommon in research, thereby further calling into question the reproducibility of the infrequently observed positive cultures.

B. burgdorferi sensu lato DNA detection.

Irrespective of cultivability, Straubinger et al. detected borrelial DNA by PCR in samples from antibiotic-treated dogs in all experiments (101–103). In the first experiments, DNA was detected in skin biopsy specimens from all infected dogs, treated and untreated, and in postmortem tissues from 62.5% of the dogs (101). Similarly, in the third experiment, there was detectable DNA in 75% of the treated dogs in more than one sample (102). It should be noted that DNA was detected in 1 to 3 postmortem samples (out of 25 different tissues) from 50% of the antibiotic-treated dogs, while numerous postmortem samples from all untreated dogs contained borrelial DNA (103). Quantification also demonstrated markedly higher DNA copy numbers in untreated controls than in treated dogs (103). Others have noted that positive PCR results in the third experiment were so scarce and at such low DNA loads that amplicon contamination and/or stochasticity may have played a role (67). Moreover, Wagner et al. later reported that borrelial DNA was absent in skin biopsy specimens from all antibiotic-treated dogs in their study, while most untreated dogs were PCR positive (106). In contrast to the experiments by Straubinger et al., older beagles were used in the study by Wagner et al., which may have affected PCR positivity. It should also be noted that in the latter study, only skin biopsy specimens were analyzed, and residual DNA in other (deeper) tissues therefore cannot be excluded.

B. burgdorferi sensu lato serology.

Additionally, multiple papers have reported strongly declining anti-C6 antibodies after antibiotic treatment, in contrast to untreated dogs, suggesting spirochetal elimination after treatment (102–104, 106). Anti-C6 antibodies declined at a higher rate and more radically than antibodies against B. burgdorferi whole-cell antigen extracts (104).

Conclusion.

Cultivable spirochetes were only sporadically recovered from antibiotic-treated dogs, and these results could not be confirmed in subsequent experiments. Antiborrelial antibodies declined after treatment, while low levels of detectable borrelial DNA may persist in antibiotic-treated dogs (101–103). Altogether, there is currently no strong evidence supporting persistent B. burgdorferi sensu lato infection after antibiotic treatment in dogs.

The Nonhuman Primate Model

General remarks and pharmacokinetics.

The multiorgan involvement of Lyme borreliosis in nonhuman primates closely resembles that of human disease (107–113). Unlike mice and dogs, monkeys can develop cutaneous lesions that contain spirochetes and show histopathological resemblance to human EM (107, 108), although the precise morphology of EM in rhesus macaques is not exactly known. Moreover, rhesus macaques can develop Lyme neuroborreliosis, affecting both the peripheral nervous system (PNS) and the central nervous system (CNS) (108–111). CNS infection was demonstrated by pleocytosis and detectable borrelial DNA in cerebrospinal fluid for up to 18 weeks after infection (108). Besides, borrelial DNA and concurrent inflammation of the PNS persisted for months to years in untreated animals (109). Nevertheless, most immunocompetent animals appeared to clear early borrelial infection without treatment, while immunosuppression resulted in disseminated disease (110–112).

Also, the pharmacokinetics of nonhuman primates may more closely resemble human pharmacokinetics than those of mice or dogs. Embers et al. performed a pharmacokinetic study with doxycycline in nonhuman primates (114). The half-life of doxycycline was determined by a bioassay and was 6.76 h after a single dose in two Indian rhesus macaques (114). This, however, seems much shorter than the half-life of 15 to 25 h in humans (74, 115, 116). Since, as mentioned above, the efficacy of doxycycline is dependent primarily on the fAUC/MIC ratio (71), it can be assumed that doxycycline should likely be administered at relatively higher doses in nonhuman primates than in humans to reach the same fAUC. However, while this study by Embers et al. calculated the AUC, those authors did not mention protein binding, which is approximately 90% in humans (https://kennisbank.knmp.nl/article/Informatorium_Medicamentorum/S547.html). Likewise, the primary pharmacokinetic parameters CI and Vd, were not mentioned (114). Of the studies by Embers et al. included in this review, one experiment lacked pharmacological data, while two out of three experiments reported only nonhuman primate peak and/or trough serum levels of doxycycline, and no fAUCs were mentioned (117, 118). In the first experiment, with the administration of 2 mg/kg doxycycline b.i.d. for 60 days after initial ceftriaxone treatment, serum peak levels exceeded the B. burgdorferi sensu lato isolate JD1 MIC of 0.31 mg/L, with a mean concentration of 1.35 mg/L, while serum trough levels dropped below this MIC (117). In the third experiment, after a doxycycline dosage of 4.2 to 5.7 mg/kg/day b.i.d. for 28 days, serum trough levels ranged from 0.98 to 1.87 mg/L in 4/5 monkeys, while one monkey had serum trough levels of 0.1 and 0.26 mg/L at weeks 19 and 20, respectively (118). In conclusion, based on the small number of animals in the pharmacokinetic study (114) and the limited pharmacological data in the other two experiments (117, 118), it seems that there is insufficient evidence to conclude that the applied doxycycline dosing regimens were sufficient to adequately treat borrelial infection or that they match the human situation.

A detailed overview of the included nonhuman primate studies is provided in Table S3 in the supplemental material.

B. burgdorferi sensu lato cultures.

Embers et al. performed three experiments with nonhuman primates, which have been reported in several papers (117–121). These authors reported that their findings support persistent B. burgdorferi sensu lato infection after antibiotic treatment in rhesus macaques, yet their conclusions were based mainly on nucleic acid detection, DFAs, and IFAs (117, 118, 121). In the second experiment, the authors reported the detection of positive postmortem cultures in 3/3 doxycycline-treated monkeys (50 mg b.i.d. for 28 days), consisting of a few spirochetes (117). However, these spirochetes could not be subcultured, calling their viability into question (117). As opposed to the other experiments performed by Embers et al., serum levels of doxycycline were not reported for this experiment (117). In all other experiments performed by the same first author, cultures were uniformly negative (117, 118, 120). An original publication reported one positive culture from an antibiotic-treated animal in the first experiment (117), but this finding has been retracted in an erratum (120). Interestingly, only one positive culture was obtained from untreated monkeys in the first experiment (117, 120), suggesting that monkeys, like humans, may be able to clear borrelial infection without treatment.

B. burgdorferi sensu lato DNA and RNA detection and indirect and direct immunofluorescence assays.

In the first experiment, after treatment with ceftriaxone (25 mg/kg q.d. for 30 days) followed by doxycycline (2 mg/kg b.i.d. for 60 days), DNA was detected in 4 postmortem tissue samples from 1 monkey, and RNA was detected in 1 to 2 tissue samples from 3/11 monkeys. Of interest, DNA and RNA detection never coincided in the same tissue, which speaks against the presence of viable spirochetes (117). A second study detected DNA in 2/5 doxycycline-treated monkeys (25 mg b.i.d. for 28 days), while RNA was undetectable (118). While both studies reported the detection of B. burgdorferi antigen (a 7.5-kDa B. burgdorferi lipoprotein and OspA, respectively) by IFAs on postmortem tissue samples (117, 118), colocalization of the B. burgdorferi sensu lato-specific 23S rRNA probe did not occur in sections reported to contain spirochetes by IFAs (121). Also, IFAs are known to suffer from reproducibility issues, and nonspecific aggregate binding could lead to false-positive results (115, 116). Since cultures were negative, such IFA results may alternatively demonstrate the presence of residual debris containing borrelial antigens. Overall, the detection of DNA and RNA as well as the IFA and DFA results may indicate residual borrelial debris containing antigens but do not provide strong evidence supporting persistent B. burgdorferi sensu lato infection.

Pathology.

Pathological examination demonstrated that no gross lesions were observed in any of the treated or untreated rhesus macaques in the included studies (117, 118, 121). In the first study, histopathology indicated inflammatory lesions in heart tissue from 3/12 antibiotic-treated monkeys, while surprisingly, no inflammation was observed among all untreated monkeys (117). In the second study, rare morphological spirochetes were observed in the CNS and heart tissues of treated monkeys, but motility could not be assessed since these were fixed sections, and subcultures of adjacent areas were not attempted. Furthermore, inflammatory lesions were present in all antibiotic-treated monkeys and 4/5 untreated monkeys but also in 3/4 uninfected controls (121). Wormser et al. previously noted that the difference between infected and uninfected animals was insignificant (116). Of interest, for this particular study, monkeys were housed in outside facilities prior to the experiments, but all monkeys were prescreened for anti-C6 antibodies by an enzyme-linked immunosorbent assay (ELISA) prior to experimentation and tested negative. Therefore, preexposure to another endemic B. burgdorferi sensu lato species does not seem to provide an explanation for the lesions observed in the uninfected controls (M. E. Embers, personal communication). Overall, these histopathological findings do not allow any meaningful conclusions regarding the existence of persistent infection after antibiotic treatment.

Xenodiagnosis.

In the second and third experiments, the researchers performed xenodiagnosis and reported positive results for some ticks that fed on antibiotic-treated animals by IFAs, DFAs, and/or nucleic acid detection. In the second experiment, the detection of intact spirochetes was noted for midgut cultures from xenodiagnostic ticks after feeding on doxycycline-treated monkeys, but neither their motility nor attempts to subculture these bacteria were reported (117). Besides, midguts from xenodiagnostic ticks that fed on untreated animals were negative by both culture and DFAs. In the third experiment, cultured midguts were uniformly negative despite the authors describing nonmotile bodies with a spirochete-like morphology (118). Curiously, in the second experiment, detectable DNA and immunofluorescence detection of spirochetes did not coincide in ticks from the same monkey (117). In the third experiment, however, transcripts of ospA and ospC were detected in xenodiagnostic ticks for which fluorescence staining was also positive (118). Embers et al. additionally reported an inability to infect SCID mice by the injection of xenodiagnostic tick contents of ticks that fed upon treated monkeys. It should also be noted that xenodiagnostic tick contents from untreated animals did not lead to infection or inflammation in SCID mice (118), which, in our view, supports that untreated monkeys may be able to clear spirochetes without treatment since the timing of this xenodiagnosis was 12 months postinfection. However, it may be that the injection of tick contents does not lead to successful infection and that effective transmission requires tick feeding for infectivity. Altogether, xenodiagnosis in antibiotic-treated nonhuman primates indicated the presence of residual nucleic acids or borrelial contents but not cultivable and/or infectious spirochetes.

B. burgdorferi sensu lato serology.

Additionally, serological data support the antibiotic clearance of spirochetes in monkeys. After treatment, multiple studies observed declining anti-C6 antibodies (117, 118). Embers et al. suggested that surviving spirochetes after antibiotic treatment are dormant, with negligible transcription of vlsE, therefore minimally stimulating anti-C6-antibody production, consequently explaining declining titers (118). In another study by Embers et al., IgM antibodies against the Borrelia antigen oligopeptide permease A2 (OppA2), a peptide transporter expressed during early B. burgdorferi infection (122), increased in 2/3 doxycycline-treated (50 mg b.i.d. for 28 days) B31-infected animals toward the end of the experiment (119). Embers et al. stated that this may be a sign of spirochete resurgence. Although an interesting thought, considering the negative cultures and the absence of symptomatic disease, in our view, these data do not support this hypothesis.

Surprisingly, one treated monkey remained C6 seronegative throughout the study despite the development of an erythematous rash, designated an EM rash in that article (118). A later comment by Wormser et al. stated that the published picture of this lesion did not resemble a characteristic EM (116). However, as mentioned above, the morphology of an EM rash in rhesus macaques has not yet been well described. Nonetheless, skin samples from this animal were culture and PCR negative, indicating the absence of spirochetes (118). Interestingly, staining with monoclonal OspA antibodies was reported to detect rare spirochetal morphologies in heart tissue from this monkey, but as also pointed out by others (116), this result seems indefinite since staining with B. burgdorferi polyclonal antibodies could not detect spirochetes in the same tissue sample (121). As also put forward by others, to us, it seems that this monkey was likely uninfected, thereby calling into question the appropriateness of the methodology utilized to evaluate persistence (116). Remarkably, a xenodiagnostic tick from this monkey had detectable spirochetes by an IFA and a positive reverse transcriptase-PCR (RT-PCR) result, also raising questions about the validity of xenodiagnosis in this particular experiment.

Conclusion.

Altogether, these studies in nonhuman primates suggest that detectable nucleic acids may persist in some tissues from antibiotic-treated rhesus macaques despite the absence of cultivable and motile spirochetes and the presence of generally declining anti-C6 antibodies (117, 118, 121). Embers et al. explain their findings by postulating that spirochetes were viable but noncultivable after host adaptation. However, borrelial host adaptation in the murine or canine model as well as in humans generally does not lead to noncultivability. Thus, in our view, another, more likely explanation is that these nucleic acids and rare nonmotile spirochetes represent nonviable bacterial remnants. Therefore, these findings cannot unambiguously confirm posttreatment persistent B. burgdorferi sensu lato infection. Moreover, the animals appeared to be completely asymptomatic after antibiotic treatment, raising further questions about the clinical significance of such persisting nucleic acids.

DISCUSSION AND CONCLUSION

This systematic review shows that studies in mice, dogs, and nonhuman primates infected with B. burgdorferi sensu lato and treated with ceftriaxone or doxycycline have regularly demonstrated the posttreatment persistence of nucleic acids, while viable and infectious spirochetes have only sporadically been found. The reported positive cultures only occurred in a small subset of animals and tissues, often in animals that received seemingly inadequate antibiotic treatment from a pharmacokinetic/pharmacodynamic point of view, while the vast majority of animals were culture negative after antibiotic treatment. Furthermore, these positive cultures could not be repeated in other studies. Overall, this review demonstrates that experimental animal models cannot provide convincing evidence of clinically relevant persistent B. burgdorferi sensu lato infection after adequate antibiotic treatment, let alone for persistent infection caused by so-called persisters.

The scientific and public debate on the etiology of persistent symptoms after antibiotic treatment for Lyme borreliosis in humans is very much alive. In these discussions, the existence of B. burgdorferi sensu lato persisters is often brought up, and reference is made to literature describing antibiotic persistence in other bacterial species that cause relapsing or chronic infections. A frequently cited paper concerns Pseudomonas aeruginosa infection in patients with cystic fibrosis (CF). Multiple samples were obtained over time from various patients, and P. aeruginosa was repeatedly cultured despite likely periodic antibiotic treatments according to the typical treatment regimen for CF patients. Upon in vitro exposure to antibiotics, these bacterial cultures showed a biphasic killing curve, and the authors claimed a first link between the existence of persisters and clinical disease (123). Reference is also frequently made to infection with Mycobacterium tuberculosis. While this organism is difficult to culture, culturing from antibiotic-treated patients became possible after extending the culture duration. These cultured M. tuberculosis bacteria were susceptible to antimicrobial treatment, as opposed to resistant organisms, thereby indicating the presence of persisters (124). B. burgdorferi sensu lato is also notoriously difficult to culture. However, the high number of positive cultures obtained from infected, untreated animals after various durations of follow-up in this review indicates that culturing B. burgdorferi sensu lato from animals is feasible. This finding also highlights that culture is a valuable technique for discriminating between infected animals and animals that cleared the infection after antibiotic treatment. Several authors, however, have proposed that culture negativity in animals treated with antibiotics for Lyme borreliosis is due to viable but noncultivable persisters (48). This theory, however, is in sharp contrast to the above-mentioned publications in which persisters from other bacterial species could be cultured from antibiotic-treated patients (123, 124). Besides, this proposed noncultivable state of B. burgdorferi persisters is contradictory to the in the introduction stated definition of an antibiotic persister, which reverts into replicable forms after the withdrawal of the stressor.

As also mentioned in the introduction, possible persistent infection after antibiotic treatment could generally be caused by host factors, treatment factors, and/or pathogen factors. Since only a few of the studies included in this review showed positive cultures and these studies provided limited information on the positive cultures, no firm conclusions can be drawn regarding these potential causes of persistent infection by B. burgdorferi sensu lato after antibiotic treatment. However, studies in genetically engineered immunocompromised mice indicate that (severe) immunodeficiency, a host factor, might lead to persistent infection after antibiotic treatment, not surprisingly primarily in the case of bacteriostatic antibiotics (62, 82, 86). Nonetheless, the most commonly observed factor that contributed to persistent infection after antibiotic administration appeared to be a treatment factor, namely, suspected inadequate treatment, because differences in pharmacokinetics, e.g., half-lives, between animals and humans appeared to be insufficiently accounted for in several studies (38, 56, 79, 101). This may have led to the application of the investigated antibiotics at dosages that were too low or with dosing intervals that were too long.

Evidence on pathogen factors that may have contributed to possible persistent infection after antibiotic treatment was scarce. While several of the included studies determined, or were aware of, the MIC of the B. burgdorferi sensu lato isolate that was used to infect the animals before the start of the experiment (38, 57, 59, 60, 62, 63, 76, 78, 81, 82, 84, 85, 88, 89, 101, 104, 117, 118, 125, 126), only one study examined the MIC of the isolate that was recultured from an infected and treated animal, in this case a ceftriaxone-treated immunosuppressed mouse (88). This posttreatment isolate had an MIC comparable to that of the isolate before antibiotic treatment, and therefore, antimicrobial resistance was ruled out. However, as mentioned above in the section on the murine model, antibiotic treatment was likely inadequate in this experiment. Therefore, a host factor and a treatment factor most likely explained the persistent B. Burgdorferi sensu lato infection after antibiotic treatment in this particular study; there was insufficient evidence for persisters being the cause of this persistent infection.

Another proposed pathogen factor for persistent infection is immune evasion by, for instance, the formation of morphological variants. There is some in vitro evidence concerning the existence of these morphological variants. Such morphological plasticity has been induced in vitro by unfavorable, sometimes supraphysiological environments, including antibiotic pressure (127–130). However, the clinical relevance of these morphological variants is controversial (131). The study by Feng et al. that was included in this review inoculated mice with log-phase, stationary-phase planktonic, and microcolony forms of B. burgdorferi sensu stricto (38). Those authors describe clinical symptoms and disease pathology following murine infection with these different morphological variants, yet this is beyond the scope of this review since these data do not describe the effects of antibiotic treatment. Importantly, there is no proof that infection caused by the inoculation of morphological variants resembles natural B. burgdorferi sensu lato infection in any way since the animal studies included in this review have not reported observations of any spherical or cystic forms by immunohistochemistry or histopathology. Nonetheless, by continuous intravital imaging in mice, Bockenstedt et al. observed two spirochetes rapidly transforming into spheres, but those authors stated that these were not bacterial cysts, and it was postulated that these morphological changes occurred due to the phagocytosis of spirochetes (62). Taken together, it seems that the rarely observed persistent B. burgdorferi sensu lato infections after antibiotic treatment in animal models can most likely be explained by host and treatment factors, while there is currently no evidence that antibiotic persistence occurred in immunocompetent animals after seemingly adequate treatment.

Some researchers have put forward that, despite negative cultures, detectable nucleic acids indicate the persistent presence of viable spirochetes. This conclusion seems insufficiently justified. Generally, no straightforward correlation exists between bacterial viability and detectable DNA and RNA (67, 132). An unrelated in vitro study, for example, has shown that bacterial Enterococcus faecalis DNA could be detected by PCR performed 1 year after cell death (133). Moreover, in vitro research showed that after the ceftriaxone-mediated killing of B. burgdorferi sensu stricto, DNA and RNA were detected for 56 and 14 days, respectively (134). In vivo environments such as the ECM might allow even longer persistence of nucleic acids. Thus, posttreatment nucleic acids may well originate from sources other than viable spirochetes.

As mentioned above in the section on the murine model, the “amber” hypothesis of Wormser et al. would help to explain the frequent PCR positivity, often without positive cultures, in this review (94). This theory states that borrelial debris entangled within fibrinous/collagenous tissues may be associated with inflammatory responses and possibly even long-term symptoms. Of interest, various studies in different animal models have demonstrated that the presence of viable spirochetes is not by definition required to elicit symptoms. Illustratively, the systemic injection of purified borrelial peptidoglycan in a mouse model induced clinical signs of arthritis with histopathological evidence of inflammation (135). Likewise, the intrathecal infusion of borrelial OspC induced prominent axonal damage in uninfected mice (136). Furthermore, a recent publication demonstrated that the ex vivo incubation of frontal cortex and dorsal root ganglion tissues from uninfected rhesus macaques with nonviable B. burgdorferi sensu stricto resulted in the release of several inflammatory mediators (137). The levels of these inflammatory mediators were often higher in response to nonviable B. burgdorferi than in response to viable spirochetes (137). These findings indicate that nonviable borrelial remnants retain pathogenicity and can lead to inflammation, thereby suggesting that the posttreatment persistence of such remnant-induced inflammation may be related to the development of persistent symptoms, which is supported by observations in humans after (repeated) antibiotic treatment for Lyme arthritis (138). Taking into account the various other possible causes of persistent symptoms, posttreatment borrelial debris and its relevance in pathology may be further explored in future research.

Various papers have suggested that animal data provide evidence for persistent borrelial infection/nonviable borrelial remnants and posttreatment Lyme disease symptoms in humans. However, these results should be extrapolated to humans with caution. First, the pharmacokinetics of the investigated antibiotics are different in animals and humans, as described in this review. However, it can be deduced from our review that an adequate intake of antibiotics is important to prevent treatment failure in patients treated for Lyme borreliosis. Second, data from research in mice should not be translated directly to humans since the immune system of laboratory mice which are kept in abnormally clean environments does not closely resemble the immunological features of human adults (139). Third, another important limitation of the murine model in this context is that, as mentioned above, mice are natural reservoir hosts for B. burgdorferi sensu lato, while humans are not, which entails fundamental differences between these species regarding borrelial persistence. Fourth, the route of infection in animal models was generally unlike that of natural human infection by a single tick bite. Dogs were commonly infected by the placement of ≥20 ticks, sometimes repeatedly (100–103). Monkeys were even injected with as many as 3.2 × 10⁸ spirochetes into multiple areas to ensure dissemination (117). Thus, borrelial loads in experimental animal infections would be markedly higher than those in naturally occurring human infections. Fifth, patients with posttreatment Lyme disease symptoms typically experience disabling subjective complaints but without objectifiable symptoms. It is difficult to investigate such subjective symptoms in animal models, and it is therefore challenging, if not impossible, to clinically translate the findings regarding borrelial nucleic acid persistence in animal models to subjective posttreatment Lyme disease symptoms in humans. In that regard, it should also be noted that objective symptoms such as swelling or histopathological lesions were commonly absent or at least much less frequent and severe in antibiotic-treated animals than in untreated controls (56, 58, 63, 78, 80, 88, 101, 102, 106).

Due to the inherent limitations of animal models, important insights into the pathophysiological mechanisms underlying posttreatment Lyme disease symptoms would likely need to come from future prospective human studies, and we strongly advocate such research. Apart from persistent B. burgdorferi sensu lato infection, there are many pathophysiological theories that aim to explain posttreatment symptoms after Lyme disease. These include the above-mentioned inflammatory responses due to residual borrelial debris (“amber” hypothesis [94]); dysfunctional or aberrant ongoing immune responses, including autoimmunity caused by molecular mimicry between borrelial and human antigens and/or underlying genetic polymorphisms; coinfections with other microorganisms; alterations of neural networks and the associated changes in neurotransmitters; and psychological factors such as beliefs and behaviors of patients regarding their symptoms (28, 29, 140, 141). Studies into these microbiological, immunological, genetic, psychological, and clinical determinants of posttreatment Lyme disease symptoms are ongoing (140). The underlying mechanisms of posttreatment Lyme disease symptoms might prove to overlap other currently unexplained postinfectious syndromes, as described for Q fever, Epstein-Barr virus infection, and, more recently, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection (142–147).

Altogether, animal studies do not provide convincing evidence that B. burgdorferi sensu lato infection persists after the recommended antibiotic treatment. Although posttreatment borrelial nucleic acids may persist after adequate treatment, this does not prove the persistence of viable and infectious spirochetes causing symptomatic disease and requiring additional or prolonged antibiotic treatment. Furthermore, the results from the animal studies discussed in this review do not allow us to draw conclusions regarding the association between posttreatment persistent infection (or persistent borrelial nucleic acids) and posttreatment Lyme disease symptoms after adequate antibiotic treatment for Lyme disease in humans.

Biographies

Y. L. Verschoor, M.D., M.Sc., obtained a double master’s degree in Medicine and Biomedical Sciences at the University of Amsterdam. She wrote her master’s thesis focusing on the posttreatment persistence of B. burgdorferi at the Amsterdam UMC, Location University of Amsterdam, Center for Experimental and Molecular Medicine. Her main interests lie with the immune system and the mechanisms involved in immune responses and immune evasion. Since 2021, she has been working as an M.D. Ph.D. candidate at the Gastrointestinal Oncology department at the Netherlands Cancer Institute, Amsterdam. Her current research focuses on immunotherapy in gastrointestinal tumors.

A. Vrijlandt is an M.D. Ph.D. candidate at the Amsterdam UMC, Location University of Amsterdam, Center for Experimental and Molecular Medicine. She is doing translational research in the field of Lyme borreliosis and mainly focuses on the antimicrobial susceptibility of Borrelia spirochetes. Besides, she is involved in patient care at the Amsterdam UMC Multidisciplinary Lyme Borreliosis Center. She finished her medical training at the University of Amsterdam in 2016. After finishing her medical training, she worked for some years in patient care, for instance, at a pediatric department and in a nursing home, before starting her Ph.D.

R. Spijker is an Information Specialist at the Amsterdam UMC and with Cochrane Netherlands located at the Julius Center for Health Sciences and Primary Care of the Utrecht University Medical Center. He had a background in immunology and molecular biology before switching to his current position. At the moment, his main interest is in systematic reviews (DTA, intervention, and prognosis), other types of literature reviews (scoping, evidence mapping, and rapid review, etc.), and the automation of systematic reviews. He is a member of the International Cochrane Council, a Cochrane information specialist executive, and a member of ICASR (International Collaboration for the Automation of Systematic Reviews). His main topics for evidence synthesis are public health and infectious diseases.

R. M. van Hest, Pharm.D., Ph.D., is a hospital pharmacist at the Amsterdam UMC, where he has worked since 2012. His main interest is in the pharmacokinetics (PK) and pharmacodynamics (PD) of antibiotics in different adult patient populations and infectious diseases. He is the principal investigator of several prospective PK/PD trials in which optimal antibiotic dosing regimens are assessed for, e.g., critically ill patients or patients with impaired renal function. He is cochairman of the antimicrobial stewardship team of the Amsterdam UMC, a member of the Antimicrobial Stewardship Committee of the Dutch Working Party on Antibiotic Policy (SWAB), and a member of the editorial board of Therapeutic Drug Monitoring. He performed his pharmacy education at the University of Utrecht and his residency at the Erasmus Medical Center Rotterdam.

H. ter Hofstede, M.D., Ph.D., is an infectious disease specialist at the Radboudumc Center for Infectious Diseases, where she has worked since 2006. She has a special interest in HIV and Lyme borreliosis, because of their multisystemic features and their impact on the patient’s daily life. Dr. ter Hofstede performed both her medical education and residency at Radboudumc, Nijmegen, and the Rijnstate Hospital, Arnhem, The Netherlands. She finished her Ph.D. trajectory about clinical aspects of toxicity in antiretroviral therapy in 2008.

K. van Kempen, B.Ec., L.L.M., studied Business Economics at Fontys Hogenscholen Eindhoven and commercial and corporate law at Maastricht University. He currently works as a secretary for the board of directors of Coram NV. Besides, he is director and secretary for the Coram Foundation and the One Coram Office Foundation and is treasurer of the latter. He has been suffering from symptoms of Lyme disease since 2004 and has had an interest in Lyme disease as a patient advocate since 2014. He has worked as director and secretary of Stichting Tekenbeetziekten (Tick Bite Disease Foundation). He became a patient advocate when he saw people suffering from long-term Lyme disease symptoms, who, in many cases, on top of that, were confronted with a variety of social consequences. Suffering from Lyme disease symptoms can cause major problems in being able to lead a normal life, ranging from having difficulty holding a job to social exclusion.

A. J. Henningsson is a specialist in infectious diseases and clinical microbiology with a position as a senior consultant in Region Jönköping County, Sweden, and also as a senior associate professor at Linköping University. Since 2016, she has been the Medical Director of the Department of Clinical Microbiology, Region Jönköping County, and leader of the Swedish National Reference Laboratory for Borreliosis in Jönköping. She has pursued research in the field of tick-borne diseases for nearly 20 years and has a special interest in laboratory diagnostics and immune responses in borreliosis. She has been Secretary of the ESCMID Study Group for Lyme Borreliosis (ESGBOR) Executive Committee since 2020.

J. W. Hovius, M.D., Ph.D., is a professor of medicine, internist, infectious disease specialist, and staff member at the Department of Internal Medicine at the Amsterdam University Medical Centers, Location AMC. He is also the founder of the Amsterdam UMC Multidisciplinary Lyme Borreliosis Center and a staff member at the Center for Experimental and Molecular Medicine. He was scientific coordinator of the European Commission-funded research project ANTIDotE and is a member of the NorthTick consortium as well as the principal investigator of several research projects funded by the Netherlands Organization for Health Research and Development (ZonMw), including the LymeProspect, LymeProspect KIDS, Ticking on Pandora’s Box, and Victory studies (www.tekenradar.nl). He is the author of numerous articles on ticks and tick-borne diseases in established scientific journals, a member of the editorial board of the journal Ticks and Tick-Borne Diseases, and a member of various national and international working groups and committees in this field.