Tuberculosis is an infectious disease caused by a bacterium called Mycobacterium tuberculosis that causes more deaths worldwide than any other infection. Individuals who are infected with the Human Immunodeficiency Virus (HIV), which weakens the immune system, are particularly vulnerable to tuberculosis. However, treating individuals who are infected with both HIV and tuberculosis is complicated because the drugs currently used to treat one infection can interfere with the effectiveness of the drugs used to treat the other.

Tuberculosis is generally treated with antibiotics. However, some strains of M. tuberculosis are difficult to treat as they have evolved to resist the effects of multiple types of antibiotics. These “multidrug-resistant” bacteria appear to be particularly common in areas where HIV infections are also common. However, it was not known whether HIV directly influences whether M. tuberculosis bacteriaevolve into drug-resistant forms.

Eldholm, Rieux et al. have now analyzed the genomes, or total genetic content, of 252 samples of M. tuberculosis taken from the largest outbreak to date of multidrug-resistant tuberculosis in South America. This made it possible to identify the genetic mutations that enable the bacteria to resist antibiotic treatment. Using mathematical models to reconstruct the spread of multidrug resistant M. tuberculosis bacteria during the outbreak also made it possible to assess who transmitted tuberculosis to whom.

The results suggest that M. tuberculosis does not evolve drug resistance any faster in patients with HIV than otherwise. Furthermore, patients infected with both HIV and tuberculosis did not transmit tuberculosis to others more often than patients who did not have HIV. However, being infected with HIV did increase the likelihood that an individual would contract tuberculosis. HIV also increased the rate at which the symptoms of tuberculosis progressed in an individual.

To clarify the effect of HIV on the spread of tuberculosis, similar studies are needed that collect more complete patient data, including their anti-HIV treatment history and their degree of immune weakening.

Among the estimated 1.5 million people who died from TB in 2013, 360,000 were HIV co-infected and 200,000 cases were caused by multidrug-resistant TB (MDR-TB) (World Health Organization, 2015). Until the late 1980s, reports of MDR-TB were rare, and transmission of such strains was even less frequent (Reves et al., 1981; Small et al., 1993; Wells et al., 2007). The MDR-TB burden surged concurrently with the human immunodeficiency virus (HIV) pandemic and most reported early MDR-TB outbreaks mainly affected HIV co-infected individuals in hospitals and prisons (Small et al., 1993; Wells et al., 2007; Ritacco et al., 1997).

There are good epidemiological reasons to suspect that the HIV and MDR-TB pandemics are fueling each other. Not only does HIV infection render people more susceptible to develop active TB by weakening their immune system, but anti-TB drugs can also directly interfere with antiretroviral treatment. Rifampicin (RIF), one of the cornerstones in anti-TB therapy, has been shown to significantly lower serum concentrations of HIV protease and reverse transcriptase inhibitors (Burman et al., 1999; Centers for Disease Control and Prevention, 1998). To make matters worse, HIV co-infection is also associated with malabsorption of anti-TB drugs. This pattern is particularly pronounced for RIF, but seems to hold true for most anti-TB drugs (Patel et al., 1995; Peloquin et al., 1993).

HIV co-infection might also directly contribute to the accumulation of resistance in Mtb. First, as resistance mutations generally entail a fitness cost to the bacterium (at least initially), some resistant strains might be more successful in HIV+ hosts with weakened immunity leading to a reduced selective pressure on the bacillus. Second, some antiretroviral drugs used to treat HIV might have a mutagenic effect on mycobacterial genomes, but this has yet to be investigated (McGrath et al., 2014).

HIV co-infection and very low CD4 lymphocyte counts (<100 cells/mm³), a hallmark of advanced HIV infection, have been shown to be risk factors for developing resistance to RIF and to a lesser degree isoniazid (INH) (Bradford et al., 1996; Burman et al., 2006; Li et al., 2005; Porco et al., 2013). However, a systematic review of 32 studies assessing MDR-TB prevalence by HIV status did not demonstrate an overall association between acquired MDR-TB and HIV, but suggested that HIV co-infection is a risk factor for contracting primary MDR-TB (Suchindran et al., 2009). In summary, the association between HIV co-infection and Mtb drug resistance remains unclear, with a number of studies yielding conflicting results (Small et al., 1993; Chum et al., 1996; Lukoye et al., 2011; Meyssonnier et al., 2012; Robert et al., 2003). Attempts have also been made to model the impact of HIV on TB incidence and resistance (Sergeev et al., 2012), but in lieu of empirical data, such studies relied on a number of assumptions on both host and pathogen biology as well as the interactions between them.

It is beyond doubt that HIV has been a driver of increased TB incidence globally, but a recent review of the subject actually found HIV co-infection to be associated with decreased rates of TB transmission within households and between close contacts (Kwan and Ernst, 2011). This observation is possibly explained by differing manifestation of TB in HIV positives, namely less frequent cavitation and lower pulmonary bacillary load (Kwan and Ernst, 2011). External factors such as social isolation or HIV infected patients being followed up more closely than HIV negatives may also contribute to this pattern (Kwan and Ernst, 2011). Indeed, in a low-incidence setting of close follow-up, HIV co-infection was associated with reduced TB transmission (inferred by clustering) and TB among HIV co-infected was at least partly due to transmission from HIV-negative patients (Fenner et al., 2012).

In the current work we aimed to directly investigate the impact of HIV co-infection on the evolution of antibiotic resistance emergence and on transmission dynamics. We analyzed the genomes of 252 isolates belonging to the largest reported outbreak of MDR-TB in South America, caused by the M strain (Ritacco et al., 1997; Eldholm et al., 2015). The isolates were collected from patients with known HIV status from the mid-90s until 2009, providing important temporal information. To assess the impact of HIV co-infection on Mtb evolutionary rates, we estimated mutation rates in the terminal branches of a time-labelled phylogenetic tree, roughly corresponding to the evolutionary history of individual clinical Mtb isolates within sampled patients. We also inferred transmission networks by implementing a novel epidemiological model accounting for the long latency of TB. Finally, we estimated the length of the latent period by combining the results of the phylogenetic reconstruction and inferred transmission networks.

We found that HIV status of the host does not affect the mutation rate of Mtb, and that drug resistance is not more likely to evolve in HIV positive than HIV negative patients. Together these findings suggest that HIV co-infection is not a direct driver of Mtb drug resistance, which fits well with the distribution of the global burden of TB, MDR-TB and HIV. Reconstructed transmission networks did not reveal a significant impact of HIV co-infection on the ability of patients to transmit TB. However, our estimates of TB latency confirm that HIV co-infection accelerates progression to active TB.

The single most important impact of HIV infection in this large multi-decade outbreak of MDR-TB seems to be an increase in the proportion of patients who develop active TB. The HIV prevalence in Argentina is approximately 0.4% (in 2001 and 2014) (World Health Organization, 2013), whereas the proportion of HIV co-infected individuals is 60.7% within the M outbreak. These numbers demonstrate that HIV infection is a massive risk factor for developing TB with the MDR M strain. We found that HIV co-infection is associated with a moderately faster, yet statistically significant, progression to active TB. As this subtle effect of HIV status on time to active TB cannot explain the far higher incidence of the M strain in HIV positives, this suggests that the main effect of HIV co-infection is to increase the absolute risk of developing active TB. In other words, we surmise that a large proportion of HIV negatives infected by the M strain will not progress to active TB but for those that do, the latency period is only slightly longer than for HIV positives.

This study encompasses an outbreak within which resistance to six common anti-TB drugs evolved early on, and our results are thus mainly restricted to the evolution of resistance to second-line drugs such as ETH and FLQ in individual isolates. Extrapolation of these findings to evolution of resistance to first-line drugs thus requires caution. However, the physiological and societal impact of HIV on TB patients as well as the fitness constraints associated with new Mtb resistance mutations should be fundamentally similar regardless of drug class.

It should be noted that free access to highly active antiretroviral therapy in Argentina from 1997 is likely to have mitigated the accelerating effect of HIV on TB progression (Waisman, 2001; Gupta et al., 2015). Clinical data on HIV progression was not available and we were thus unable to quantify the effect of anti-retroviral treatment by stratifying our analyses by CD4 counts. We predict CD4 counts would correlate negatively with TB progression. However, we do not expect that antiviral therapy (and increased CD4 counts) would nullify the effect of HIV on TB progression. Indeed, it has been shown that TB incidence during highly active anti-retroviral treatment is significantly higher than background levels even though a number of possible confounders makes the exact quantification of the effect of antiretroviral therapy challenging (Gupta et al., 2015; Girardi et al., 2005; Lawn et al., 2005; Lawn and Wood, 2005).

Accurate reconstruction of transmission chains of bacteria with extended periods of within-host evolution remains challenging (see Materials and methods). But even though some artefactual transmissions are bound to be included among the reconstructed high-confidence events, we are confident that the overall pattern of transmission is shaped by actual events and is hence robust. Restricting the transmission network analyses to the most likely transmission events did not affect our finding that HIV status does not significantly impact the transmissibility of Mtb (Table 3 and Table 4). We also found that HIV co-infection does not affect the rate of Mtb evolution within patients. In fact, Mtb was found to accumulate more mutations in HIV negatives. This likely reflects the slower progression to active disease in this group, with these patients harboring Mtb for a more extended period relative to HIV positives. This pattern holds true also for antimicrobial resistance mutations, which were found to evolve significantly more often in HIV negatives than in HIV positives.

We previously showed that the largest clade in the M outbreak had evolved resistance to six antimicrobials by 1979, well before the HIV epidemic reached Argentina (Eldholm et al., 2015), a finding which has been replicated for another highly resistant Mtb lineage in South Africa (Cohen et al., 2015). To put our results in a global context, we retrieved data on the burden of TB, MDR-TB and HIV globally from the World Health Organization (WHO) Global Health Observatory Data Repository (http://apps.who.int/gho/data/node.main). We observed a strong correlation between TB and MDR-TB prevalence (Figure 5a) as well as a correlation between HIV and TB burden between countries (Figure 5b). We also recovered a highly significant correlation between HIV and MDR-TB (Figure 5c). However, when correcting the MDR-TB burden for total TB burden, the correlation vanished (Figure 5d). This is in line with our results on the M outbreak that HIV is a driver of TB in general, but does not disproportionately contribute to the rise of MDR-TB lineages.

Figure 5 with 1 supplement see all

Download asset Open asset

Figure 5—source data 1

Global per-country health, economy and disease metrics.

https://doi.org/10.7554/eLife.16644.018

Download elife-16644-fig5-data1-v1.xlsx

By combining Bayesian evolutionary analyses and the reconstruction of transmission networks based on a new epidemiological model, we were able to directly assess the impact of HIV on the evolution and transmission of the single most widespread MDR-TB strain reported to date in South America. The main pre-extensively resistant (pre-XDR) clade within the outbreak evolved before the HIV epidemic in Argentina, but HIV patients at a major hospital in Buenos Aires played a central role in fueling the epidemic in the early 1990s (Ritacco et al., 1997; Eldholm et al., 2015), by providing the strain with a large and spatially restricted reservoir of individuals susceptible to develop active TB. Once the outbreak erupted, we find that HIV co-infection did not play a role in accelerating Mtb mutation rates; neither did HIV co-infected patients cause secondary TB cases at significantly higher rates than their HIV negative peers did. Our findings confirm that HIV co-infected patients have increased susceptibility to contract TB, but strongly suggest that they do not drive the evolution of Mtb resistance within an outbreak, nor do they act as super-spreaders of MDR-TB.

1. 1. Eldholm V
   2. Monteserin J
   3. Rieux A
   4. Lopez B
   5. Sobkowiak B
   6. Ritacco V
   7. Balloux F

   (2015) Whole genome analysis of time-structured samples from an outbreak of Mycobacterium tuberculosis M strain in Buenos Aires

   Publicly available at the EBI European Nucleotide Archive (accession no: PRJEB7669).

   http://www.ebi.ac.uk/ena/data/view/PRJEB7669

1. 1. Biek R
   2. Pybus OG
   3. Lloyd-Smith JO
   4. Didelot X

   (2015) Measurably evolving pathogens in the genomic era

   Trends in Ecology & Evolution 30:306–313.

   https://doi.org/10.1016/j.tree.2015.03.009

   • Google Scholar
2. 1. Bradford WZ
   2. Martin JN
   3. Reingold AL
   4. Schecter GF
   5. Hopewell PC
   6. Small PM

   (1996) The changing epidemiology of acquired drug-resistant tuberculosis in San Francisco, USA

   The Lancet 348:928–931.

   https://doi.org/10.1016/S0140-6736(96)03027-9

   • Google Scholar
3. 1. Brust JC
   2. Berman AR
   3. Zalta B
   4. Haramati LB
   5. Ning Y
   6. Heo M
   7. van der Merwe TL
   8. Bamber S
   9. Moll AP
   10. Friedland GH
   11. Shah NS
   12. Gandhi NR

   (2013) Chest radiograph findings and time to culture conversion in patients with multidrug-resistant tuberculosis and HIV in Tugela Ferry, South Africa

   PLoS One 8:e73975.

   https://doi.org/10.1371/journal.pone.0073975

   • Google Scholar
4. 1. Burman W
   2. Benator D
   3. Vernon A
   4. Khan A
   5. Jones B
   6. Silva C
   7. Lahart C
   8. Weis S
   9. King B
   10. Mangura B
   11. Weiner M
   12. El-Sadr W
   13. Tuberculosis Trials Consortium

   (2006) Acquired rifamycin resistance with twice-weekly treatment of HIV-related tuberculosis

   American Journal of Respiratory and Critical Care Medicine 173:350–356.

   https://doi.org/10.1164/rccm.200503-417OC

   • Google Scholar
5. 1. Burman WJ
   2. Gallicano K
   3. Peloquin C

   (1999) Therapeutic implications of drug interactions in the treatment of human immunodeficiency virus-related tuberculosis

   Clinical Infectious Diseases 28:419–429.

   https://doi.org/10.1086/515174

   • Google Scholar
6. Software

   1. Carey V
   2. Long L
   3. Gentleman R

   (2016)

   RBGL: An interface to the BOOST graph library, version R package version 1.44.0

   Bioconductor.
7. 1. Cauchemez S
   2. Carrat F
   3. Viboud C
   4. Valleron AJ
   5. Boëlle PY

   (2004) A bayesian MCMC approach to study transmission of influenza: application to household longitudinal data

   Statistics in Medicine 23:3469–3487.

   https://doi.org/10.1002/sim.1912

   • Google Scholar
8. 1. Centers for Disease Control and Prevention

   (1998)

   Prevention and treatment of tuberculosis among patients infected with human immunodeficiency virus: principles of therapy and revised recommendations

   Morbidity and Mortality Weekly Report 47:1–58.

   • Google Scholar
9. 1. Chum HJ
   2. O'Brien RJ
   3. Chonde TM
   4. Graf P
   5. Rieder HL

   (1996) An epidemiological study of tuberculosis and HIV infection in Tanzania, 1991-1993

   AIDS 10:299–310.

   https://doi.org/10.1097/00002030-199603000-00009

   • Google Scholar
10. 1. Cohen KA
    2. Abeel T
    3. Manson McGuire A
    4. Desjardins CA
    5. Munsamy V
    6. Shea TP
    7. Walker BJ
    8. Bantubani N
    9. Almeida DV
    10. Alvarado L
    11. Chapman SB
    12. Mvelase NR
    13. Duffy EY
    14. Fitzgerald MG
    15. Govender P
    16. Gujja S
    17. Hamilton S
    18. Howarth C
    19. Larimer JD
    20. Maharaj K
    21. Pearson MD
    22. Priest ME
    23. Zeng Q
    24. Padayatchi N
    25. Grosset J
    26. Young SK
    27. Wortman J
    28. Mlisana KP
    29. O'Donnell MR
    30. Birren BW
    31. Bishai WR
    32. Pym AS
    33. Earl AM

    (2015) Evolution of extensively drug-resistant tuberculosis over Four decades: whole genome sequencing and dating analysis of mycobacterium tuberculosis Isolates from kwaZulu-natal

    PLoS Medicine 12:e1001880.

    https://doi.org/10.1371/journal.pmed.1001880

    • Google Scholar
11. 1. Didelot X
    2. Eyre DW
    3. Cule M
    4. Ip CL
    5. Ansari MA
    6. Griffiths D
    7. Vaughan A
    8. O'Connor L
    9. Golubchik T
    10. Batty EM
    11. Piazza P
    12. Wilson DJ
    13. Bowden R
    14. Donnelly PJ
    15. Dingle KE
    16. Wilcox M
    17. Walker AS
    18. Crook DW
    19. Peto TE
    20. Harding RM

    (2012) Microevolutionary analysis of clostridium difficile genomes to investigate transmission

    Genome Biology 13:R118.

    https://doi.org/10.1186/gb-2012-13-12-r118

    • Google Scholar
12. 1. Didelot X
    2. Gardy J
    3. Colijn C

    (2014) Bayesian inference of infectious disease transmission from whole-genome sequence data

    Molecular Biology and Evolution 31:1869–1879.

    https://doi.org/10.1093/molbev/msu121

    • Google Scholar
13. 1. Didelot X
    2. Nell S
    3. Yang I
    4. Woltemate S
    5. van der Merwe S
    6. Suerbaum S

    (2013) Genomic evolution and transmission of helicobacter pylori in two South African families

    PNAS 110:13880–13885.

    https://doi.org/10.1073/pnas.1304681110

    • Google Scholar
14. 1. Didelot X
    2. Walker AS
    3. Peto TE
    4. Crook DW
    5. Wilson DJ

    (2016) Within-host evolution of bacterial pathogens

    Nature Reviews Microbiology 14:150–162.

    https://doi.org/10.1038/nrmicro.2015.13

    • Google Scholar
15. 1. Drummond AJ
    2. Rambaut A

    (2007) BEAST: Bayesian evolutionary analysis by sampling trees

    BMC Evolutionary Biology 7:214.

    https://doi.org/10.1186/1471-2148-7-214

    • Google Scholar
16. 1. Eldholm V
    2. Monteserin J
    3. Rieux A
    4. Lopez B
    5. Sobkowiak B
    6. Ritacco V
    7. Balloux F

    (2015) Four decades of transmission of a multidrug-resistant mycobacterium tuberculosis outbreak strain

    Nature Communications 6:7119.

    https://doi.org/10.1038/ncomms8119

    • Google Scholar
17. 1. Eldholm V
    2. Norheim G
    3. von der Lippe B
    4. Kinander W
    5. Dahle UR
    6. Caugant DA
    7. Mannsåker T
    8. Mengshoel AT
    9. Dyrhol-Riise AM
    10. Balloux F

    (2014) Evolution of extensively drug-resistant Mycobacterium tuberculosis from a susceptible ancestor in a single patient

    Genome Biology 15:490.

    https://doi.org/10.1186/s13059-014-0490-3

    • Google Scholar
18. 1. Fenner L
    2. Gagneux S
    3. Helbling P
    4. Battegay M
    5. Rieder HL
    6. Pfyffer GE
    7. Zwahlen M
    8. Furrer H
    9. Siegrist HH
    10. Fehr J
    11. Dolina M
    12. Calmy A
    13. Stucki D
    14. Jaton K
    15. Janssens JP
    16. Stalder JM
    17. Bodmer T
    18. Ninet B
    19. Böttger EC
    20. Egger M
    21. Swiss HIV Cohort Study
    22. Molecular epidemiology of tuberculosis study group

    (2012) Mycobacterium tuberculosis transmission in a country with low tuberculosis incidence: role of immigration and HIV infection

    Journal of Clinical Microbiology 50:388–395.

    https://doi.org/10.1128/JCM.05392-11

    • Google Scholar
19. 1. Girardi E
    2. Sabin CA
    3. d'Arminio Monforte A
    4. Hogg B
    5. Phillips AN
    6. Gill MJ
    7. Dabis F
    8. Reiss P
    9. Kirk O
    10. Bernasconi E
    11. Grabar S
    12. Justice A
    13. Staszewski S
    14. Fätkenheuer G
    15. Sterne JA
    16. Antiretroviral Therapy Cohort Collaboration

    (2005) Incidence of tuberculosis among HIV-infected patients receiving highly active antiretroviral therapy in Europe and North America

    Clinical Infectious Diseases 41:1772–1782.

    https://doi.org/10.1086/498315

    • Google Scholar
20. 1. Gupta RK
    2. Rice B
    3. Brown AE
    4. Thomas HL
    5. Zenner D
    6. Anderson L
    7. Pedrazzoli D
    8. Pozniak A
    9. Abubakar I
    10. Delpech V
    11. Lipman M

    (2015) Does antiretroviral therapy reduce HIV-associated tuberculosis incidence to background rates? A national observational cohort study from England, Wales, and Northern Ireland

    The Lancet HIV 2:e243–e251.

    https://doi.org/10.1016/S2352-3018(15)00063-6

    • Google Scholar
21. 1. Kwan CK
    2. Ernst JD

    (2011) HIV and tuberculosis: a deadly human syndemic

    Clinical Microbiology Reviews 24:351–376.

    https://doi.org/10.1128/CMR.00042-10

    • Google Scholar
22. 1. Lawn SD
    2. Badri M
    3. Wood R

    (2005) Tuberculosis among HIV-infected patients receiving HAART: long term incidence and risk factors in a South African cohort

    AIDS 19:2109–2116.

    https://doi.org/10.1097/01.aids.0000194808.20035.c1

    • Google Scholar
23. 1. Lawn SD
    2. Wood R

    (2005) Incidence of tuberculosis during highly active antiretroviral therapy in high-income and low-income countries

    Clinical Infectious Diseases 41:1783–1786.

    https://doi.org/10.1086/498308

    • Google Scholar
24. 1. Lekone PE
    2. Finkenstädt BF

    (2006) Statistical inference in a stochastic epidemic SEIR model with control intervention: Ebola as a case study

    Biometrics 62:1170–1177.

    https://doi.org/10.1111/j.1541-0420.2006.00609.x

    • Google Scholar
25. 1. Li J
    2. Munsiff SS
    3. Driver CR
    4. Sackoff J

    (2005) Relapse and acquired rifampin resistance in HIV-infected patients with tuberculosis treated with rifampin- or rifabutin-based regimens in New York City, 1997-2000

    Clinical Infectious Diseases 41:83–91.

    https://doi.org/10.1086/430377

    • Google Scholar
26. 1. Lukoye D
    2. Cobelens FG
    3. Ezati N
    4. Kirimunda S
    5. Adatu FE
    6. Lule JK
    7. Nuwaha F
    8. Joloba ML

    (2011) Rates of anti-tuberculosis drug resistance in kampala-uganda are low and not associated with HIV infection

    PLoS One 6:e16130.

    https://doi.org/10.1371/journal.pone.0016130

    • Google Scholar
27. 1. Ma M-J
    2. Yang Y
    3. Wang H-B
    4. Zhu Y-F
    5. Fang L-Q
    6. An X-P
    7. Wan K-L
    8. Whalen CC
    9. Yang X-X
    10. Lauzardo M
    11. Zhang Z-Y
    12. Cao J-F
    13. Tong Y-G
    14. Dai E-H
    15. Cao W-C

    (2015) Transmissibility of tuberculosis among school contacts: An outbreak investigation in a boarding middle school, China

    Infection, Genetics and Evolution 32:148–155.

    https://doi.org/10.1016/j.meegid.2015.03.001

    • Google Scholar
28. 1. McGrath M
    2. Gey van Pittius NC
    3. van Helden PD
    4. Warren RM
    5. Warner DF

    (2014) Mutation rate and the emergence of drug resistance in mycobacterium tuberculosis

    Journal of Antimicrobial Chemotherapy 69:292–302.

    https://doi.org/10.1093/jac/dkt364

    • Google Scholar
29. 1. Meyssonnier V
    2. Veziris N
    3. Bastian S
    4. Texier-Maugein J
    5. Jarlier V
    6. Robert J

    (2012) Increase in primary drug resistance of mycobacterium tuberculosis in younger birth cohorts in France

    Journal of Infection 64:589–595.

    https://doi.org/10.1016/j.jinf.2012.01.013

    • Google Scholar
30. 1. Patel KB
    2. Belmonte R
    3. Crowe HM

    (1995) Drug malabsorption and resistant tuberculosis in HIV-infected patients

    New England Journal of Medicine 332:336–337.

    https://doi.org/10.1056/NEJM199502023320518

    • Google Scholar
31. 1. Peloquin CA
    2. MacPhee AA
    3. Berning SE

    (1993) Malabsorption of antimycobacterial medications

    New England Journal of Medicine 329:1122–1123.

    https://doi.org/10.1056/NEJM199310073291513

    • Google Scholar
32. 1. Porco TC
    2. Oh P
    3. Flood JM

    (2013) Antituberculosis drug resistance acquired during treatment: An analysis of cases reported in California, 1994-2006

    Clinical Infectious Diseases 56:761–769.

    https://doi.org/10.1093/cid/cis989

    • Google Scholar
33. 1. Pybus OG
    2. Rambaut A

    (2009) Evolutionary analysis of the dynamics of viral infectious disease

    Nature Reviews Genetics 10:540–550.

    https://doi.org/10.1038/nrg2583

    • Google Scholar
34. 1. Reves R
    2. Blakey D
    3. Snider DE
    4. Farer LS

    (1981)

    Transmission of multiple drug-resistant tuberculosis: report of a school and community outbreak

    American Journal of Epidemiology 113:423–435.

    • Google Scholar
35. 1. Ritacco V
    2. Di Lonardo M
    3. Reniero A
    4. Ambroggi M
    5. Barrera L
    6. Dambrosi A
    7. Lopez B
    8. Isola N
    9. de Kantor IN

    (1997) Nosocomial spread of human immunodeficiency virus-related multidrug-resistant tuberculosis in Buenos Aires

    Journal of Infectious Diseases 176:637–642.

    https://doi.org/10.1086/514084

    • Google Scholar
36. 1. Robert J
    2. Trystram D
    3. Truffot-Pernot C
    4. Jarlier V

    (2003) Multidrug-resistant tuberculosis: eight years of surveillance in France

    European Respiratory Journal 22:833–837.

    https://doi.org/10.1183/09031936.03.00014103

    • Google Scholar
37. 1. Sergeev R
    2. Colijn C
    3. Murray M
    4. Cohen T

    (2012) Modeling the dynamic relationship between HIV and the risk of drug-resistant tuberculosis

    Science Translational Medicine 4:135ra67.

    https://doi.org/10.1126/scitranslmed.3003815

    • Google Scholar
38. 1. Shea KM
    2. Kammerer JS
    3. Winston CA
    4. Navin TR
    5. Horsburgh CR

    (2014) Estimated rate of reactivation of latent tuberculosis infection in the United States, overall and by population subgroup

    American Journal of Epidemiology 179:216–225.

    https://doi.org/10.1093/aje/kwt246

    • Google Scholar
39. 1. Small PM
    2. Shafer RW
    3. Hopewell PC
    4. Singh SP
    5. Murphy MJ
    6. Desmond E
    7. Sierra MF
    8. Schoolnik GK

    (1993) Exogenous reinfection with multidrug-resistant mycobacterium tuberculosis in patients with advanced HIV infection

    New England Journal of Medicine 328:1137–1144.

    https://doi.org/10.1056/NEJM199304223281601

    • Google Scholar
40. 1. Sreeramareddy CT
    2. Panduru KV
    3. Menten J
    4. Van den Ende J

    (2009) Time delays in diagnosis of pulmonary tuberculosis: a systematic review of literature

    BMC Infectious Diseases 9:91.

    https://doi.org/10.1186/1471-2334-9-91

    • Google Scholar
41. 1. Storla DG
    2. Yimer S
    3. Bjune GA

    (2008) A systematic review of delay in the diagnosis and treatment of tuberculosis

    BMC Public Health 8:15.

    https://doi.org/10.1186/1471-2458-8-15

    • Google Scholar
42. 1. Suchindran S
    2. Brouwer ES
    3. Van Rie A

    (2009) Is HIV infection a risk factor for multi-drug resistant tuberculosis? A systematic review

    PLoS One 4:e5561.

    https://doi.org/10.1371/journal.pone.0005561

    • Google Scholar
43. 1. Waisman JL
    2. Palmero DJ
    3. Alberti FA
    4. Güemes Gurtubay JL
    5. Francos JL
    6. Negroni R

    (2001)

    Improved prognosis in HIV/AIDS related multi-drug resistant tuberculosis patients treated with highly active antiretroviral therapy

    Medicina 61:810–814.

    • Google Scholar
44. 1. Wells CD
    2. Cegielski JP
    3. Nelson LJ
    4. Laserson KF
    5. Holtz TH
    6. Finlay A
    7. Castro KG
    8. Weyer K

    (2007) HIV Infection and multidrug‐resistant tuberculosis—The perfect storm

    Journal of Infectious Diseases 196:S86–S107.

    https://doi.org/10.1086/518665

    • Google Scholar
45. 1. Whittles LK
    2. Didelot X

    (2016) Epidemiological analysis of the eyam plague outbreak of 1665–1666

    Proceedings of the Royal Society  283:20160618.

    https://doi.org/10.1098/rspb.2016.0618

    • Google Scholar
46. 1. Worby CJ
    2. Lipsitch M
    3. Hanage WP

    (2014) Within-host bacterial diversity hinders accurate reconstruction of transmission networks from genomic distance data

    PLoS Computational Biology 10:e1003549.

    https://doi.org/10.1371/journal.pcbi.1003549

    • Google Scholar
47. 1. World Health Organization

    (2013)

    Global report: UNAIDS report on the global AIDS epidemic

    • Google Scholar
48. 1. World Health Organization

    (2015)

    Global Tuberculosis Report

    • Google Scholar
49. 1. Yang Z

    (2007) PAML 4: Phylogenetic analysis by maximum likelihood

    Molecular Biology and Evolution 24:1586–1591.

    https://doi.org/10.1093/molbev/msm088

    • Google Scholar
50. 1. Zerbini E
    2. Chirico MC
    3. Salvadores B
    4. Amigot B
    5. Estrada S
    6. Algorry G

    (2008)

    Delay in tuberculosis diagnosis and treatment in four provinces of Argentina

    The International Journal of Tuberculosis and Lung Disease 12:63–68.

    • Google Scholar

Thank you for submitting your article "Impact of HIV co-infection on the evolution and transmission of multidrug-resistant tuberculosis" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Prabhat Jha as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

The reviewers and Reviewing Editor were satisfied with the general direction of the analysis and found the work to be sophisticated and to hold an important public health message. However, there were a number of issues around the model and its parameterisation that we felt were inadequate, and which would need to be addressed for the manuscript to be considered further.

Essential revisions:

1) In the first paragraph of the subsection “Effect of HIV co-infection on TB transmission”, you appear to describe a new method for inferring transmission pairs (i.e. person A infected person B), which you have developed. You use a simple SEIR transmission model in determining who infected whom and use an SEIR simulation to help validate the method.

It is important that you further explain this method so that readers can better understand it and researchers will be better able to apply it judiciously in their own work. For example, you conducted a sensitivity analysis on the values used for the latent period and the removal rate but you did not include HIV status in the model, and HIV status could affect each of these parameters. Furthermore, you do not state what mixing assumption you used and likely assumed random mixing. It would be useful if you could explore how the method would perform in the context of non-random mixing (e.g. discuss what the expected effects would be and not necessarily conduct new simulations). You also reported that keeping two-thirds of cases in the second half of the simulated outbreak corresponded to the sampling frame used for the actual genomic data analyses but you need to explain where this number came from (subsection “Simulations and Sensitivity analyses”). If necessary, you should also test the impact of different sampling fractions.

You state "These results conform with our expectation given that there is significant uncertainty about who infected whom based on genomic data alone when accounting for extended periods of within-host evolution" but this high degree of uncertainty does not appear in the conclusions in the main text.

2) In the third paragraph of the subsection “Effect of HIV co-infection on TB transmission”, you tested whether HIV-positive or HIV-negative individuals transmit TB more often conditional on having TB disease. As you suggest, the analysis including all transmissions seems like it would not be very informative since it likely includes many pairs who did not actually transmit between them. While the top X% analyses are affected by the same issue to a lesser extent, there could be another bias in these analyses since enriching for actual transmission pairs would seem to condition on there being at least 1 transmission from the putative source case and thereby eliminates (many of) the instances in which no TB transmission occurred. This would skew the distribution of onward transmissions which could make HIV-positives and negatives appear more similar in terms of their transmission potential than is actually the case.

To provide a simple numerical example of the potential issue, suppose that out of 100 HIV-negative people with active TB 60 of them transmit to 1 other person, while 80 out of 100 TB-HIV coinfected people each transmit TB to 1 person. In this scenario, TB would appear to have the same transmissibility among those who transmitted (i.e. 1 TB transmission per TB transmitter in each HIV group). However, this would give a misleading picture of the actual TB transmission potential of HIV-positive vs. HIV-negative individuals since coinfected people had an 80% probability of transmitting whereas HIV-negative people had a 60% chance of transmitting.

3) In the last paragraph of the subsection “Effect of HIV on progression of Mtb infection to active TB”, you correlate cross-sectional data on HIV and TB prevalences and HIV and MDR-TB prevalences across countries in an effort to determine whether HIV increases TB in general or if it specifically drives the spread of MDR-TB. However, you should explain why this analysis is appropriate in light of the findings of Sergeev et al. (Sci Trans Med 2013, reference 20) who find that the "individual-level association between HIV and drug-resistant forms of TB is dynamic, and therefore, cross-sectional studies that do not report a positive individual-level association will not provide assurance that HIV does not exacerbate the burden of resistant TB in the community."

Also, while the analysis includes HIV and TB/MDR-TB, it's not clear on how it specifically fits in with the rest of paper or if it could be cut without loss of any key information.

You report in Figure 5, the correlation between global patterns of MDR-TB, TB and HIV prevalence. In the Methods you refer to the World Health Organization for the underlying data. This reference should be extended as it now only states "World Health Organization". Data from several countries are known to have limitations in the sense that the prevalence of HIV or TB may not always been known or collected in a rigorous manner. The prevalence of a particular disease can also vary within a single country. Use of these data can therefore also result in biased estimates, as estimates for HIV and TB may be derived from different sites in a single country.

This entire analysis needs to be either strengthened or simply removed from the paper.

4) More explanation is required for key numbers reported in their paper.

In particular, in the Results, you state: "Based on the available data we considered that the sequenced outbreak isolates represented about 35% of the total number of individuals belonging to the outbreak." You should explain how you came up with the figure of 35%. What are you assuming about the fraction of TB cases that are diagnosed and recorded? How, and over what time scale, are you defining the "outbreak" since reactivation TB cases may have been infected decades prior to disease onset?

Again, in the Results, you state: "This left 13 resistance mutations that evolved with high probability during therapy in 11 patients (Table 2). Seven of the patients were HIV negative and four were positive. A χ2 analysis revealed a statistically significant overrepresentation of acquired resistance in HIV-negative cases relative to positives (p = 0.027). This finding strongly suggests that HIV was not a driver of Mtb drug resistance within the outbreak." Given that you indicate that these results strongly back up a key part of your findings, it is important that more information be provided. Specifically, it is unclear exactly which numbers were being compared in the test (e.g. was the null hypothesis that there should be a 50:50 split between HIV-positives and negatives, or was it directly comparing the fraction of mutations found in HIV-infected vs. uninfected individuals [though this would not seem to give the p-value reported]). Also, a Fisher's exact test, not a chi-squared test, would likely be required given that apparently some of the cell sizes were very small.

Essential revisions:

1) In the first paragraph of the subsection “Effect of HIV co-infection on TB transmission”, you appear to describe a new method for inferring transmission pairs (i.e. person A infected person B), which you have developed. You use a simple SEIR transmission model in determining who infected whom and use an SEIR simulation to help validate the method.

It is important that you further explain this method so that readers can better understand it and researchers will be better able to apply it judiciously in their own work. For example, you conducted a sensitivity analysis on the values used for the latent period and the removal rate but you did not include HIV status in the model, and HIV status could affect each of these parameters. Furthermore, you do not state what mixing assumption you used and likely assumed random mixing. It would be useful if you could explore how the method would perform in the context of non-random mixing (e.g. discuss what the expected effects would be and not necessarily conduct new simulations). You also reported that keeping two-thirds of cases in the second half of the simulated outbreak corresponded to the sampling frame used for the actual genomic data analyses but you need to explain where this number came from (subsection “Simulations and Sensitivity analyses”). If necessary, you should also test the impact of different sampling fractions.

We have expanded our description of the methodology used to estimate transmission probabilities, both in the Results and Methods section. For a given putative infector-infectee pair, the calculation depends only on the dates of detection of the two cases, their relative position on the phylogeny, and whether the infector was smear positive or negative. It does not rely on HIV status, so that the effect of HIV status on transmission can be tested independently. We have clarified this important point in the Results section (subsection “Effect of HIV co-infection on TB transmission”, first paragraph). The underlying epidemiological model is a SEIR model, which assumes random mixing of the population, and we have clarified this in the Methods section (subsection “Reconstruction of transmission chains and assessment of the impact of HIV co-infection”, first paragraph) and now discuss how information on the host population structure could be integrated into our inferential framework. The code used to reconstruct transmission events is now available on GitHub https://github.com/xavierdidelot/TransPairs and the paper now includes a pointer to this website (in the aforementioned paragraph, the code should be available now or within a few days). The two-thirds sampling in the latter half of the outbreak reflects the fact that we know the completeness of sampling to be 63% (143/228) in the years 2003 – 2009 (2009 being the last year covered by the study). We have clarified the estimated sample loss in the early and late stages of sampling in the last paragraph of the subsection “Impact of HIV co-infection on Mtb mutation rates and resistance development”.

You state "These results conform with our expectation given that there is significant uncertainty about who infected whom based on genomic data alone when accounting for extended periods of within-host evolution" but this high degree of uncertainty does not appear in the conclusions in the main text.

We agree that this uncertainty could have been acknowledged more clearly in our conclusions. We have now added some reservations regarding the accuracy of the reconstruction of individual transmission events in the Discussion section, fifth paragraph.

2) In the third paragraph of the subsection “Effect of HIV co-infection on TB transmission”, you tested whether HIV-positive or HIV-negative individuals transmit TB more often conditional on having TB disease. As you suggest, the analysis including all transmissions seems like it would not be very informative since it likely includes many pairs who did not actually transmit between them. While the top X% analyses are affected by the same issue to a lesser extent, there could be another bias in these analyses since enriching for actual transmission pairs would seem to condition on there being at least 1 transmission from the putative source case and thereby eliminates (many of) the instances in which no TB transmission occurred. This would skew the distribution of onward transmissions which could make HIV-positives and negatives appear more similar in terms of their transmission potential than is actually the case.

To provide a simple numerical example of the potential issue, suppose that out of 100 HIV-negative people with active TB 60 of them transmit to 1 other person, while 80 out of 100 TB-HIV coinfected people each transmit TB to 1 person. In this scenario, TB would appear to have the same transmissibility among those who transmitted (i.e. 1 TB transmission per TB transmitter in each HIV group). However, this would give a misleading picture of the actual TB transmission potential of HIV-positive vs. HIV-negative individuals since coinfected people had an 80% probability of transmitting whereas HIV-negative people had a 60% chance of transmitting.

We appreciate the reviewers’ thorough assessment of our transmission reconstruction analyses. We agree that the distribution of transmission per potential transmitter is indeed worth a closer examination than we presented in the original submission. That said, we believe that the potential bias identified by the reviewers as illustrated by their numerical example is based on a misunderstanding of how these analyses were actually conducted. Based on the general depth and quality of the reviewers’ comments, this suggests to us we should explain our methodology better, as pointed out by the reviewers under point 1.

In the original submission we did not quantify the number of transmitters, but the number of reconstructed transmission events. We will illustrate this point with the same numerical example: 60% of HIV- and 80% of HIV+ transmitting to one secondary case and assuming we can perfectly reconstruct all transmissions. If we had counted the number of transmissions per person who have transmitted, we would indeed find 1 for both HIV+ and HIV-. However, what we counted were the number of transmissions and for the same numerical example we would thus have observed 4/3 more transmissions from HIV+ relative to HIV-.

This being said, we recognize that skewed distributions of transmissions, e.g. in the form of a limited number of transmitters responsible for a large portion of transmissions within one of the groups could indeed mask differences in transmissibility. We therefore extended our analyses to investigate the distribution of transmissions in more detail. For each subset of transmissions (all and top-X%) we counted the number of transmissions per transmitter (0, 1, 2, … n) and used non-parametric Kruskal-Wallis tests to assess whether the distributions were significantly different. When all transmission events were included, HIV-positive transmitters tended towards having transmitted more often than HIV-negatives, but the difference was not significant. After application of the various top-X% cut-offs the two groups became even more similar (Table 4). This suggests that even when taking into account the distribution of transmissions per transmitter, there is no major impact of HIV co-infection on transmissibility.

The extended results are included in Table 4 and discussed in the fourth paragraph of the subsection “Effect of HIV co-infection on TB transmission” in the revised manuscript.

3) In the last paragraph of the subsection “Effect of HIV on progression of Mtb infection to active TB”, you correlate cross-sectional data on HIV and TB prevalences and HIV and MDR-TB prevalences across countries in an effort to determine whether HIV increases TB in general or if it specifically drives the spread of MDR-TB. However, you should explain why this analysis is appropriate in light of the findings of Sergeev et al. (Sci Trans Med 2013, reference 20) who find that the "individual-level association between HIV and drug-resistant forms of TB is dynamic, and therefore, cross-sectional studies that do not report a positive individual-level association will not provide assurance that HIV does not exacerbate the burden of resistant TB in the community."

Also, while the analysis includes HIV and TB/MDR-TB, it's not clear on how it specifically fits in with the rest of paper or if it could be cut without loss of any key information.

You report in Figure 5, the correlation between global patterns of MDR-TB, TB and HIV prevalence. In the Methods you refer to the World Health Organization for the underlying data. This reference should be extended as it now only states "World Health Organization". Data from several countries are known to have limitations in the sense that the prevalence of HIV or TB may not always been known or collected in a rigorous manner. The prevalence of a particular disease can also vary within a single country. Use of these data can therefore also result in biased estimates, as estimates for HIV and TB may be derived from different sites in a single country.

This entire analysis needs to be either strengthened or simply removed from the paper.

We appreciate our conclusion may seem in variance with the sentence copied from the Abstract of the paper by Sergeev et al. However, we feel it is important to put their statement in its context. The Sergeev et al. paper deals with the dynamics of drug resistant TB (DRTB) over the course of the HIV pandemic in a single setting (Zimbabwe). They focus on a quantity defined as the “ratio of the proportion of TB cases that are drug resistant among HIV-seropositive individuals to the proportion of TB cases that are drug resistant among HIV-seronegative individuals”.

The paper highlights the interesting behavior over time of this ratio, but this is not a quantity we consider. The paper does not explore the effect of different rates of incidence/prevalence of HIV and does not make predictions for other settings than the situation in Zimbabwe they model. The sentence from their Abstract makes an interesting point about the behavior of this ratio in their simulations and states that depending if it is recorded early or late in the same epidemic, it can take different values. However, we do not think the statement can be expanded to warn against cross-comparisons devoid of individual HIV/TB status over multiple settings all recorded at the same time, late in the HIV pandemic (in our case 2013).

The only figure from the Sergeev paper that can be directly compared to the quantities we consider in our correlations is their Figure 3B, which reports the simulated incidence of TB and DRTB in the presence or absence of HIV. Interestingly the simulated patterns they report are actually in line with our global correlations. Except in the very early stages of the epidemic (before ~1990), populations experiencing an HIV epidemic also suffer a higher burden of TB (our panel 5B), a higher burden of MDR-TB (our panel 5C), but the ratio of MDR-TB/ TB remains constant (our panel 5D).

The Sergeev paper also makes a couple of seemingly reasonable assumptions that are likely important determinants of the modelled individual-level rates of drug resistant TB by HIV status as well as overall resistant/susceptible ratio over time. Their model assumes a lower fitness threshold in HIV+ individuals, which necessarily leads to resistant strains (that have lower fitness) being more likely to cause disease in HIV positives (although the model includes compensatory mechanisms that reduce the fitness cost over time in resistant bacteria).

Less important for individual-level relationships but probably an important shaper of the modelled overall rates of drug resistant TB is the parametrization of transmissibility in the model, which assumes that HIV+ TB patients transmit TB less efficiently. It is difficult for us to assess this in detail, but the combination of these two assumptions (1. HIV+ patients are more prone to become infected and contagious with drug resistant TB. 2. HIV+ transmit TB less efficiently than HIV-), that are expected to have opposing effects on the modelled association between HIV prevalence and DR-TB/TB ratio, might explain why the overall MDR/DR/DS ratio is constant irrespective of the presence of HIV (Figure 3B in the Sergeev paper).

However, as our goal was to assess the effect of HIV on transmission, resistance and latency using real data, we did not parametrize HIV+ and HIV- people separately, as this would defeat the point of our analyses. There is however (surprisingly) little support in our results for different fitness thresholds depending on HIV status (both resistance load and estimated Mtb mutation rates are identical between HIV classes) and also little support for a negative association between HIV co-infection and Mtb transmissibility.

The strength of WHO epidemiological data is their availability for most of the world. Even if country-wide summaries can mask some heterogeneities at a smaller scale, we believe the high level of replication (large number of countries) allows capturing robust qualitative global patterns. However, we recognize that there are differences in methodologies between countries and that data from low income countries are likely to be less reliable (which could lead to biases).

To ensure the robustness of the correlations we report, we have now tested whether GDP per capita or public health expenditure per capita might color our results. Limiting the analyses to countries with the highest GDP or health expenditure did indeed increase the correlations between HIV prevalence, TB prevalence and MDR_TB prevalence but did never lead to a significant correlation between HIV prevalence and the ratio of MDR-TB over susceptible TB prevalence (new Figure 5 and new Figure 5—figure supplement 1).

We appreciate these correlations are not directly linked to the rest of the analyses in the paper. In order to confirm our findings in multiple settings, it will be important to replicate our analyses to other outbreaks or epidemics. However, we feel that the inclusion of Figure 5 offers some support for our findings being general throughout different settings. That said, we recognize that the best place for opening up the debate on the generality of our results is probably not the Results section, and we have now moved the figure and the associated text to the Discussion.

Finally, we agree that the rationale, justification and methodological details were inappropriately detailed in the previous draft and wish to apologize for this oversight. We are now far more explicit on the origin of the data and the analysis and have attached the source data to Figure 5 (Figure 5—source data 1). The changes to the manuscript are listed below”

• Inclusion of a fourth panel in Figure 5, showing the correlation between TB prevalence and MDR-TB prevalence;

• Figure 5 moved to Discussion and a re-write of the section in the text pertaining to Figure 5 (Discussion, fifth paragraph);

• Description of the data and analyses in Methods (subsection “Collection of data on global HIV, TB and MDR-TB burden”);

• Inclusion of a supplementary figure with the same correlations than in Figure 5 but limited to the countries with top 50% GDP per capita.

4) More explanation is required for key numbers reported in their paper.

In particular, in the Results, you state: "Based on the available data we considered that the sequenced outbreak isolates represented about 35% of the total number of individuals belonging to the outbreak." You should explain how you came up with the figure of 35%. What are you assuming about the fraction of TB cases that are diagnosed and recorded? How, and over what time scale, are you defining the "outbreak" since reactivation TB cases may have been infected decades prior to disease onset?

We have added an explanation for our estimates of sampling density in the second paragraph of the subsection “Impact of HIV co-infection on Mtb mutation rates and resistance development”. We now refer to one third rather than 35% which is actually more accurate (252 sequenced genomes out of an estimated total of 778). The outbreak start is assumed to be 1992, when the outbreak was detected (Ritacco et al. J infect Dis 1997). We have however shown that the MRCA of the large six-drug resistant clade of the outbreak had evolved already by around 1979, but the strain seems to have circulated in very small numbers prior to 1992, when it caused an outbreak at the Muniz hospital in Buenos Aires (Eldholm et al. 2015 Nat Commun).

Again, in the Results, you state: "This left 13 resistance mutations that evolved with high probability during therapy in 11 patients (Table 2). Seven of the patients were HIV negative and four were positive. A χ2 analysis revealed a statistically significant overrepresentation of acquired resistance in HIV-negative cases relative to positives (p = 0.027). This finding strongly suggests that HIV was not a driver of Mtb drug resistance within the outbreak." Given that you indicate that these results strongly back up a key part of your findings, it is important that more information be provided. Specifically, it is unclear exactly which numbers were being compared in the test (e.g. was the null hypothesis that there should be a 50:50 split between HIV-positives and negatives, or was it directly comparing the fraction of mutations found in HIV-infected vs. uninfected individuals [though this would not seem to give the p-value reported]). Also, a Fisher's exact test, not a chi-squared test, would likely be required given that apparently some of the cell sizes were very small.

We would like to thank the reviewer for identifying this rather weak piece of statistics. The expected numbers are based on the fraction of HIV positives and negatives in the sampled outbreak. We have now redone the statistics using the Fisher’s exact test, which indeed shows that the difference is not significant:

We would however like to stress that we were aware that the numbers are too small to draw strong conclusions and that we originally stated (as you acknowledge above) that these results strongly suggested that HIV co-infection was not a driver of Mtb resistance, rather than the much stronger alternative, namely that HIV co-infection could somehow be said to be “protective” against resistance development. We still feel that the results back up a key finding, as the distributions do not deviate from the 0-hypothesis that there is no difference between the two groups, i.e. HIV is not a direct driver of resistance emergence in the study setting. We have updated the subsection “Impact of HIV co-infection on Mtb mutation rates and resistance development” (last paragraph) accordingly and moderated our statement.

Author details

1. Vegard Eldholm

   Division of Infectious Disease Control, Norwegian Institute of Public Health, Oslo, Norway

   Contribution

   VE, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   Contributed equally with

   Adrien Rieux

   For correspondence

   v.eldholm@gmail.com

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0001-6721-3375

2. Adrien Rieux

   UCL Genetics Institute, University College London, London, United Kingdom

   Present address

   CIRAD, UMR PVBMT, La Réunion, France

   Contribution

   AR, Analysis and interpretation of data, Drafting or revising the article

   Contributed equally with

   Vegard Eldholm

   Competing interests

   The authors declare that no competing interests exist.

3. Johana Monteserin

   1. Instituto Nacional de Enfermedades Infecciosas, ANLIS Carlos Malbrán, Buenos Aires, Argentina
   2. Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Buenos Aires, Argentina

   Contribution

   JM, Acquisition of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

4. Julia Montana Lopez

   UCL Genetics Institute, University College London, London, United Kingdom

   Contribution

   JML, Acquisition of data, Analysis and interpretation of data

   Competing interests

   The authors declare that no competing interests exist.

5. Domingo Palmero

   División Tisioneumonología, Hospital Muñiz, Buenos Aires, Argentina

   Contribution

   DP, Acquisition of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

6. Beatriz Lopez

   1. Instituto Nacional de Enfermedades Infecciosas, ANLIS Carlos Malbrán, Buenos Aires, Argentina
   2. Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Buenos Aires, Argentina

   Contribution

   BL, Acquisition of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

7. Viviana Ritacco

   1. Instituto Nacional de Enfermedades Infecciosas, ANLIS Carlos Malbrán, Buenos Aires, Argentina
   2. Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Buenos Aires, Argentina

   Contribution

   VR, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

8. Xavier Didelot

   Department of Infectious Disease Epidemiology, Imperial College London, London, United Kingdom

   Contribution

   XD, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   xavier.didelot@gmail.com

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0003-1885-500X

9. Francois Balloux

   UCL Genetics Institute, University College London, London, United Kingdom

   Contribution

   FB, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   f.balloux@ucl.ac.uk

   Competing interests

   The authors declare that no competing interests exist.

• 3,041

  Page views

• 560

  Downloads

• 36

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.