Case Report: Disseminated, rifampicin resistant Mycobacterium bovis (BCG) infection in an immunocompromised child

Background: Bacillus Calmette–Guérin (BCG) is a live-attenuated vaccine used world-wide for prevention of tuberculosis disease. In some immunocompromised hosts it has the potential to cause disease. As with other members of the M. tuberculosis complex it has the potential for acquiring drug resistance. Methods: We reviewed 10 years of paediatric clinical BCG strains referred to our clinical microbiology laboratory in Oxford where they underwent whole genome sequencing. We present a case series comparing clinical, pathogen genetic and pathogen phenotypic data, and consider the clinical implications. Results: We identified 15 BCG isolates from 8 children under 16 years old. Only one child had clinical disease with the other seven reported as local inoculation-site reactions. Case 1 suffered disseminated disease secondary to an undiagnosed IL-12/IFNγ receptor defect and the BCG isolates evolved two different rifampicin resistance mutations. Across all 15 isolates, phenotypic resistance to each first line drug was seen. Conclusions: BCG is a safe and effective vaccine in children. Most clinical specimens in our series were not related to disease. However, in the context of rare pathogen-specific immunocompromise, BCG can cause pathology and acquire drug resistance under selection from therapy.


Introduction
The Bacillus Calmette-Guérin (BCG) vaccine has been in use worldwide for many years as prophylaxis against Mycobacterium tuberculosis infection. It is a live-attenuated vaccine derived from many cell culture passages of Mycobacterium bovis. There are multiple different M. bovis BCG strains 1 but the Danish Staten Serum Institute (SSI) 1331 strain has been the most widely used in the United Kingdom (UK).
All M. bovis BCG strains are inherently resistant to pyrazinamide. The in vitro susceptibility of vaccine strains to other antibiotic agents has been investigated by previous studies 1,2 . Unusually, the Danish BCG strain 1331 has only intermediate isoniazid susceptibility (sensitive at 0.4 μg/ml but resistant at 0.1 μg/ml) 1 and is resistant to ethionamide. In vivo, however, serum concentrations of isoniazid would often be >0.1 μg/ml 3 and its low-grade resistance has not been considered a clinical problem. Alternative vaccine strains have occasionally been used during periods of interruption to the supply of the Danish 1331 preparation in the UK. These include the InterVax (Canada) BCG vaccine (Bulgarian substrain [Sofia] SL222) 4 and the Connaught strain 2 .
BCG adenitis following vaccination is relatively common, with an incidence of 1 in 1500-2800 5 . The history has been well described, and local infection usually regresses spontaneously 6 . In contrast, disseminated BCG infection is extremely rare (estimated to affect 3.4 children per million given BCG 5 ) and is characteristically associated with primary or secondary immunodeficiency. There are no clear guidelines on how best to treat children with disseminated M. bovis BCG strain infection. However, treatment regimens generally include a prolonged course of rifampicin, isoniazid and ethambutol with adjustments made according to antibiotic susceptibility testing, led by an expert in mycobacterial infection, similar to recommendations for treating patients with multidrug-resistant (MDR) M. tuberculosis 7 .
In Oxfordshire, United Kingdom (UK), the annual incidence of TB is relatively low (approximately 4 per 100,000) and children are only given the BCG vaccine if they are in a high-risk group as defined by Public Health England (PHE) guidance 8 . This study was triggered by the isolation of a rifampicin-resistant BCG strain from a child in Oxfordshire who was vaccinated due to being in a high-risk epidemiological group, and went on to develop disseminated BCG disease. To better contextualise this case, we sought to assess the antibiotic susceptibilities of other M. bovis BCG isolates from children submitted to our clinical microbiology laboratory.

Study population
This study was conducted at Oxford University Hospitals NHS Foundation Trust, UK, a large tertiary referral teaching hospital for which specialist services serve a population of approximately 2.5 million people. Clinical isolates referred to the regional laboratory at Oxford University Hospital NHS Foundation Trust, UK, and identified as Mycobacterium tuberculosis complex have been whole-genome-sequenced using Illumina technology since 2007 9 . Following the identification of an index case of a child with disseminated BCG infection ('Case one'), we used this sequence database to identify BCG isolates from children, and reviewed clinical notes and routine microbiology results of these cases.
Phenotypic susceptibility testing Culture-based susceptibility testing was performed using liquid culture (Bactec MGIT 960) at the Public Health England National Mycobacterial Reference Laboratory, London.

Mycobacterial sequencing
For isolates identified as BCG, Illumina reads were mapped to a Mycobacterium bovis reference genome (AF2122/97 NC_002945.3) using Stampy (version 1.0.17), with repetitive regions masked. Variant calls were made using SAMtools mpileup (version 0.1.18), based on a minimum depth of 5X and at least one read on each strand. A mixed-call was assigned where the minority allele composed >10% of read depth 10 . Genotypic susceptibility predictions were based on a catalogue of resistance mutations previously described 11 .

Ethics approval
Our retrospective review of laboratory data did not require formal ethical approval but the Caldicott guardian for Oxford University Hospitals NHS Foundation Trust approved the study and we obtained written informed consent from the parents of our index case ('Case one'). Mycobacterial sequencing was undertaken as part of service development within clinical microbiology.

Case report and results
Clinical context of M. bovis samples We identified 15 clinical samples from eight children <16 years old (Table 1). Case one was a girl of Pakistani descent who first presented at 4 months of age. She had been vaccinated routinely with BCG (Danish strain 1331) soon after birth as she was from a high-risk epidemiological group. She developed disseminated BCG infection and was subsequently diagnosed with an Interleukin-12 receptor B1 (IL-12 RB1) gene mutation, which is known to increase susceptibility to mycobacterial disease 12 . The samples from the remaining seven cases all arose from either injection site collections or local adenitis, and all arose in children who had received the same vaccine strain.
Antibiotic management of M. bovis in an immuncompromised host Case one initially received rifampicin (15 mg/kg daily), isoniazid (10 mg/kg daily), pyrazinamide (35 mg/kg daily) and ethambutol (20 mg/kg daily) for presumed M. tuberculosis infection. Following identification of the isolate as M. bovis by line-probe assay in the National Mycobacterial Reference Laboratory, London, approximately two months into therapy, pyrazinamide was stopped. A month later, phenotypic susceptibilities became available, suggesting resistance to ethambutol. This agent was therefore switched to moxifloxacin (10 mg/kg daily). She received 15 months of therapy and then commenced azithromycin (10 mg/kg daily, three days a week) for prophylaxis against non-tuberculous mycobacterial infection. She subsequently relapsed with disseminated BCG infection nine months after stopping initial antimycobacterial treatment, requiring multiple surgical procedures (drainage of abdominal collections) and prolonged antibiotic therapy, which are still ongoing three years later.
Antibiotic susceptibility of M. bovis isolates in Case one Phenotypic susceptibility profiles are summarised in Table 1.
All 5 of 7 isolates with phenotypic results from this child were reported as isoniazid susceptible. In total, two of five isolates were reported as phenotypically resistant to ethambutol. In the penultimate isolate (2015), in vitro rifampicin resistance was also reported. This corresponded to the identification of a rifampicin resistance mutation (rpoB D435V) in the 2015 isolate, a well characterised mutation in the rifampicin resistance determining region that is the target of the widelyused MTB/RIF Xpert assay 13 . Interestingly, after rifampicin therapy was stopped and the patient suffered a further relapse in 2017, the organism was reported to be phenotypically susceptible and the D435V mutation was no longer present, although an alternative rpoB mutation (L430P) was present on WGS, and was also detected by Xpert MTB/RIF at the reference laboratory. The fact that the subsequent isolate had a different mutation is suggestive of a significant ongoing within-host selective pressure. That the rifampicin phenotype was susceptible despite the L430P resistance mutation is not unprecedented 13 .
Antibiotic susceptibility of other M. bovis isolates Among the other seven children from whom BCG was isolated, one exhibited phenotypic resistance to ethambutol and one to isoniazid (Table 1). Despite resistance being reported to all first-line antibiotics within our set of isolates, no single isolate was resistant to more than two drugs (i.e. more than one in addition to pyrazinamide). We were unable to identify any genetic basis for phenotypic resistance to isoniazid or ethambutol, despite searching through a list of known resistancemechanisms with 94-98% sensitivity in M. tuberculosis infection 13 . As expected, based on derivation from a single vaccine strain, the paediatric isolates were all closely related, with a maximum of seven SNPs separating any two samples ( Figure 1). Interestingly, we observed phenotypic variation in drug susceptibility even among isolates that are genomically indistinguishable.

Sequence data
All raw sequence data have been uploaded to the ENA and are available under the project number PRJNA548242.

Discussion
We have confirmed that isolation of M. bovis BCG strain from clinical samples is rare, with only eight positive cases recorded over an 11-year period in a large teaching hospital laboratory. The majority of these reflected local adenitis, which is a well-recognised complication of vaccination 14,15 . There was only one case of disseminated M. bovis BCG, which occurred in a child with an immunodeficiency known to increase susceptibility to severe mycobacterial infection 12 . In this study, we are able to draw on full-length sequence data for M. bovis to assess the presence of drug-resistance mutations and correlate these with phenotypic resistance testing. Rifampicin resistance in M. bovis (BCG) has been reported previously, in each case in an immunocompromised child, including its emergence on treatment 19-21 . Case one in our series had multiple isolates tested, with genotypic rifampicin resistance twice selected for, but only detected phenotypically once. This may be due to variations in the subpopulations of M. bovis organisms that are sampled in a patient with a high bacterial burden, or reflect antibiotic pressure (the rifampicin our patient had been receiving was stopped when the resistant isolate was detected). The presence of phenotypic resistance to isoniazid and ethambutol despite the absence of any identifiable genetic predictors of resistance is interesting: it might represent hitherto unrecognised mechanisms of resistance, or alternatively error in phenotypic determination of resistance. The latter is suggested by the fact that in vitro susceptibility reporting varies despite indistinguishable genomic sequences.
In our small series of cases we have observed in vitro resistance to each first-line anti-tuberculosis drug, despite BCG being a laboratory strain that is administered to patients. Despite known low-level isoniazid resistance in the Danish SSI 1331 BCG strain 19 , to our knowledge no underlying mechanism for this has been identified. Whatever the explanation, as far as we know the presence of resistance to any drug has only been clinically relevant in Case one in our cohort, whose primary problem was one of immunodeficiency predisposing to severe mycobacterial infection. Since local infection commonly either resolves spontaneously, or can be managed by aspiration or surgical excision 14,15 , the antibiotic resistance profile may not be clinically relevant to developing a safe management plan.
Although the BCG vaccine is contraindicated in patients with immunodeficiencies, in the UK infants in high-risk groups for developing tuberculosis are usually vaccinated soon after birth, before any immunodeficiency will have become clinically apparent. In the cases with local vaccine site collections or adenitis, it could be argued that laboratory diagnosis did not inform the management, and thus in only one of the seven cases with clinical information available was the mycobacterial culture and/or sequence data clinically useful.
It is not known how many doses of BCG are given to UK children per year, but according to Public Health England (PHE) data for high-risk areas, it will be considerably more than 30,000 doses 22 . A recent systematic review and meta-analysis 23 of the effect of BCG vaccination in children demonstrated a 19% protective efficacy against tuberculosis infection among vaccinated children after exposure compared with unvaccinated children, and protection by BCG against progression from infection to active disease of 58%. Thus, that only one case had invasive disease from all the children receiving the BCG in our large population catchment area, suggests BCG is safe, and that it is the host response that is important in the development of disease rather than any specific attribute of the vaccine strain.
In view of the dramatic changes in the development of whole genome sequencing of mycobacteria in the last few years, routine phenotypic antibiotic resistance testing has been discontinued in England in favour of genotypic testing. On the one hand this will provide more reproducible results, but on the other hand it will not detect resistance due to unknown mechanisms, such as that reported for isoniazid in the Danish BCG strain. However, work to enrich our knowledgebase of molecular resistance determinants is accelerating 13 and the clinical relevance of this low-level isoniazid resistance that is only rarely detected even by phenotypic methods is questionable. Phenotyping only detected this purported isoniazid resistance in one of our clinical isolates.
In summary, we have shown isolation of M. bovis BCG strain is rare, and that the BCG vaccine is safe and well-tolerated in immunocompetent children. Only in the setting of profound immunocompromise do we see molecular and phenotypic rifampicin resistance emerging, most likely facilitated by a large bacillary burden. Because invasive disseminated BCG disease is rare, reporting and data sharing is important to allow assimilation of experience, particularly with respect to unusual cases of antibiotic resistance. were administered to children in the study area over the study period? Any suggestions on how the single case could/should have been dealt with differently to prevent the recurrent relapses observed? Should the initial treatment period have been extended in light of the immune abnormality and ineffectiveness of PZA?

4.
Can we accept 'significant within host selective pressure' as a likely explanation in a child with significant immune compromise? It seems as if treatment induced selective pressure, differential drug penetration and selective sampling of bacterial sub-populations offer the most likely explanation in this instance.

5.
Overall this is an informative and well-presented report that makes an important contribution to the existing literature. Consideration of minor comments above should help to refine it further.

Are enough details provided of any physical examination and diagnostic tests, treatment given and outcomes? Yes
Is sufficient discussion included of the importance of the findings and their relevance to future understanding of disease processes, diagnosis or treatment? Yes Is the case presented with sufficient detail to be useful for other practitioners? Yes ensure the most appropriate anti-mycobacterial treatment course for this rare condition. It also explains why multiple drugs are included in regimens to treat patients such as these. The authors do not comment on toxicity/adverse drug reactions in this child. They also don't comment on drug levels -metabolism of rifampicin in children differs greatly to adults, and higher drug doses are required to achieve therapeutic drug levels -this child may have benefited from therapeutic drug monitoring -the authors could comment on its role in these complex cases. It may be that low levels of rifampicin contributed to the within-host selective pressure.
As it is a case report, I would be interested to see more details on the clinical case, but understand why the authors have concentrated on the drug sensitivity element of the presentation -as this has relevance for the readers, in particular if they have a similar case in the future. I wonder if the team have considered the role of phage therapy for this child, as she seems to be still be on an antibiotic regimen 3 years later.
Reporting rare cases such as these has value as clinical decisions are made on expert opinion, rather than evidence based data. This case will ensure future similar cases are considered in the context of drug sensitivity/resistance.
It would be interesting for the authors to comment on screening for primary immune deficiencies and planned changes to BCG delivery in the UK. Of particular note, this child would not have been identified by TREC screening, so the case would not have been avoided.

Is the background of the case's history and progression described in sufficient detail? Yes
Are enough details provided of any physical examination and diagnostic tests, treatment given and outcomes? Partly Is sufficient discussion included of the importance of the findings and their relevance to future understanding of disease processes, diagnosis or treatment? Yes Is the case presented with sufficient detail to be useful for other practitioners? Yes