We screened 67 adults within 6 months of TB treatment, of whom 48 met inclusion criteria and underwent FDG PET-CT scans, spirometry, respiratory questionnaires, and exercise tolerance tests (6MWT and 5 TSTS), at a median 62 days after TB treatment completion (range 4–160 days). Due to challenges with equipment (gas stock-out) or participant compliance, acceptable plethysmography and DLCO could only be acquired in 31 participants. 40 participants underwent BAL, seven were excluded due to safety concerns, and one lost to follow up. Figure 1 shows the flow of recruitment and study procedures.
Abnormalities on pulmonary function tests and FDG PET-CT were common and highly variable after successful TB treatment
Table 1 shows demographics and other characteristics for all participants, subdivided into those with normal and abnormal spirometry. We classified 26 (54.2%) of the participants with normal spirometry, and 22 (46%) in the abnormal range. Of those in the abnormal range, there were 6 (12.5%) with an obstructive pattern, 9 (18.7%) with possible restriction, and 7 (14.6%) with a mixed pattern. Only one participant displayed significant bronchodilator response (> 10% increase in FEV1 or FVC). Impaired forced mid-expiratory flow (FEF25-75) was found in 23 (48%) of the participants, and distributed in all spirometry class groups, including 15% of those with FVC and FEV1 within the normal range, 83% in the obstructive group, 78% in the restrictive group and all in the mixed group (Supplementary Table 4a). Seven of the 23 (30.4%) showed a > 10% improvement in FEF25-75 with bronchodilation. 17 of 31 (55%) had some diffusion abnormality mostly attributable to a pulmonary vascular abnormality (Supplementary Table 4b). Abnormalities on plethysmography or DLCO were detected in ten of the 26 (38.5%) participants with spirometry in the normal range.
On CT, all participants had lesion(s) attributable to the aftermath of TB. Residual FDG avidity on PET was present in 41 (85.4%), defined by total glycolytic activity (TGA) > 5 or SUVmax higher than mild (compared to aortic blood pool as reference). There was a wide range of morphological features, usually co-occurring in an individual (Fig. 1). Small nodules were the most common, affecting 31% of segment groups, followed by fibrotic scarring (27%), cavitation (16%), large nodules (15%), bronchiectasis (12%), bullae (10%), and consolidation (4%). Differences in the gross pathology between the groups were not as clear as expected, with adjacent lobes and segments often reflecting contrasting fates; and areas of low density often surrounding dense inflammatory or fibrotic lesions.
Transverse and sagittal views of CT images with PET overlay.
Figure1a: A participant with no FDG-avid lesions in lungs, but some small nodules and linear fibrotic lesions in the right apex and right middle lobe. Figure 1b: A participant with the most residual glycolytic activity on PET. The image shows large areas of destruction, cavitation, fibrosis, and nodules – especially in the upper lobes. Figure 1c: A participant with the median FEV1 value in the normal spirometry (FEV1 and FVC normal) category. The right upper lobe shows diffuse small nodules and posterior dense lesions with calcifications, as well as decreased density of lung tissue. Figure 1d: A participant with the median FEV1 value in the obstructive category (low FEV1, normal FVC, and low FEV1/FVC ratio) on spirometry. The right upper lobe shows numerous small nodules, cavitation, and an apical dense lesion with calcifications, as well as decreased density lung tissue; the right lower lobe shows numerous small nodules. Figure 1e: A participant with the median FEV1 value in the restrictive category (low FEV1 and low FVC and normal FEV1/FVC ratio) on spirometry. There are dense lesions and loss of volume in right upper and lower lobes, with heterogenous pathology in other lobes. Figure 1f: A participant with the median FEV1 value in the mixed category (low FEV1, low FVC with low FEV1/FVC ratio) on spirometry. There are emphysematous changes and cavitation in right upper lobe, with heterogenous pathology in other lobes.
There was overall agreement on severity of disease between FDG PET-CT and pulmonary function testing
Those with abnormal spirometry had a worse median for several severity of disease indicators as shown in Table 1, including: (1) lower BMI, (2) higher number of lobes with any lesions in keeping with previous TB on CT, (3) larger cavity volume, and (4) larger volume of high-density lesions (>−100 HU).
Quantified image metrics correlated with pulmonary function test results (Fig. 2a) as follows: (1) Volume of low density lung on CT to FEV1, FEF25-75, and Residual Volume/Total lung capacity ratio – all indicative of large or small airway obstruction or gas trapping, (2) Features of inflammation, destruction, and fibrosis (respectively TGA, cavity volume, high density lesion volume) to low FVC, indicative of loss of parenchyma. TGA, cavity volume and dense lesion volume correlated more strongly with FEV1 and FEF25-75; and FVC and FEV1 values were also strongly correlated (p <.0001). When further exploring the correlation between the lobes affected and pulmonary function as shown in Table 2, we found a trend for both the decrease in total lung capacity, as well as the increase in residual volume when more lobes were affected.
Exercise tolerance and quality of life after TB treatment were dissociated from imaging severity and pulmonary function testing
Most participants performed well in the exercise tests and did not indicate severe subjective disability on the questionnaires (Table 1). We found no difference in exercise tolerance (6MWT and 5 TSTS), or quality-of-life questionnaires (the SGRQ and CAT COPD) between the spirometry groups. Furthermore, these measures did not correlate with conventional pulmonary function test measures of severity (FEV1, FVC) and CT characteristics, but the SGRQ and 6MWT was positively correlated with measures of inflammation (TGA & SUVmax) on PET (Fig. 2b).
(a) Correlation of quantified imaging variables to pulmonary function. Cross marks indicate non-significance (p ≥.05). Note that increasing lung function values indicate improving function (except for residual volume/total lung capacity ratio) while high imaging values indicate more pathology, hence the inverse correlations. (b): Correlation plot of quantified imaging variables and lung function to effort tolerance tests, symptom scores and clinical variables showing the lack of correlation between these tests.
Concentrations of proinflammatory cytokines in serum were correlated with measures of inflammation and the number of lung segments with lesions on FDG PET-CT
Correlations were observed between inflammation markers on PET (TGA or SUVmax), and both Type 1 (interferon (IFN)γ, tumour necrosis factor (TNF)α, interleukin (IL)−2, IL-12) and Type 2 (IL-4, IL-33) pro-inflammatory responses, regulatory response (IL-10), chemo-attractants (CXCL-9, CXCL-12), selected growth factors, matrix metalloproteinases (MMPs 1 and 8), and tissue inhibitors of MMPs (TIMP) 1 and 3) (Fig. 3a). The number of affected bronchopulmonary segment groups on CT had similar association patterns, but with significant correlations to MMPs (1,8) and chemo-attractants (IL-8, CCL-2), and fewer correlations with Type 1 and Type 2 response proteins. Pulmonary function tests were not correlated with inflammatory cytokines, but worse function did correlate with levels of selected growth factors (PDGF, FGFβ), TIMP-1 and MMPs (1,8). Surfactant protein D (SP-D), which is crucial for surfactant regulation and innate immunity and may induce immune pathology6, correlated with markers of severity, extent of lung involvement and worse pulmonary function. Conversely, the immunomodulatory cytokine, uteroglobin, correlated to better pulmonary function.
Worsening severity of cavity volume, FDG PET inflammation, FEV1, and symptoms were correlated with higher levels of markers of leukocyte trafficking,collagen degradation and remodelling, regulation and repair in BAL
The analysis of BALF proteins was markedly influenced by age, smoking, and illicit drug use (which appeared to have a global suppressive effect) more clearly than serum proteins (Supplementary Table 3), although some notable correlations remained after correction. More extensive lesions and worse lung function were correlated to lower IL-2 (in contrast to serum values), higher IL-8 and other chemo-attractants like CCL-18, levels of adiponectin, fibronectin, MMP-2, and TIMP-1 (Fig. 3b). The cavity volume correlated positively with levels of BAL CCL2, TIMP-1 and adiponectin, while FEV1 was inversely correlated with IL-8, CCL18, fibronectin, TIMP-1, adiponectin, and MMP-2, and positively with IL-2 levels. The amount of glycolytic activity on FDG-PET correlated positively with CCL18, A2M, fibronectin, adiponectin, and MMP-2. Interestingly the SGRQ was positively correlated with CCL2, CCL18, and adiponectin.
(a) Correlation matrix of quantified imaging variables, pulmonary function tests and participant characteristics to protein factors in serum. Cross marks indicate non-significance (p ≥.05). Note that increasing lung function values indicate improving function (except for residual volume/total lung capacity ratio), while high imaging values indicate more pathology, hence the inverse correlations. (b): Correlation matrix of quantified imaging variables, pulmonary function tests and participant characteristics to protein factors in bronchoalveolar lavage fluid.
Higher measures of inflammation and the severity of lesions on FDG PET-CT correlated with reduced regulation of inflammation by T helper cells, and priming of cytotoxic T cells for migration to the site of disease
Participants with higher CT scan measures of involvement (low- and high-density volumes, cavity volume, number of segment groups involved) and higher FDG-PET avidity (TGA, and SUVmax) correlated positively with clusters of helper T cells displaying memory (CD45RA-) and effector phenotypes (CD197-), and negatively with naive T cells (CD45RA + CD197+). Additionally, these features of radiologic involvement were positively correlated(Kendall’s tau < 0.2) with percentages of CD194 + Th2 and CD183 + Th1 cells, and negatively with the presence of CD196 + Th17 subpopulations (Fig. 4a). The Treg population (CD25 + CD127 lo) showed a weak negative correlation with TGA. We hypothesize that this may represent reduced Treg regulation of inflammation, with downstream dysregulation of TGFβ contributing to fibrosis.
CD8 T cells displayed similar results (Fig. 4b). The naive CD45RA + CD197 + population displayed a negative correlation with extent of disease, while chemokine receptor-expressing effector cells (Teff and Tem) and central memory (Tcm) populations displayed positive correlations. Specifically, we observed more subsets expressing CD194 (Tc2), CD196 (Tc17.1), and CD25 (Treg), and conversely low percentages of CD183 (Tc1) expressing subsets among participants with more extensive lesions, suggesting that Tc17.1, Tc2, and Tcreg subsets may be associated with lung pathology. CD103 expression was also present among those CD8 subsets correlated with lung pathology. Altogether suggesting that activated effector CD8 T cells are primed for migration to the site of inflammation.
Percentages of B cells (Fig. 4c) expressing CD25 and CD196 (denoted activated B cells) were lower among participants with severe lung pathology. These B cells are also CD183-, suggesting they are less migratory to inflammatory sites. Conversely, memory B cells expressing CD45RA, CD27 and CD183 were positively correlated with some features of involvement. Results of the sensitivity analysis are shown in the Supplementary materials.
Heatmaps displaying correlations of cell phenotypes with clinical features of lung disease. Cells were clustered according to their expression of the markers shown in each expression heatmap. Metaclusters were annotated according to expected phenotypes for memory and functional subsets. Frequencies of each subset were correlated to clinical features of pulmonary function and disease using Kendall’s rank correlation test. The frequency of each subset among the total cell population across all participants is shown as a boxplot. The median expression of markers used in clustering was scaled and is shown for each metacluster, defining cell function and phenotype; the range and median of expression was used to denote expression intensity in each expression heatmap. Figure 4a: CD4 + T cells. Treg were defined as CD25 + CD127 lo/- among total CD4 + cells. Memory phenotypes were inferred by the expression of CD45RA and CD197 among non-Treg CD4 + T cells. Subpopulations of effector T cells were identified by the expression of CD194, CD183, or CD196 among CD4 + CD197- non-Treg cells. Figure 4b: CD8 + T cell subsets were inferred similar to the CD4 + T cell criteria, using the CD8 + CD4- population of T cells. Figure 4c: B cells were clustered according to their expression of CD45RA, CD27, CD25, CD197, CD196, and CD183 among CD19 + CD3- lymphocytes. Activation and memory phenotype was defined by the expression of CD25 for activation and CD27 for memory. Homing was inferred by the expression of CD197 for movement to lymph nodes and CD183 for homing to sites of inflammation.
