Molecular investigation into outbreak of HIV in a Scottish prisonBMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7092.1446 (Published 17 May 1997) Cite this as: BMJ 1997;314:1446
- D L Yirrell, senior research fellowa,
- P Robertson, research techniciana,
- D J Goldberg, deputy directorb,
- J McMenamin, lecturer in public health medicineb,
- S Cameron, top grade scientistc,
- A J Leigh Brown, convenora
- a Centre for HIV Research, Institute of Cell, Animal and Population Biology, University of Edinburgh, Edinburgh EH9 3JN
- b Scottish Centre for Infection and Environmental Health, Ruchill Hospital, Glasgow G20 9BN
- c Regional Virus Laboratory, Ruchill Hospital, Glasgow G20 9NB
- Correspondence to: Dr Yirrell
- Accepted 27 February 1997
Objectives: To support already established epidemiological links between inmates of Glenochil prison positive for HIV infection by using molecular techniques and thus provide evidence of the extent of acquisition during a recent outbreak of the disease resulting from needle sharing. To identify possible sources of the outbreak, and to demonstrate the ability of the methodology to make further links beyond the original outbreak.
Design: Viral sequences obtained from the blood of HIV positive prisoners previously identified by standard epidemiological methods were compared with each other and with sequences from other Scottish patients.
Setting: Glenochil prison for men, central Scotland.
Subjects: Adult inmates and their possible contacts.
Results: Phylogenetic analysis of viral sequences in two different genomic regions showed that 13 of the 14 HIV positive prisoners had been infected from a common source. Previous research had shown that six of these had acquired their infection in Glenochil; molecular evidence suggests that more than double this number were infected while incarcerated. Virus from two long term HIV positive patients who were in the prison at the time of the outbreak but who were not identified in the original or subsequent surveys was sufficiently different to make it unlikely that they were the source. A viral sequence from heterosexual transmission from one inmate showed the ability of these techniques to follow the infection through different routes of infection.
Conclusion: The number of prisoners infected with HIV during the 1993 outbreak within Glenochil prison was more than twice that previously shown. This shows the potential for the spread of bloodborne diseases within prisons by injecting drugs.
Originally, standard serological studies established that six out of 14 HIV positive people had acquired their infection in Glenochil prison in 1993
A subsequent survey estimated that up to 20 prisoners were HIV positive at that time
Molecular linkage techniques showed that 13 of the original 14 infected men had viral sequences similar enough to indicate that one source of infection was common to all
Molecular techniques provide a powerful epidemiological tool either to link or to distance infections with HIV
Injecting drug use is a potentially explosive health care problem in prisons
Inmates who inject drugs in prison are at risk of acquiring HIV and other bloodborne infections.1 Although several investigations into the prevalence of HIV infection have been conducted in correctional institutions,2 3 4 measuring the incidence of HIV transmission within prison is difficult because of problems in identifying the probable date of transmission. The first report to provide direct evidence of an outbreak of HIV occurring within a prison documented an HIV counselling and testing initiative of inmates from HM Prison Glenochil conducted at the end of June 1993; this exercise was precipitated by the reporting of eight symptomatic cases of acute hepatitis B and two diagnosed cases of primary infection with HIV between April and June 1993.5 Widespread drug injecting and needle sharing were also reported by the prison doctor, and the possibility of an outbreak was recognised.
Of 378 inmates who were incarcerated during the survey period, 227 (60%) agreed to counselling and subsequently to an HIV test. This resulted in the identification of a further 12 cases of HIV infection. From known dates of imprisonment in Glenochil, the most recent negative result for HIV antibody, the presence of p24 antigenaemia, and clinical presentation of symptomatic primary HIV infection, it was concluded that eight inmates had acquired HIV infection while in prison. Of these eight, two could possibly have acquired HIV in another prison before transfer, thus for only six patients was there definite evidence of an infection acquired in Glenochil itself.5
HIV evolves at a rate estimated to be one million times faster than that of higher eukaryotes.6 This is because the rate of viral replication is rapid (about 1010 viral particles are generated daily)7 8 and the virus's reverse transcriptase enzyme, which is essential for replication, is error prone and lacks any proofreading capacity. Thus, any one infected person harbours a complex population of closely related but distinguishable viruses. Whereas this is a major problem for the design of therapeutic interventions, it can facilitate certain epidemiological investigations. An epidemiological link among infected people is likely if their viral nucleic acid sequences are similar and unlikely if they are different. To compare sequences from different people it is important to study the most appropriate region of the viral genome. The env gene evolves at such a rapid rate that an epidemiological link can be detected only if the interval between transmission and sampling is extremely short. In contrast, the pol gene is highly conserved in people not taking antiviral drugs and can be very similar among those who acquired their HIV infection from unrelated sources. The p17 region of the gag gene has been shown to be effective as a molecular epidemiological tool particularly when used in conjunction with the env gene.9
Molecular epidemiological analysis of nucleotide sequence data has been used in several investigations of HIV-1 transmission.10 11 12 13 14 Analysis of nucleotide sequence data established the occurrence and extent of an infection cluster associated with a dental practice in Florida10 1112 13 14 15 and distinguished between two possible sources of infection in an investigation by the Maryland Department of Health.13 Similar methods were used here to conduct a molecular epidemiological investigation on samples from the HIV infected inmates identified at the time of the counselling and testing exercise.5 The aim of the study was to determine if all diagnosed cases were, as postulated, linked and if there had been a single or multiple source of infection.
Subjects and methods
Between June and August 1993 peripheral blood mononuclear cells were obtained from 13 of the 14 HIV positive inmates identified in the previously published survey (cases 1-7 and 9-14).5 In addition, a plasma sample was obtained from the other member of the cohort (case 8), from whom a follow up whole blood sample was unavailable. Despite repeated attempts only a single viral sequence from this sample was generated. Case numbers used in this paper correspond with those published elsewhere.5
Peripheral blood mononuclear cells and plasma samples were obtained from two long term patients infected with HIV (S1 and S2, respectively) who had been identified by their general practitioners, who, in response to publicity surrounding the outbreak, realised that they had been in Glenochil during the “at risk” period (January to June 1993). They were therefore not considered part of the outbreak but as putative sources. Peripheral blood mononuclear cells were also obtained from the wife of one inmate (HC1), who seroconverted after heterosexual contact with her husband after his release from prison.
Proviral DNA and viral RNA was extracted from patients' peripheral blood mononuclear cells and plasma, respectively, and amplified by using nested polymerase chain reaction in both the V3/V4 region of the env gene and the p17 region of the gag gene.16 17 18 19 Direct sequencing of polymerase chain reaction products derived from single molecules was carried out by using an Applied Biosystems 373A automated sequencer.20 Ideally, two sequences from each region from each person were generated but because of the quality of starting material this was not always possible.
In the case of gag sequences a maximum likelihood phylogenetic tree, rooted by the inclusion of the subtype D isolate HIV-1 ELI, was obtained for duplicate sequences from cases 1, 2, 4-7, and 9-14 and single sequences from cases 3 and 8 and compared with duplicate sequences from the heterosexual contact of case 5 (HC1) and one putative source S1 (see fig 1). Single sequences within the homologous region of HIV, from seven injecting drug users, five haemophilic subjects, three heterosexual contacts of injecting drug users, all of whom were Scottish and not known to be related for HIV transmission, and 10 unrelated subtype B isolates obtained from international databases were included for comparison.
A similar tree was used to compare a 246 base pair region of the env gene (see fig 2). Duplicate sequences from Glenochil cases 1, 2, 4-7, and 9-13, a single sequence from case 3, and duplicate sequences from the sexual contact (HC1) and both putative sources S1 and S2 were aligned with the homologous region from 14 haemophilic subjects and five injecting drug users and heterosexual contacts, all of whom were Scottish and not known to be related for HIV transmission, and 10 unrelated subtype B isolates from international databases.
Analysis of p17 gag sequences
This analysis grouped 13 of the 14 Glenochil inmates into a single cluster; indeed 16 sequences from nine different people had identical sequences over the 350 base pairs used for the comparison (fig 1). The heterosexual contact of case 5 (HC1) clearly groups within the main cohort whereas the sequence from putative source S2 groups with other Scottish drug users unrelated to the outbreak. The single sequence obtained from the 14th Glenochil inmate (case 8) was distinct. Analysis in a wider background showed that it grouped among sequences found in several homosexual men from Scottish cities (A J Leigh Brown et al, unpublished data).
Env gene sequences
Figure 2) shows the maximum likelihood phylogenetic tree that illustrates that all of the 12 Glenochil patients from whom sequences were obtained in this particular region not only fall into a single cluster, supporting the gag data, but also have no specific association with either of the potential sources of the infection.
Statistical assessment of phylogenetic analysis
The phylogenetic trees for both gag and env were tested by bootstrap resampling and results presented as a percentage. The bootstrap procedure identifies clusters of sequences which are still present after the nucleotide dataset has been resampled. Clusters that are present in 70% or more bootstrap samples are considered to be well supported.15 Every significant group–that is, supported in 70% or more samples–is identified in the figures. With the single exception of case 8 the entire Glenochil cohort has a bootstrap support of 100% in the p17 gag tree and of 85% in the V3 env region tree. The only other supported clusters (apart from multiple sequences from the same patient) were those which have previously been described for haemophilic subjects in Edinburgh.19
A more specific evaluation was required to test the possibility of linkage between the major Glenochil cluster of sequences and either of the putative sources. This was done by constructing alternative trees in which either of these two patients was grouped with the cohort and testing the significance of the difference by a maximum likelihood ratio test.21 Three rearrangements of the V3 tree data were tested by artificially altering the tree to see if the result was a significantly worse fit. Versions that either grouped any individual members of the main Glenochil cluster with other Scottish drug users or that introduced putative source S2 into the main Glenochil cluster were significantly less likely (P<0.05) than that shown in figure 2). The rearrangement that grouped the other putative source (S1) with this group, however, was not significantly less likely than the tree shown.
Similar analyses on the gag data were performed. Both the inclusion of sequences from either case 8 or putative source S2 into the main Glenochil cluster resulted in a tree that was significantly less likely than the one shown (P<0.05).
Summary of results
The similarity between the viral sequences from different people is remarkable. In the gag region the average divergence within the cluster of 13 was 0.2%. This compares with an average divergence between Scottish drug users of 5.4% (P<0.001) and subtype B viruses generally of 6.7%. Even within the V3 region of the envelope gene where the average divergence within B subtypes and Scottish drug users is higher at 11.4% and 8.6%, respectively, divergence within the cohort is only 0.65%. Indeed, the values for the cohort are comparable with or even lower than those that would be expected from sequential samples obtained from the same person. As six of this group had previously been shown to have acquired their infection while in Glenochil prison,5 it follows that all 13 were part of the same epidemiological cluster. The sequence from the 14th patient (case 8) is significantly different. The two putative sources of infection (S1 and S2), identified serendipitously by local general practitioners and not by prison surveys, were known to have acquired their infection before the outbreak in Glenochil and are therefore not considered as part of this particular cohort. The sequences from these two people were sufficiently different from the cluster of 13 for us to discount linkage, although only in the case of S2 was this significant. The linkage of sequences from the heterosexual contact HC1 showed a significant linkage with the cluster of 13.
HIV in prisons
The prevalence of HIV infections in different prisons within and across countries can vary considerably. Prevalence has ranged from none in a young male offenders institution in Scotland3 to 33.6% in an adult prison in Catalonia22 and to over 50% in a female correctional facility in New York City.2 Most HIV infected inmates probably acquired their infection through injecting drug use, very little being ascribed to homosexual contact.23 In Australia an estimated half of male injecting drug users have at some time been imprisoned,4 and several reports have indicated that not only do a high proportion of drug users continue to inject and share injecting equipment while incarcerated but some inject for the first time while in prison.1 24 25 It is unknown but could be expected that outbreaks such as that reported in Glenochil prison are more widespread. Several surveys in Scottish prisons have indicated that the medical authorities have underestimated the number of HIV positive people in their prison by about 25%.1 26 For example, the initial survey of Glenochil prison in June 19935 identified 14 HIV positive prisoners, whereas a follow up study 12 months later estimated that up to 20 inmates had been part of the outbreak.27
Application of molecular technology
Before this investigation the combination of conventional serological data and dates of prison entry had been used to show that eight out of the 14 HIV positive inmates of Glenochil prison had acquired their infection while incarcerated, although two of these could have been infected in another prison before transfer. By applying molecular techniques we have shown that 13 of the 14 HIV positive subjects were infected with an almost identical virus. This finding strongly suggests that not only were these men infected while in Glenochil prison but also the infection came from a single source. As the two inmates previously known to be infected (S1 and S2) had viral sequences that were unrelated to those seen in the cohort, we conclude that the source of the infection was either an unidentified Glenochil inmate or one of the 13 cases who acquired his infection outside the prison just before being incarcerated there early in 1993. Infection was then probably transmitted by this single source directly or indirectly along a chain to either 12 or 13 inmates over a short period of time. The rapid spread of HIV among the cohort probably occurred while concentrations of circulating virus were at their highest in each subject during the interval between infection and seroconversion.28
The 14th person identified by the survey (case 8) was clearly infected with a different virus, with an average sequence divergence from the rest of the cohort of 12.4%. Only a single sample, which had shown an early banding pattern on western blot characteristic of seroconversion, was available for this patient. In the absence of a follow up sample it seems that he was infected at the same time as the others but not from the same source. Indeed, although no sex between men was reported from the Glenochil survey, analysis within a more extensive dataset shows that the sequence from case 8 is most closely related to viral sequences obtained from homosexual men, suggesting a different source of infection. In this context the Glenochil cohort will refer to the 13 patients infected with a similar viral strain.
Public health implications
The application of molecular techniques to trace the origin or chain of an outbreak of infection is not new and has been particularly useful in the context of food poisoning. For HIV the benefits of viral sequencing may also be considerable. Before 1993 no outbreak of HIV within a prison had been reported. While there was much indirect evidence to suggest that epidemics had occurred in prisons, the absence of proof made it difficult for those working in the specialties of HIV and AIDS to convince prison officials and authorities that preventive measures needed to be implemented.
The finding that at least 13 prisoners definitely acquired their infection inside Glenochil prison was the catalyst for the introduction of a wide range of harm reduction measures for people who inject in prisons throughout Scotland. These include the provision of bleach tablets for sterilising injecting equipment; a methadone detoxification programme for inmates; the availability of hepatitis B vaccine; increased training for prison officers; and improved access to drug harm minimisation counselling for prisoners. Thus the public health implications of this molecular investigation have already been considerable. As drug injecting inside prison is a global problem, it is hoped that the findings of this study will have far reaching consequences.
The Glenochil cohort strain belongs to the subtype B category, which is the predominant strain in Europe and North America. Non-B subtypes, which are prevalent in the developing world, however, are being detected increasingly in the United Kingdom as a consequence of travel and mixing of populations. Of particular concern is HIV subtype E, which is highly prevalent among heterosexuals in South East Asia and for which there is laboratory evidence of properties that suggest that it may be transmitted heterosexually more efficiently than B subtype strains (A J Leigh Brown et al, unpublished data). Already the public health laboratory service in England has identified several different viral subtypes.29 It is thus essential that surveillance of HIV infection includes not only data on numbers of prevalent cases but also information on strain subtype. Such monitoring would identify the entry of new strains into the population from abroad, the occurrence of which could have a major impact on the dynamics of HIV spread locally.
Further work that has stemmed from the findings of this study is under way. The Glenochil cohort is unique; all members are of similar age, the same sex, ethnic group, and risk category, and all were infected with the same virus at the same time. This feature affords the opportunity to study progression of HIV disease in a cohort for which viral as well as host factors are controlled. Although the cohort comprises 13 cases, there could be more. In July 1994 a voluntary anonymous prevalence study of HIV infection among inmates of Glenochil prison estimated that 20 had been infected during January to June 1993. It is therefore likely that some cases have yet to be diagnosed or associated with the Glenochil outbreak. Accordingly, approval has been obtained to link the prison register for the period January to June 1993 with the HIV register held at the Scottish Centre for Infection and Environmental Health; the latter contains details on all new laboratory diagnoses in Scotland. If a match is identified efforts will be made, on condition of consent, to verify the association through an interview with the subject and molecular investigations being conducted on his blood to compare his strain with that of the original cohort.
We thank Mr J Whitelaw (HIV Immunology, Royal Infirmary of Edinburgh, Edinburgh), Dr A J France (Kings Cross Hospital, Dundee), and Dr J R Robertson (Muirhouse Medical Centre, Edinburgh) for providing samples.
Funding: Medical Research Council programme grant (G9209918).
Conflict of interest: None.