High dose aerosol challenge with Mycobacterium tuberculosis fails to overcome BCG vaccination-induced protection in cynomolgus macaques of Chinese origin: implications of natural resistance for TB vaccine evaluation.

This study describes the use of cynomolgus macaques of Chinese origin (CCM) to evaluate the ecacy and immunogenicity of the BCG vaccine against high dose aerosol Mycobacterium tuberculosis challenge. Progressive disease developed in three of the unvaccinated animals within ten weeks of challenge, whereas all six vaccinated animals controlled disease for 26 weeks. Three unvaccinated animals limited disease progression, highlighting the intrinsic ability of this macaque species to control disease in comparison to macaques of other species and genotypes. Low levels of IFNγ were induced by BCG vaccination in CCM suggesting that IFNγ alone does not provide a suciently sensitive biomarker of vaccination in this model. An early response after challenge, together with the natural bias towards terminal effector memory T-cell populations and the contribution of monocytes appears to enhance the ability of CCM to naturally control infection. The high dose aerosol challenge model of CCM has value for examination of the host immune system to characterise control of infection which would inuence future vaccine design. Although it may not be the preferred platform for the assessment of prophylactic vaccine candidates, the model could be well suited for testing post-exposure vaccination strategies and drug evaluation studies.


Introduction
Tuberculosis is responsible for 1.2 million deaths each year, with ten million new infections by the causative agent Mycobacterium tuberculosis (M. tb) annually 1 . Vaccination is widely accepted as the most effective method for control of infectious disease, and although the only licenced TB vaccine Bacille Calmette-Guérin (BCG) protects children from developing severe TB disease 2 , the protection afforded to adults is limited 3 , and BCG is unsuitable for use in people whose immune system is compromised. Improved vaccines against TB are desperately needed. Until a surrogate marker able to predict the potential e cacy of new TB vaccine candidates is identi ed, the slow, logistically complex Animal models are valuable in the development of new vaccines, as studies of infectious challenge can be used to predict the effectiveness of vaccines in humans and provide the opportunity to identify and validate correlates of protection. Non-human primates (NHP), with their close similarity to humans, in terms of immunology and physiology provide the most relevant models of human tuberculosis [4][5][6] . The potential of the cynomolgus macaque species as a model for human tuberculosis following administration of M. tb via the trachea was rst described by Walsh et al 7 , and subsequent studies have demonstrated the propensity of the species to develop a spectrum of outcomes from active disease to chronic and latent infection 8,9 . Consequently, when using cynomolgus macaques, the number of animals required for TB vaccine evaluation experiments must be su cient to ensure statistically robust results can be obtained in light of this inherent variability in disease control, and animals must be monitored for an extended period for the outcome of infection to be determined 9 . For these reasons, the rhesus macaque is commonly used as a stringent, but more practical model for TB vaccine evaluation due to the uniform and relatively rapid disease progression that can be observed in this species [10][11][12][13] . However, we hypothesised that a uniform pro le of disease progression may be achievable in cynomolgus macaques of Chinese genotype by application of a high dose aerosol M. tb challenge to ensure widespread distribution of bacteria within the lung 14 , thus providing an alternate primate model for evaluation of prophylactic TB vaccination strategies.
Validated correlates of protection against M.tb infection are yet to be identi ed, although both a cell mediated T-helper 1 (Th1) response from CD4 T-cells 15 and a MHC-I restricted CD8 T-cell response 16 are known to be important for successful control of disease. Until a true correlate of protection is available, functional markers such as IFN-g expression are measured to assess vaccine immunogenicity. There is evidence that the quality of the T-cell response is important to the induction of T-cell memory 17,18 and multifunctional CD4 T-cells expressing combinations of the cytokines IFN-g, TNF-α and IL-2 are involved in the active phase of disease [19][20][21] and provide potential correlates of vaccine-induced protection [22][23][24][25] although not all reports support this latter role 26,27 .
This study aimed to establish a high dose aerosol challenge model of TB in cynomolgus macaques of Chinese origin (CCM), in order to evaluate the e cacy afforded by BCG and to elucidate correlates of protective immunity.

Experimental animals
Twelve three-year-old, male cynomolgus macaques (Macaca fascicularis) of Chinese genotype were obtained from a Home O ce approved breeding colony. Absence of previous exposure to mycobacterial antigens was con rmed by a tuberculin skin test whilst in their original breeding colony and screening using an ex-vivo IFN-γ ELISPOT (MabTech, Nacka. Sweden) to measure responses to PPD (SSI, Copenhagen, Denmark), and pooled 15-mer peptides of ESAT6 and CFP10 (Peptide Protein Research LTD, Fareham, U.K.) immediately prior to study entry as described elsewhere 13 .
Throughout the study animals were housed in compatible social groups, in accordance with the Home O ce (UK) Code of Practice for the Housing and Care of Animals Used in Scienti c Procedures (1989), and the National Committee for Re nement, Reduction and Replacement (NC3Rs), Guidelines on Primate Accommodation, Care and Use, August 2006 (NC3Rs, 2006)). Animals were sedated by intramuscular (IM) injection with ketamine hydrochloride (Ketaset, 100mg/ml, Fort Dodge Animal Health Ltd, Southampton, UK; 10mg/kg) to allow procedures requiring removal from their housing as described elsewhere 13,28 . None of the animals had been used previously for experimental procedures and each socially compatible group was randomly assigned to a study treatment. All animal procedures and study design were approved by the Public Health England, Porton Down Establishment Animal Welfare and Ethical Review Body, and authorized under an appropriate UK Home O ce project license and as in previous studies 13,28,29 . Study was carried out in compliance with the ARRIVE guidelines.

Vaccination
Six macaques (Group A) were immunised intradermally (ID) in the upper left arm with 100 μl BCG vaccine, Danish strain 1331 (SSI, Copenhagen, Denmark). The method is as described elsewhere 13,29 but in brief; BCG vaccine was prepared for ID administration according to manufacturer's instructions for preparation of vaccine for administration to human adults, by addition of 1 ml Sautons diluent to a vial of vaccine, to give a suspension of BCG at an estimated concentration of 2 x 10 6 to 8 x 10 6 CFU/ml. All vaccinations were administered within one hour of vaccine reconstitution. The viability of the BCG vaccine was con rmed to be within the expected range for the batch. Vaccination sites were monitored and assessed for local reactions after vaccination with BCG.

M. tb challenge strain
The stocks of the M. tb Erdman strain K 01 (BEI Resources) used for challenge were provided as frozen suspensions at a stated titre of 3x10 8 colony forming units (CFU)/ml when grown on Middlebrook 7H11 OADC selective agar as described previously . On the day of challenge, six vials were thawed, pooled and diluted appropriately, in sterile distilled water.

Aerosol exposure
Twenty-one weeks after vaccination with BCG the six immunised animals (Group A) together with six unvaccinated animals (Group B) were challenged by the aerosol route with M. tb. The methodology and apparatus used to deliver M. tb via the aerosol route was as previously described 31 . In brief, monodispersed bacteria in particles were generated using a 3-jet Collison nebuliser (BGI) and, in conjunction with a modi ed Henderson apparatus 32 , delivered to the nares of each sedated primate via a modi ed veterinary anaesthesia mask. Challenge was performed on sedated animals placed within a 'head-out', plethysmography chamber (Buxco, Wilmington, North Carolina, USA) to enable the aerosol to be delivered simultaneously with the measurement of respired volume. The aerosol delivery process was calculated to result in the deposition of an estimated target dose of 1000 CFU in the lungs. The calculations to derive the presented dose (PD) and the retained dose (the number of organisms assumed to be retained in the lung) have been described previously for high/medium aerosol doses 33,34 . The challenge was conducted such that, one animal from each group was exposed in sequence with the cycle repeated until all animals were exposed.

Clinical procedures
Animals were monitored daily for behavioural and clinical changes. Behaviour was evaluated for contraindicators including depression, withdrawal from the group, aggression, changes in feeding patterns, breathing pattern, respiration rate and cough. Prior to blood sample collection, vaccination, aerosol challenge and euthanasia, animals were weighed, examined for gross abnormalities and body temperature measured. As in previous studies 29 , red blood cell haemoglobin (RBC[Hb] levels were measured using a HaemaCue haemoglobinometer (Haemacue Ltd, Dron eld, UK) to identify the presence of anaemia, and erythrocyte sedimentation rates (ESR) were measured using the Sediplast system (Guest Medical, Edenbridge, UK) to detect and monitor in ammation induced by infection with M. tb. The time of necropsy, if prior to the end of the planned study period, was determined by experienced primatology staff and based on a combination of the following adverse indicators: depressed or withdrawn behaviour, abnormal respiration (dyspnoea), loss of 20% of peak post-challenge weight, ESR levels elevated above normal (>20 mm), haemoglobin level below normal limits (< 100g/dL), increased temperature (>41 o C) and severely abnormal chest X-ray.

Peripheral blood mononuclear cells (PBMC) preparation
Peripheral blood mononuclear cells (PBMC) were isolated from heparin anti-coagulated blood using standard methods 35 .

Memory T-cell intracellular cytokine staining assay
Antigen speci c memory T-cells secreting multiple cytokines were detected using PPD stimulated PBMCs using the method previously described (Sharpe et al., 2016).

Interferon-gamma (IFN-g) ELISpot
An IFN-g ELISpot assay was used to estimate the numbers and IFN-g production capacity of mycobacteria-speci c T cells in PBMCs using a human/simian IFN-g kit (MabTech, Nacka. Sweden), as described previously 31 . PBMCs were cultured with 10 mg/ml PPD (SSI, Copenhagen, Denmark), or a pool containing overlapping 15mer peptides spanning ESAT6 (Peptide Protein Research Ltd, Wickham, UK), or without antigen, in triplicate, and incubated for 18 hours. Phorbol 12-myristate (Sigma-Aldrich Dorset, UK) (100 ng/ml) and ionomycin (CN Biosciences, Nottingham, UK) (1 mg/ml) were used as a positive control. After culture, spots were developed according to the manufacturer's instructions. Spot forming units were counted and average spot areas measured using AID CADAMA ELISPOT reader and software (CADAMA, Stourbridge, UK). Determinations from triplicate tests were averaged. Data were analysed by subtracting the mean number of spots in the wells with cells in medium alone from the mean counts of spots in wells with cells combined with PPD or peptide pools.
2.10 Quanti cation of secreted IFN-γ IFN-γ production was measured in serum samples prepared from clotted blood using standard methods and supernatants collected from cultures of M. tb antigen stimulated whole blood 31 .Heparinised blood was diluted 1 in 10 with serum-free medium (RPMI supplemented with L-glutamine, penicillin and streptomycin (All Sigma-Aldrich, UK)) and cultured with either puri ed protein derivative from M. tb (PPD; 5 μg/ml, SSI, Copenhagen, Denmark), BCG (Danish strain 1331, 2 x 10 5 CFU, SSI, Copenhagen, Denmark), or mitogen Phytohaemaglutinin (PHA; 10μg/ml, Sigma-Aldrich, Dorset, UK), or in medium alone for 6 days. Supernatants were harvested at day 6 and stored at -80 o C.
The quantity of IFN-γ in the supernatants and serum was estimated using a commercially available human/monkey IFN-g ELISA kit (MabTech, Nacka. Sweden). A puri ed human IFN-γ was used for the standard curve on each plate that was included with the kit. The ELISA was developed using streptavidin-HRP and 3,3',5,5'-tetramethylbenzidine (TMB) liquid substrate system (Sigma-Aldrich, Dorset, UK) and the reaction stopped with 2M sulphuric acid (May & Baker Ltd, Dagenham, UK). Well absorbances at 450 nm were determined using a Multiskan reader (ThermoScienti c, USA). A standard curve was plotted for each plate and used to calculate the concentrations of IFN-g in each sample using Ascent Software (ThermoScienti c, USA) to calculate the reciprocal from the standard curve. Lower limits of detection were determined on a plate by plate basis based on the standard curve. The media only wells were subtracted from the PPD or PHA stimulated sample wells.

Whole blood intracellular cytokine staining
Immediately after collection, 500 μl undiluted, sodium heparin anticoagulated whole blood was added to tubes containing a 0.25 mg/ml solution of anti-CD28 and anti-CD49d (both from BD Biosciences, Oxford, UK), and either PPD (20 μg/ml, SSI, Copenhagen, Denmark), or nothing (negative control), or SEB (5 mg/ml Sigma Aldrich, Gillingham, UK) (positive control) and were transferred to a 37 o C water bath. After six hours, 10 μl Brefeldin-A (10 μg/ml, Sigma Aldrich, UK) was added and the waterbath programmed to switch off after a further 5 hours. The red blood cells were lysed using 1 ml Pharmlyse (BD Biosciences, Oxford, UK) and cells stored in cryomedia (heat inactivated foetal calf serum (Labtech International, Uck eld, UK) with 10% DMSO (Sigma Aldrich, Gillingham, UK) in nitrogen vapour.
2.12 Flow cytometric acquisition and analysis for WBICS.
Cells were analysed using a 4b SORP LSRII (BD Biosciences, Oxford, UK). Analysis was performed using an established method 18 ; cytokine-secreting T cells were identi ed using a forward scatter-height (FSC-H) versus side scatter-area (SSC-A) dot plot to identify the lymphocyte population, to which appropriate gating strategies were applied to exclude doublet events, non-viable cells, monocytes (CD14 + ) and B cells (CD20 + ) prior to sequential gating through CD3 + , CD8and CD4 + versus IFNγ, and CD3 + , CD8 + and CD4versus IFNγ histograms. All data were analysed using FlowJo (version 9.7.6, Treestar, Ashland, US). Polyfunctional cells were identi ed using Boolean gating combinations of individual cytokine-producing CD4 or CD8 T-cells. The software package PESTLE (version 1.7) was used for background subtraction to obtain antigen-speci c ICS assay responses, and SPICE (version 5.35) was used to generate graphical representations of ow cytometry data (Mario Roederer, Vaccine Research Centre, NIAID, NIH).

Necropsy
Animals were anaesthetised, and clinical data collected. Blood samples were collected prior to euthanasia by intra-cardiac injection of a lethal dose of anaesthetic (Dolelethal, Vétoquinol UK Ltd, 140mg/kg). A post-mortem examination was performed immediately, and gross pathological changes were scored using an established system based on the number and extent of lesions present in the lungs, spleen, liver, kidney and lymph nodes, as described previously 31 . Samples of spleen, liver, kidneys and tracheobronchial, inguinal and axillary lymph nodes were removed and sampled for quantitative bacteriology. The lungs, including the heart and attached tracheobronchial and associated lymph nodes, were removed intact. The lymph nodes were measured and examined for lesions. Following examination, the left bronchus of lung was clamped off using artery forceps, and the left upper and left lower lung lobes dissected away from lung. The upper and lower left lobes were collected for quantitative bacteriology. The left bronchus was tied off with string to the left of the clamp and right-hand side of lung gently infused with 10% neutral buffered formalin (NBF) using a 10 ml syringe attached to a 14 CH Netalon catheter (J.A.K. Marketing, York, UK) whilst observing closely for in ation. The trachea was tied off and the lungs immersed in formalin to complete xation. In addition, samples of kidneys, liver, spleen, and sub clavicular, hepatic inguinal and axillary lymph nodes were xed in 10% NBF.

Lung imaging
Thoracic radiographs (SP VET 3.2, Xograph Ltd) were acquired using mammography lm (Xograph Imaging Systems Ltd, Tetbury, UK) before and every 2 weeks after exposure to M. tb. Evaluation of disease was performed by an experienced consultant thoracic radiologist blinded to the animal group and clinical status, using a pre-determined scoring system based on the amount and distribution of in ltrate 37 . The lung was divided into four zones (left and right sides divided at the anterior aspect of the fth rib) and the disease burden scored in each area according to the system 0 = normal; 1 = < 10% abnormal; 2 = > 10% < 50% abnormal; 3 = >50% <75% abnormal; 4 = >75 % abnormal. The scores for each zone were summed to provide the total X-ray score where the maximum score was 16.

Magnetic resonance imaging
The ex-vivo expanded, xed lungs were set in 2% agarose (Sigma-Aldrich, UK) and images were taken using a 3.0 T 750 MRI Scanner (General Electric Healthcare, Milwaukee, WI, USA) as described previously 31 . This enabled evaluation of the pulmonary disease burden at the end of the study period. Lung lesions were identi ed in MR images from their signal intensity and nodular morphology relative to normal lung parenchyma.

Lesion analysis / quanti cation (Stereology)
Lung lesions were identi ed on MR images based on their signal intensity and nodular morphology relative to normal lung parenchyma. The total lung and lesion volume relative to the xed tissue was determined using the Cavalieri method applied to MRI image stacks, and then expressed as a ratio to provide a measure of disease burden in each animal as previously described 31,37 . Analyses of lesion volume on magnetic resonance (MR) images were performed with the investigators reading the images blind to treatment groups.

Histopathological examination
Representative samples from each lung lobe and other organs, were processed to para n wax, sectioned at 3-5 µm and stained with haematoxylin and eosin (HE). For each lung lobe, tissue slices containing obvious lesions were selected for histological examination. Where gross lesions were not visible, a sample was taken from a pre-de ned anatomical location from each lobe to establish consistency between animals. Sections of lung associated lymph nodes (trachea-bronchial at the bifurcation and cranial and caudal to the bifurcation) and other tissues were evaluated for the presence of tuberculous lesions. Lesions were classi ed according to the scheme used by Lin et al 9 .

Bacteriology
The lung lobes, spleen, kidneys, liver and tracheobronchial lymph nodes were sampled for the presence of viable M. tb post-mortem as described previously 31 . Weighed tissue samples were homogenized in 2 ml (spleen, liver, kidney, hilar lymph nodes) or 10 ml (lung lobes) of sterile water, then either serially diluted in sterile water prior to being plated or plated directly onto Middlebrook 7H11 OADC selective agar. Plates were incubated for 3 weeks at 37ºC and resultant colonies were con rmed as M. tb and counted.

IDEXX Lasercyte Haematology analyser.
The IDEXX Lasercyte analyser (IDEXX, USA) was used according to the manufacturer's instructions using 500 µl of whole blood anticoagulated with EDTA (1.8 mg/ ml of blood), in a Vacutainer (BD Biosciences, USA).

Enzyme-linked immunosorbant assay (ELISA)
ELISAs were performed as previously described ( 38 ). Antigens used were puri ed protein derivative (PPD) (Statens Serum Institut (SSI), Denmark) at a concentration of 5µl/ml and whole BCG SSI at a concentration of 5×10 5 CFU/ml; plates were coated with 50 µl of antigen and incubated at 4°C overnight.
Samples were prepared by diluting test serum and positive/negative control serum 1:100 in casein and 50 µl added per well for 2 hours. Secondary antibody (goat anti-monkey γ-chain-speci c whole IgG alkaline phosphatase conjugate (Rockland Immunochemicals, PA, US) was diluted 1:750 and 50 µl added per well for 1 hour. 100 µl of p-nitrophenyl phosphate (pNPP) development buffer was added to each well and plates were read every 10 minutes using a Model 550 Microplate Reader (Bio-Rad, UK) until the positive control reached a predetermined OD 405 that was consistent across plates. Reported values are the mean OD 405 values of triplicate samples with the mean OD 405 values of triplicate negative controls subtracted.

Mycobacterial growth inhibition assay
The direct PBMC mycobacterial growth inhibition assay (MGIA) was carried out using PBMCs at a concentration of 13x10 6 cells inoculated with ~500 CFU Pasteur Aeras in a total volume of 480 µl RPMI (containing 2mM l-glutamine and 25mM HEPES), plus 120 µl autologous serum matched to animal and time-point in a 48-well plate, as described previously 39 . After incubation at 37°C with 5% CO 2 for 96 hours, co-cultures were transferred to 2ml screw-cap tubes and centrifuged at 15,300g for 10 minutes. During this time, 500 µl sterile water was added to each well to lyse adherent monocytes and release intracellular mycobacteria. Supernatants were removed from the screw-cap tubes by pipetting, and water from the corresponding well added to the remaining pellet. Tubes were pulse vortexed and the suspension was transferred to Mycobacterium growth indicator tubes (MGIT) supplemented with PANTA antibiotics and OADC (Becton Dickinson, UK) for enumeration of surviving mycobacteria using the BACTEC MGIT instrument (Becton Dickinson, UK). On day 0, duplicate direct-to-MGIT inoculum controls were set up by inoculating supplemented BACTEC MGIT tubes with the same amount of mycobacteria as the samples. The time to positivity (TTP) read-out was converted to log10 CFU using stock standard curves of TTP against inoculum volume and CFU. Results were normalised by subtracting log10 CFU of the inoculum control from log10 CFU of each sample, and 'vaccine response' calculated as (post-vaccination normalised growth -baseline normalised growth), presented as Δlog10 CFU.

Statistical analyses
Comparisons of ex-vivo IFNγ ELISpot pro les were completed using the area under the curve (AUC) of each animal's response calculated using Sigmaplot version 10 (Systat Software Inc, Hounslow, UK). AUC values were compared between test groups using the non-parametric Mann-Whitney U test, Minitab version 15 (Minitab Ltd, Coventry, UK).
To compare T-cell functional pro les measured by polyfunctional ow cytometry, between vaccination groups, T-cell subset frequencies were compared using a Wilcoxon-rank test at each analysis time point

Disease progression
In the period after aerosol challenge with M. tb, three (4389, 9623, 1027) of the six animals that did not receive a BCG vaccination showed changes in behaviour and clinical parameters consistent with progression of tuberculosis (decreased body weight, anaemia, raised in ammatory markers, severely pulmonary changes) and were euthanized ahead of the planned end of the study at ve, seven and ten weeks after challenge. By contrast, all six BCG vaccinated animals controlled disease progression during the 26 weeks of study ( Figure 1A). The length of time that animals in each group controlled disease progression was compared using two statistical approaches, the Log Rank Test and Cox regression analysis. Both tests demonstrated trends for improved control of disease progression in the BCG vaccinated group although neither reached statistical signi cance (Log rank test, p = 0.055, Cox regression, p = 0.058).
Trends for body weight to increase over time and body temperature to remain within normal ranges were seen in both the BCG vaccinated and unvaccinated animals that completed the study period. At the end of the study signi cantly more weight loss was recorded in the unvaccinated group than the BCG vaccinated group (p = 0.0130, Figure 1B). The BCG vaccinated group showed (RBC[Hb])levels consistently within the normal range for the species, while the levels in the unvaccinated animals reduced over time and tended to be low at termination in the animals in which disease progressed to meet humane end criteria. Erythrocyte sedimentation rate (ESR) increased in animals as disease progressed to meet humane endpoint criteria but remained within the normal range in all animals that completed the study. At the end of the study, RBC[Hb] levels were signi cantly higher (p = 0.0022, Figure 1C) and ESR signi cantly lower (p = 0.0022, Figure 1D) in the BCG vaccinated group than the unvaccinated group.
Disease burden was evaluated, either at the planned study endpoint, 26 weeks after aerosol exposure, or when disease progressed to meet pre-set, humane, endpoint criteria. At necropsy, all animals were assessed for gross pathological changes. All BCG vaccinated animals were seen to have good body condition and the abdominal organs were free of observable abnormalities. In the thoracic cavity scattered lesions were seen on the lung surfaces of all six animals with enlarged lymph nodes observed in two animals. In contrast, two of the six unvaccinated animals showed thin body condition, and abnormalities were detected in the gut or abdominal region of three of the six animals (for example; enlarged gall bladder or spleen, pale liver colour) and all six animals possessed abnormalities in the thoracic cavity (e.g. adhesions to thorax wall). TB-induced disease burden was higher in animals in which disease met humane endpoint criteria during the study, than in those which controlled disease progression to the end of the study. The total pathology score determined at necropsy, based on the number and size of lesions, revealed the level of disease in all BCG vaccinated animals to be signi cantly lower than in the unvaccinated animals (p = 0.0013, Figure 1E).
The pathology and imaging scores used to assess pulmonary disease revealed differences in outcome between the vaccinated and unvaccinated groups. The lung pathology scores ( Figure 1F), comprising the number of pulmonary lesions determined by serial sectioning and manual counting of discrete and coalesced lesions in the right lung lobes ( Figure 1G), and pulmonary disease burden quanti ed using stereology of MR images to determine the ratio between lesion volume and lung volume in the left lung lobes ( Figure 1H), were signi cantly lower in the BCG vaccinated group (lesion count: p = 0.0346; Lung pathology score: p = 0.0087; lesion:lung volume: p = 0.0411) than in the unvaccinated group. Chest X-ray scores also revealed a non-signi cant trend for a reduced level of pulmonary disease in the BCG vaccinated animals (p = 0.106, Figure 1I).
Macroscopic assessment of the lungs revealed tuberculous lesions in all animals in both study groups with a trend for disease to be more severe in the unvaccinated group (Supplementary data 1), which was con rmed by microscopic assessment. A range of granuloma types with variable presence of mineralisation and coalescing lesions were seen in the lungs of both BCG vaccinated and unvaccinated animals. Unorganised granulomas were observed more frequently in BCG vaccinated animals than in the unvaccinated animals, and coalesced lesions were observed more frequently in the unvaccinated animals particularly those which developed progressive disease during the study.
Evidence of dissemination to the hilar lymph nodes, liver, kidneys and spleen was sought by gross and microscopic pathological examination and by culture of tissue samples for the presence of viable M. tb ( Figure 1J). Tuberculous lesions were observed in the hilar lymph nodes of all study animals and M. tb was isolated from the hilar lymph nodes of four of six animals in each group. Spread to the thoracic walls was seen only in the unvaccinated animals where tuberculosis lesions were grossly visible in four of the six animals. In the BCG vaccinated animals, gross lesions were not observed in any abdominal organs but microscopic examination and bacterial culture, suggested a trend for reduced spread to the spleen (three of six), liver (three of six) and kidneys (three of six) in the BCG vaccinated animals compared to the unvaccinated animals (spleen: ve of six; liver: ve of six; kidneys four of six). ( Figure   1J). Tuberculous lesions were not detected in the brain of animals from either group. Where bacteria were isolated, the bacterial burden in tissues was at a similar level in both study groups. Mycobacteria were not isolated from the urine, faeces, saliva or PBMCs of any of the vaccinated or unvaccinated animals.

The mycobacterium speci c IFNγ response to BCG vaccination and challenge with M. tb
IFNy is an important biomarker for TB, consequently a series of approaches were used to characterize the IFNγ response induced by BCG vaccination and M. tb infection in CCM.
Firstly, application of an ex-vivo ELISPOT assay prior to and after vaccination revealed the frequency of PPD-speci c IFNγ secreting cells increased to levels signi cantly above pre-vaccination levels in BCG vaccinated animals six weeks after vaccination (p=0.031) (Figure 2A) and when the AUC of the IFNγ pro le was compared, a trend for the responses made by the BCG vaccinated group to be higher than those of the unvaccinated group (p=0.065) was seen (Supplementary data 2A). Challenge with M. tb induced an increase in the frequency of PPD-speci c IFNγ secreting cells in both the vaccinated and unvaccinated animals. In the rst four weeks after challenge, a higher PPD-speci c response was observed in the BCG vaccinated group than in the unvaccinated group, although statistical analysis showed the difference in frequency at two (p = 0.132) and AUC up to four weeks after challenge did not reach signi cance (p = 0.18) (Supplementary data 2B and C). A signi cant increase in the number of PPD-speci c IFNγ secreting cells was observed in the BCG group two weeks after challenge compared to pre-challenge numbers (p = 0.0313). Twelve weeks after challenge there was a trend for the frequency of PPD-speci c IFNγ secreting cells to be signi cantly higher in BCG vaccinated animals than in the three unvaccinated animals that controlled disease progression (p = 0.048) (Supplementary data 2D) . Fourteen weeks post-challenge (week 35) there was a transient increase in PPD-speci c IFNγ secreting cells measured in the unvaccinated group, which decreased and then remained stable for the rest of the study.

Page 14/29
The ex-vivo ELISPOT assay was applied to evaluate the response to ESAT6, a TB speci c marker associated with disease burden that is absent in BCG ( Figure 2B). ESAT6-speci c IFNγ-secreting cells were not detected in any of the animals prior to, or after vaccination. Following exposure to M. tb, signi cantly more ESAT6-speci c IFNγ-secreting cells were measured in the unvaccinated animals than in the BCG vaccinated group two weeks after challenge (p = 0.0346); the number of spot forming units (SFU) was very low (median of 34 SFU) ( Figure 2C). The frequency of ESAT6-speci c IFNγ secreting cells increased in the peripheral blood of all BCG vaccinated and unvaccinated animals with the two groups showing a similar magnitude of responses measured at four (Mann Whitney: p = 0.178), and twelve weeks (Mann Whitney: p = 0.167) after challenge and over the 26 weeks post challenge period (AUC, Mann Whitney: p = 0.619).
An ELISA was used to measure f IFNγ secreted in response to stimulation with PPD by cells in whole blood (WB) samples collected following vaccination and challenge, ( Figure 2D). Responses to PPD were detected in ve of the BCG vaccinated animals but this was not signi cantly different to the unvaccinated group (p = 0.0827). Following exposure to M. tb, the PPD-speci c responses induced in both the BCGvaccinated and unvaccinated groups showed a bi-phasic pattern with peak IFNγ secretion detected between weeks four and six and weeks 10 and 12 post challenge with non-signi cant trends for responses to be higher in the unvaccinated group (PPD: p = 0.329).
The Spearman's Rank correlation test was applied to evaluate potential relationships between the IFNγ responses induced by vaccination, or challenge and the TB-induced disease burden evaluated at termination using a range of measures, including gross pathology score, pulmonary lesion counts, lesion to lung volume ratio and changes in clinical parameters. When results from BCG vaccinated and unvaccinated animals were combined and tested, a higher frequency of PPD-speci c IFNγ secreting cells measured two weeks after challenge correlated with lower disease burden in the lungs after challenge (week 2: lung lesion ratio: p = 0.013, lung score: p = 0.049, total score p = 0.0075) (Supplementary Data 3). Signi cant relationships between IFNγ measures and measures of protection were not identi ed within individual test groups.

Polyfunctional and memory T-cell analysis
The frequency of polyfunctional CD4 + and CD8 + T-cells producing IFNγ, TNFα and IL-2 in response to PPD stimulation of whole blood was measured both the immunisation and challenge phases of the study. Low levels of CD4+ T-cells producing IFNγ + and TNFα + or IL-2 + and TNFα + were detected after vaccination, when compared to frequencies measured before vaccination, or in the unvaccinated group ( Figure 3A and B), although these did not reach signi cance.
Following M. tb challenge, increases in the frequency of CD4 T-cell populations producing IFNγ + TNFα + IL-2 + , IFNγ + TNFα + or IL-2 + TNFα + were detected in both vaccinated and unvaccinated animals, with higher proportions of these populations occurring in the unvaccinated group.
Minimal levels of cytokine production were detected in CD8 + T-cells after vaccination, though TNFα producing cells were detected in some of the animals ( Figure 3C and D) independent of vaccination status. Post-challenge, IFNγ + TNFα + double positive cells were detected in both vaccinated and unvaccinated animals, as well as cells producing IFNγ + , TNFα + and IL2 + individually. All cytokines showed trends towards higher levels in the unvaccinated group than in BCG vaccinated animals.
When cytokine responses were summed, a signi cantly higher level of cytokine production was detected in CD4 + T-cells collected four weeks post-BCG vaccination compared to that in the unvaccinated group (p=0.0476) ( Figure 3E). In general, higher levels of cytokine production were observed in CD8 + T-cells from BCG vaccinated animals during the vaccination period ( Figure 3F). Overall, summed cytokine levels measured in CD4 + T-cells were not signi cantly different in vaccinated and unvaccinated groups during the rst eight weeks post-challenge, and for the latter part of the study, higher levels of cytokine production were detected in the unvaccinated animals. An IFNγ ELISA was used to measure the quantity of IFNγ in serum and revealed an increase in the levels of IFNγ measured following vaccination in ve of the six BCG vaccinated animals relative to the unvaccinated group although the difference did not reach statistical signi cance (p = 0.179). This trend reversed following challenge when higher levels of IFNγ were detected in the unvaccinated animals ( Figure 4A).
Using a haematology analyser, populations of neutrophils, monocytes, lymphocytes, eosinophils and basophils were measured throughout the vaccination phase and up to the fourth week after infection (Supplementary data 5). Signi cantly more lymphocytes were detected in the BCG vaccinated group four weeks post-infection (Supplementary data 5). The M:L ratio was calculated and whilst signi cant differences were not seen between the vaccinated and unvaccinated animals during the vaccination phase, four weeks after infection, signi cantly higher M:L ratios were recorded in the unvaccinated group in comparison to the BCG vaccinated animals ( Figure 4B). Furthermore, the M:L ratio calculated at baseline correlated with total pathology score after infection (r = 0.6245, p = 0.033) (Supplementary data 6).
Due to the trend seen in the M:L ratio data, monocytes were examined in more detail at baseline using a ow cytometric immunophenotyping assay (Supplementary data 7). Monocyte subtypes were analysed using CD14 and CD16, to elucidate classical (CD14 + CD16 -), intermediate (CD14 + CD16 + ) and nonclassical (CD14 -CD16 + ) monocytes. When correlated with lung pathology, there was a signi cant positive correlation between classical monocytes and lung pathology, and an inverse trend with the non-classical monocytes ( Figure 4C and D).
The quantity of IgG speci c to either PPD ( Figure 4E), or BCG ( Figure 4F) was evaluated by ELISA in serum collected at baseline and eight weeks after vaccination to determine whether speci c antibodies were induced by BCG vaccination. A signi cant increase in both PPD and BCG speci c-IgG was detected at week eight post-vaccination in the BCG group using a paired Wilcoxon T-test (p=0.0313 for both) with no change in the unvaccinated group.
The direct mycobacterial growth inhibition (MGIA) assay was carried out using PBMCs and matched serum collected at baseline and week eight post-vaccination. Improved control of mycobacterial growth was observed in three of the six BCG vaccinated animals following vaccination, and none of the unvaccinated animals, although the difference between groups was not statistically signi cant ( Figure  4G). There were no signi cant associations between MGIA outcome and measures of protection, although there was a trend towards a correlation between mycobacterial growth at week eight postvaccination and X-ray score (r=0.55, p=0.07).

Vaccine assessment
In this high dose infection model, BCG vaccination afforded a consistent and noteworthy level of protection, with all vaccinated animals showing a signi cantly reduced disease as determined by every measure of pulmonary disease, namely; lesion counts, stereology and total pathology score; overall disease burden (total gross pathology score); disease dissemination (histology, bacterial culture) and changes in clinical parameters (body weight, anaemia (RBC [Hb]) and in ammatory markers (ESR). The strong consistent protective effect conferred by BCG vaccination in CCM against high dose aerosol challenge in the present study is consistent with the protective effect afforded by BCG to similar animals against bronchoscopically delivered high dose M tuberculosis challenge 42 . A feature of aerosol delivered challenge is that the inoculum is more evenly distributed throughout the lungs 14 . Thus, we aimed to use this feature to minimise variation in disease readouts and end-point measures by standardising the disease progression with a post challenge follow up period that is more amenable to vaccine evaluation than might be required following lower dose exposure in this species 9,43 .
Recent studies have used low doses of M. tb for intrabronchial challenge in CCM to provide a dose that is more representative of natural infection and under these circumstances the CCM shows a spectrum of outcomes , from the development of active disease through to the development of latent disease 9 . Whilst this re ects the outcome of M. tb infection in humans, for this approach to be used for the evaluation of vaccine e cacy, test group sizes need to be su cient to discriminate a vaccine effect against the background of inter-animal variability. The remarkable ability of the CCM to naturally control M. tb infection, even after challenge with high titres of M. tb, is highlighted by the outcome observed in unvaccinated animals in the present study, where half successfully controlled disease for up to sixmonths, illustrating the challenge of optimal group size selection for vaccine evaluation in this species.
The enhanced natural ability to control M. tb-induced disease may also play a role in the strong protective e cacy afforded by BCG vaccination to cynomolgus macaques that is in contrast to the lower level of protection conferred by BCG to rhesus macaques 31,29 . Therefore, the natural resistant to TB of CCM populations is likely to confound e cacy measures when this species is used for the evaluation of prophylactic vaccines; however, this same feature is a useful attribute for the study of latent or chronic disease states.

Immune responses
As the mycobacterium-speci c IFNγ response provides a biomarker often used in clinical trials [44][45][46][47] and TB diagnostics 48 , the IFNγ response to vaccination and high dose aerosol challenge in the CCM was measured using several complimentary detection methods. Low levels of IFNγ were induced by BCG vaccination in CCM, which contrasts with the signi cantly increased IFNγ responses measured in similarly treated rhesus macaques 31,29 and suggests that IFNγ alone does not provide a su ciently sensitive biomarker of vaccination in this model. IFNγ SFU as measured by ELISPOT increased after exposure to M. tb in both vaccinated and unvaccinated groups. The frequency of PPD-speci c IFNγ producing cells increased more rapidly in BCG vaccinated animals during the initial phase after challenge with an early peak in response measured two weeks after challenge. In comparison, the PPD-speci c IFNγ responses reported in similarly vaccinated and challenged rhesus macaques typically occurred four weeks after challenge 18,31 . This rapid onset of the adaptive cellular immune response in the CCM may be a factor in the improved protection conferred against challenge and is in keeping with similar observations of early increases in pro-in ammatory cytokines reported by Dijkman et al in comparison to rhesus macaques 49 .
Following TB infection, populations of multifunctional CD4 + T cells increased in both vaccinated and unvaccinated groups of cynomolgus macaques, with a trend for frequencies of these cell populations to be greater in the unvaccinated group in which the highest pathology scores were recorded. This supports the hypothesis that high numbers of multifunctional CD4 + T-cells are related to active disease 19 .
However, within the unvaccinated group, animals that were able to control disease to the end of the study had the highest levels of multifunctional cells in comparison to animals in which disease progressed to meet end-point criteria. Therefore, it appears that polyfunctional T-cells contributed to disease control for unvaccinated animals, and that polyfunctional T-cells are a central component of the mycobacterial response but are not necessarily a correlate of protection.
There appears to be a much larger population of terminally differentiated memory (TEM2) cells in CCM in comparison to rhesus macaques, where memory T-cell populations were more biased towards transitional effector memory cells in naïve animals 18,29 . These terminally differentiated memory cells are considered to be a short lived population capable of exerting a rapid but more limited range of functional parameters 17 . TEM2 cells have recently been shown to be promoted following vaccination with cytomegalovirus (CMV) vectored TB vaccines and were associated with the improved protective e cacy imparted to rhesus macaques following M. tb infection 12 , The natural bias towards TEM2 populations we have observed in CCM in this study and the early responses after infection may contribute to the inherent ability of this species to control TB. M:L ratio and monocyte subtype at baseline correlated with pathology scores, and a role for the innate M:L ratio and host factors in determining the risk of developing active disease may exist. In our previous work, we have demonstrated that different populations of macaques have varying M:L ratios, which seems to correspond to their ability to control disease 50 . The Chinese cynomolgus macaques have a signi cantly lower M:L ratio than rhesus macaques and Mauritian cynomolgus macaques 50 and are better at controlling TB disease 30,49,51 , which corresponds to human data suggesting that high M:L ratio is a risk factor for TB 52,53 . Differences in monocyte sub populations between rhesus and CCM have been reported elsewhere; higher levels of mycobacterial permissive monocytes were found in rhesus macaques and were thought to have a negative effect on T-cell responses and have a role in the susceptibility of this species to TB 49 . Speci c cell populations have been identi ed that may be implicated, namely; the TEM2 T-cell phenotype with a low threshold of activation, and pro-in ammatory monocyte subsets. Further work