The spleen assumes a major role in blood glucose regulation in type 1 diabetes patients treated with BCG

Abstract

The Bacillus Calmette-Guérin (BCG) vaccine, which has been used for > 100 years to prevent tuberculosis, is well-established for bladder cancer treatment, and under study for neurological and autoimmune diseases. In patients with type 1 diabetes (T1D), BCG vaccinations have been shown in randomized clinical trials to gradually lower blood sugar to near normal levels. This effect appears to be driven by a BCG-induced shift in lymphoid cells’ glucose metabolism from oxidative phosphorylation to aerobic glycolysis. The latter is a state of high glucose utilization that draws more glucose from the blood. Apart from blood, it is unknown whether BCG establishes residence in any organs and alters their glucose metabolism. In this two-year-long clinical trial in type 1 diabetics, we use positron emission tomography (PET) and x-ray computed tomography (CT) to map organs that increase their uptake of the glucose analogue ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) before versus after BCG vaccinations. We also injected BALB/c mice with BCG to test for the presence of BCG in various organs. Results from both studies point to the spleen as the dominant site for glucose uptake and BCG residence. The human spleen is significant because its 47% increase in ¹⁸F-FDG uptake by a large population of lymphocytes and monocytes might help to explain BCG’s systemic lowering of blood glucose to near normal levels. Findings suggest that the spleen, triggered by BCG, assumes a critical role in systemic glucose regulation in the absence of a functional pancreas.

Introduction

Mycobacterium tuberculosis (Mtb) and other pathogens can change the energy metabolism of the host^1–3. Murine hosts infected with Mtb in the lung switch monocyte’s energy metabolism from oxidative phosphorylation to aerobic glycolysis; the switch occurs in the lung granuloma niche⁴. Aerobic glycolysis is a state of high glucose utilization that draws significantly more glucose from the blood. While aerobic glycolysis is a faster means to energy (ATP) production, it is less efficient than oxidative phosphorylation. The Mtb-induced shift from oxidative phosphorylation to aerobic glycolysis is a multi-step process involving upregulated glycolysis enzymes and transporters for glucose uptake and downregulation of enzymes of the tricarboxylic acid cycle (TCA cycle) and oxidative phosphorylation pathways^1–4.

Lessons learned from tuberculosis (TB)-induced changes to host metabolism apply to the closely related Mycobacterium bovis (MB) organism rendered avirulent in the Bacillus Calmette-Guérin (BCG) vaccine. The BCG vaccine is a > 100 year old vaccine developed for tuberculosis protection, and it has been administered to over 3 billion people globally⁵. Since the 1970s it has become a well-established treatment for bladder cancer⁶. Over the last two decades BCG is under study for its therapeutic benefits in neurology, infectious disease, and autoimmunity^7–19.

We have shown that BCG is a promising experimental treatment for type 1 diabetes (T1D). T1D patients with longstanding disease have no endogenous insulin production due to autoimmune destruction of their insulin-secreting pancreatic islet cells. These patients respond to multidose BCG vaccinations in placebo-controlled trials by gradually lowering their blood sugar levels over a 2 to 3 year period to near normal values for the next 5 years¹⁷. One important mechanism driving this beneficial effect appears to be a BCG-induced shift in host energy metabolism from oxidative phosphorylation to aerobic glycolysis^{12,16,17,20,21}. The acceleration of glucose metabolism, also called the Warburg effect, occurs gradually over a 2–3 year time period in association with lowering of blood sugars¹⁷. We have shown that T1D patients have an underlying defect in glucose uptake in a variety of white blood cell lineages such as monocytes, T cells and PBLs^12,16,17,22. The underlying defect leaves the lymphoid cells of these patients with overactive oxidative phosphorylation and suppressed aerobic glycolysis. BCG vaccination shifts their glucose metabolism to aerobic glycolysis. The BCG-induced shift to aerobic glycolysis is demonstrated in patients’ monocytes as well as their lymphocytes and peripheral blood lymphocytes (PBLs)^{12,16,17,20,21}.

In this clinical trial of T1D patients, we map over a two-year period the organ-specific uptake of the glucose analogue fluorine-18 fluorodeoxyglucose (¹⁸F-FDG) before versus after BCG vaccination. (For ethical reasons we were unable to study placebo patients.). The major question is what organ(s) show increased glucose uptake to an extent that may explain the systemic benefit of BCG, i.e., its long-term lowering of blood sugar levels. Additionally, we ask whether the BCG-vaccinated mouse exhibits similar anatomic niches of microbial residence.

Materials and methods

Overview of trial design

We began by confirming prior research on the in vivo benefit of multidose BCG for blood sugar control (via HbA1c monitoring) and by demonstrating that the effect is not the result of endogenous insulin production. We then sought to define metabolic niches for BCG vaccine-induced increase in glucose metabolism and to investigate the possible multiple human metabolic niches for this long-term microorganism effect. We mapped the changes in glucose uptake before vs after multidose BCG vaccination on an organ-by-organ basis. To do this, we employed positron emission tomography (PET) and x-ray computed tomography (CT) scanning to acquire images from injected ¹⁸F-FDG, a positron-emitting analogue of glucose whose biodistribution is similar. The ¹⁸F-FDG PET/CT mapping study in T1D patients was an open-label clinical trial. The scanning technology allowed us to accurately define human organ regions in 3D space over 2 years after BCG. To buttress our findings, we performed experiments in mice to evaluate organ-by-organ tracking of BCG microbe colonies after BCG injections in the footpad. We then compared glucose uptake distribution in human organs before and after BCG vaccinations to actual organ niches of residence in mice.

Human participants and clinical trial design

All human studies had full institutional approvals through Massachusetts General Hospital and Partners Health Care (Study# 2013P002633). Clinical trial subjects (n = 6) were open-label BCG-treated subjects with longstanding T1D (duration over 13 years at the time of baseline injection) given ¹⁸F-FDG. We also performed an in vitro study that compared glucose uptake in isolated monocytes from BCG-treated and BCG-untreated T1D subjects (total n = 11) as measured with the fluorescent derivative of glucose 2-NBDG. These two BCG human studies were formally approved by the FDA (IND#2013P16434) (ClinicalTrials.gov Identifier: NCT02081326 07/03/2014) and involved the Tokyo 172 BCG strain from Japan BCG Laboratory, Tokyo, Japan. Informed consent was obtained from all subjects, and the experiments conformed to the principles set out in the WMA Declaration of Helsinki and the Department of Health and Human Services Belmont Report. Characteristics for the PET/CT patients at the trial baseline injection visit were as follows: Average Diabetes Onset (years) 12.5 ± 4.5; Average Age (years) 28 ± 4.5; Average Diabetes Duration (years) 14.8 ± 1.5; %female 16.67%; %male 83.33%; Average HbA1c 8.07 ± 0.46.

The BCG treatment regimen for the patients enrolled in the ¹⁸F-FDG study involved one BCG dose at baseline, another dose 4 weeks later, and then yearly repeat doses, for a total of 6 doses over a period of 2 years with this study. Since this functional PET study required irradiation, the ethics committee did not approve inclusion of an untreated T1D control group. However, since the first PET/CT scan was before the first BCG treatment, each patient was their own control and subsequent scan results were compared to this baseline scan. BCG dosing was according to the manufacturer’s recommendation. Briefly, each vial of BCG Tokyo contained 0.5 mg of lyophilized BCG and averaged 27.2 × 10⁶ CFU/vial. No longer than 3 h before the treatment, a vial of BCG was reconstituted with 1 mL of sterile saline for injection, yielding 10 adult doses of 100 µL each. Administration of the BCG dose was performed by subdermal injection in the posterior deltoid region of the upper arm using a 1 mL TB syringe with 27G x ½” needle.

Glucose uptake in culture: 2-NBDG glucose uptake by human monocytes

All human studies are restricted in the total number of blood tubes that can be drawn for research studies at each visit. Although the white blood cell lineage defects in glucose utilization are present in white blood cells in general, in this study we performed the supporting in vitro studies of augmented glucose uptake on one cell type only, monocytes to have enough harvested cells for the assays. The EasySep™ Direct Human Monocyte Isolation Kit (Stemcell Technologies, Cambridge, MA) was used to isolate human primary monocytes from blood donated by BCG-treated, as well as untreated T1D subjects, according to the manufacturer’s protocol (Supplemental Fig. S1). Briefly, 1000 μL of Isolation Cocktail and 1000 μL of RapidSpheres were mixed with 20 mL of whole blood in a 50 mL centrifuge tube and incubated for 5 min at room temperature. 30 mL of PBS containing 1 mM EDTA (Sigma Aldrich, St. Louis, MO) was added and the tube placed into an “Easy 50” magnet. After 10 min, the monocyte-enriched suspension was transferred into a new tube. The magnetic separation process was then repeated for five minutes with the same amount of fresh RapidSpheres. The resulting suspension was transferred into a new tube and purified for a third time using the magnet. Typical purity was > 94%.

1 × 10⁶ of isolated human monocytes were then cultured in 24-well Nunc UpCell culture plates (Thermofisher Scientific, Waltham, MA) using 1 mL Immunocult™-SF Macrophage Medium (Stemcell Technologies, Cambridge, MA) supplemented with 100 U/mL penicillin, and 100 µg/mL streptomycin (ThermoFisher Scientific). Half of the wells were incubated with 1 × 10⁵ CFU of BCG (JAPAN BCG Laboratory, Tokyo, JAPAN) in culture medium and the others with culture medium only. Overnight culture was at 37 °C in 5% CO₂ and 95% air. The next day the warm culture media was replaced with 1 mL of cold XFP buffer (Agilent Technologies, Santa Clara, CA) and the plates were left at room temperature for 20 min to allow the cells to detach.

Harvested monocytes were washed with XFP buffer, counted, resuspended at 250 k monocytes in 50 μL of XFP, mixed with 50 μL 2-NBDG in XFP (ThermoFisher, end concentration 0·1 mM) and incubated at 37 °C in a CO₂-free incubator for 20 min. Five μL of APC anti-human CD14 antibody (Stemcell Technologies, Cambridge, MA, USA) was then added and the solution incubated for a further 10 min. Unstained autofluorescence control samples were incubated with neither 2-NBDG nor CD14 antibody. Following incubation, the cells were washed with XFP buffer, resuspended in 300μL of XFP buffer and analyzed on a FacsCanto II flow cytometer (BD Biosciences, San Jose, CA). FlowJo software (FlowJo LLC, Ashland, OR; division of Becton, Dickinson & Company) was used to analyze the data by quantifying 2-NBDG fluorescence intensity (MFI). Monocytes were identified by creating a lymphocyte APC-CD14^high gate and 2-NBDG fluorescence data was collected for these gated cells in the FITC channel. Typically, CD14^high purity for monocytes was > 85%. An example of the gating strategy is shown in Supplemental Fig. S1.

¹⁸F-FDG PET/CT scan analysis

The six open-label T1D clinical trial patients treated with BCG were monitored annually via ¹⁸F-FDG-PET/CT scans, with a baseline scan conducted before BCG treatment and annual follow-ups for two years afterwards. The BCG treatment regimen for these patients involved one BCG dose at baseline, another dose 4 weeks later, and then yearly repeat doses. Since this functional PET study required irradiation, the ethics committee did not allow inclusion of a diabetic control group that was not treated with BCG. The baseline PET/CT scan was performed before the first BCG treatment, so we analyzed this data by comparing subsequent PET/CT scans for each patient to this first baseline scan.

A GE Discovery MI 5-Ring PET/CT clinical research scanner with software version 3.0 SP2 was used for the scans. For protocol 2013P002633, patients were required to exhibit a blood glucose range of 80–225 mg/dL and to have been fasting for at least four hours prior to the scan. ¹⁸F-FDG PET data (injected dose: 10 mCi ± 10%) were collected for 60 min starting at the time of injection of ¹⁸F-FDG. The following parameters were applied: chest CT Parameters: Scan Type: Helical/0.5/Full, 40 mm detector coverage/3.75 mm helical thickness/1:531:1 pitch & speed (mm/rot)/0·5 rotation time; kV:140 & mA:20. 60 Minute Dynamic Imaging of the Chest: arms up; VPFX w/ SharpIR; no z-axis filter; 5·0 mm cutoff; 3 iterations, 16 subsets; 256 MATRIX MAC; DFOV 70 cm; Binning: 8 × 15 s, 4 × 30 s, 6 × 60 s, 5 × 120 s, 8 × 300 s.

Immediately following the dynamic study, a whole-body PET scan was performed from the Apex to mid-thigh. Whole Body CT Parameters: Scan Type: kV:120 & mA:40 Helical/0·5/Full, 20 mm Detector coverage,3.75 mm helical thickness/1:531:1 pitch & speed (mm/rot), DFOV 70 cm, Std Wideview, Full 400/40 width/level. Whole Body Imaging vertex to midthigh: arms down & remove insulin pump/ Static/VIP Off/Binning: 5 m per bed (height based) /VPHD, Z axis filter: std, 5·0 filter cutoff, Dynamic PET data were reconstructed with ordered subset estimation maximization (OSEM) algorithm with 3 iterations and 16 subsets. CT scans (Thorax scan: Helical CT with 3.75 slice thickness, 140 kV and 20 mA; Whole Body scan: Helical CT with 3.75 slice thickness, 120 kV and 40 mA) were acquired for attenuation correction.

Standardized Uptake Values (SUV) were calculated in Osirix Lite (v·12·0·3) by normalizing the PET activity by the participants’ injected dose and body mass in regions of interests (ROIs). Spherical ROIs were manually drawn based on the CT images and placed bilaterally for anatomically symmetrical tissues if appropriate. In the sagittal view, ROIs were placed for the spleen (5 cm³), liver (50 cm³), 2 × kidneys (1 cm³), 2 × biceps (15 cm³), 2 × thighs (15 cm³); and in the axial view, the descending aorta (1.5 cm³), 2 × Pectorals (5 cm³), bone marrow (spine, L2 lumbar region) (1.5 cm³). Lung segmentations were generated automatically using AI Rad Companion Chest CT Solution software (Siemens Healthineers, Erlangen Germany), based on the CT images. The mean SUV was collected for each ROI.

For the brain, the PET images were cropped to isolate the organ, and then warped into the Montreal Neurological Institute (MNI) space by co-registering to a FDG template²² using FSL Flirt (the FMRIB [Functional Magnetic Resonance Imaging of the Brain] Software Library). The mean SUV was extracted in brain regions: 3–12, 19–22 (the lateral frontal cortex) from the brain Automated Anatomical Atlas 3v1 (AAL3v1)²³ in the MNI space as this region has been observed to show the biggest changes in SUV²⁴.

Target region SUV ratios (SUVR) were generated by normalizing the target region SUV by the reference region (cerebellum) SUV. The reference region was extracted from numbers 95 to 112 from the AAL3v1 atlas²³. Bilateral target region SUVRs were averaged.

Serology values

HbA1c values were measured by Lab Corp (Burlington, NC). C-peptide was quantified using Mercodia’s Ultrasensitive C-peptide ELISA (Uppsala, Sweden) following the manufacturer’s protocol.

Mice BCG culture experiments

Male BALB/c mice (n = 17) were ordered at 4 weeks old from Charles River Laboratories (Wilmington, MA) and were housed at the MGH animal facilities. The care of mice and experimental procedures complied with the ‘‘Principles of Laboratory Animal Care’’ (Guide for the Care and Use of Laboratory Animals, National Institutes of Health publication 86-23, 1985). The experimental protocol was approved by the Institutional Animal Care and Use Committee of Massachusetts General Hospital (Approval protocol# 2017N000137). The starting weights of the BALB/c mice were tracked and had tight measurements of 17.1 g ± 0.2 g at 4 weeks. All BCG treatments were then started at 8 weeks old when the mice weighed 26.1 g ± 0.5 g. The BCG was injected into the footpad at a volume of 25 μl to simulate an intradermal/subcutaneous injection (2 mg/mL in saline; Japan BCG Laboratory, Tokyo Japan). For our batches of BCG Tokyo, the viability counts average 27.2 × 10⁶ CFU per vial containing 0.5 mg lyophilized BCG. For human use, one vial was reconstituted in 1 mL of saline and each adult dose is 100 µL which thus contained 2.72 × 10⁶ CFU BCG. We used this same dose in mice, but since the volume that can be injected into the footpad was very limited, we prepared a 4 × BCG concentration but administered 25 µL. So, the BCG dose per mouse was equal to the human dose at 2.72 × 10⁶ CFU.

The BALB/c mice were euthanized by CO₂ asphyxiation at various intervals post-BCG treatment to understand the possible migration of the BCG organisms into different organ structures. The BCG time migration intervals after BCG infections were 4 weeks at 1–4 (n = 7), 5–8 (n = 3), 9–12 (n = 3), 13–16 (n = 3) and 17–20 (n = 1) weeks after BCG treatment. Tissues from liver, spleen, lungs, bone marrow (from femurs) and muscle were collected. We used the whole spleen and the bone marrow from both hind femur bones. For the lung, liver and muscle we used only a part, comparable in size to the spleen. The tissues were homogenized under sterile conditions by mashing the tissues with the help of sterile syringe plungers through 100 µm mesh cell strainers into ice cold PBS. The tissue homogenates were then centrifuged, resuspended in 300 µL sterile cold saline and plated on Middlebrook 7h11 agar plates (Hardy Diagnostics, Catalog Number W40, Santa Maria, CA), specifically designed to only grow Mycobacterium. Plates were wrapped in aluminum foil and incubated at 37 °C with 5% CO₂. The plates were set up for serial weekly reads to detect colony formation. Plates were checked weekly; colonies were counted, and images taken at week three. The number of colonies was considered indicative of the number of BCG bacteria per mouse organ. To confirm that the colonies were specific for BCG, we prepared samples using the Capilia TB-Neo Extraction Buffer (TAUNS Laboratories, Catalog no. CATB0877, Izunokuni, Japan) for use with Capilia TB-Neo anti-MBP64 chromatography strips (TAUNS Laboratories, Catalog no. CATB0870, lot no. TN73201, Izunokuni, Japan). In addition, we matched morphology to published images of mycobacterium colonies²⁵. No anesthesia was required for any of the procedures requiring animals; mice were euthanized via CO₂ overdose.

Data analysis and statistics

The percent change in SUVR values was calculated between baseline and each sequential year (sequential year − baseline)/baseline*100). Analysis was conducted using Microsoft Excel version 16.43, graphs were created in Graphpad Prism version 9.0.2, and statistical analysis was run in SAS software to calculate P-values using the linear mixed effects model with subject level random effects to test for significance of longitudinal % change in the SUVR values. A mixed effects analysis with a Holm-Šídáks test for multiple comparisons was used to test the significance for change in HbA1c values in the six open-label patients. A one-tailed Wilcoxon signed ranked test was used to analyze the BCG positive mice percentages and BCG colony averages for the weeks that mice showed the presence of colonies. All statistical testing was assessed at p = 0.05 level.

Ethics approval and consent to participate

The mouse experimental protocol was approved by the Institutional Animal Care and Use Committee of Massachusetts General Hospital (Approval protocol# 2017N000137). All human studies had full institutional approvals through Massachusetts General Hospital and Partners Health Care (Study# 2013P002633) and the subjects in this study were open-label subjects specific to the FDG PET/CT study. The BCG interventional studies were also formally approved by the FDA (IND#2013P16434).

Results

BCG vaccinations in Type I diabetic subjects lowers hemoglobin A1c values

Hemoglobin A1c (HbA1c) values, which reflect average blood sugar levels over the past 3 months, were assayed for the six open-label T1D patients, once at baseline and then at 6-month intervals post-BCG treatment. The data show the 6 open-label subjects had a significant decrease (p = 0.026) in HbA1c values to near normal levels over the three-year study period (Fig. 1A). At the group-level, the HbA1c values decreased by almost 10% after BCG treatment. A 10% reduction is clinically significant because it is associated with a 37% lower rate of diabetes complications. The reductions in HbA1c values were statistically significant as early as week 0 vs week 52 (p = 0.0006), vs week 78 (p = 0.0237). It should be noted that as reported previously^17,26, T1D subjects with early age of onset (< 21 years of age) have a more rapid response to therapy, typically less than 2 years. These subjects had an average age of onset of 12.5 ± 4.5 (Fig. 1D).

Figure 1.

BCG vaccinations affect lymphoid glucose uptake in humans with T1D. (A) Average percentage change in HbA1c values over two years of time after four BCG vaccinations in adults with long-standing T1D (n = 6). A Repeated Measure one-way ANOVA yields a p-value of 0.0362; a Holm-Šídák's multiple comparisons test when compared to baseline, provided significant p-values for Year 0 vs Year 0.5 through to 2 (p-values are 0.0311, 0.0004, 0.0119, 0.0354 respectively), Shapiro-Wilks and Kolmogorov–Smirnov test showed normality. (B) C-peptide, a biomarker of residual pancreatic insulin secretion, was measured in all T1Ds receiving in vivo BCG (n = 6). The graph shows the sequential values for the 6 patients at baseline, and at year 1 and at year 2 after repeat BCG vaccinations. The data show negligible C-peptide production, indicating that no residual C-peptide can account for the restored glucose metabolism. The dotted line represents the average reference value of C-peptide in Healthy Non-Diabtic Controls (Mercodia Ultrasensitive C-peptide ELISA kit, Mercodia, Uppsala, Sweden). (C) Human monocytes harvested from T1D patients who were (n = 11) or were not (n = 11) treated with in vivo BCG vaccination (approx. 2 years prior) were cultured in vitro with or without BCG exposure. Glucose uptake was analyzed using a 2-NBDG assay. The 2-NBDG assays in the in vivo BCG group were performed at 3.5 (n = 9) or 4.5 years (n = 2) years after first BCG vaccination. Both groups of T1Ds, BCG-naïve (unvaccinated) and BCG-vaccinated, showed a significant increase in glucose uptake after in vitro BCG stimulation (p-values of 0.0026 and 0.0013 respectively, paired t-test). BCG vaccinated individuals also showed a significant increase in glucose uptake compared to BCG naïve individuals (p- value of 0.062, unpaired t-test). (D) Characteristics of the 28 T1D subjects presented in (A–C).

Lowered HbA1c is not due to increased insulin production

All six open-label T1D subjects in this study were adults with early age of onset of their disease, i.e., juvenile onset (average 12.5 ± 4.5 years). Also, they had long-standing T1D averaging 14.8 ± 1.5 years duration (Fig. 1D). To confirm here that the lower HbA1c systemic effect was associated with BCG-induced metabolic changes in glucose metabolism and not due to pancreas recovery, C-peptide was measured from serum (Fig. 1B). C-peptide is a molecule co-secreted with insulin and thus measures in vivo the actual amount of endogenous insulin from the pancreas, not exogenous insulin administered daily. Participants’ C-peptide values (Fig. 1B) clearly show a lack of insulin production when compared to the reference value for healthy non-diabetic controls (Mercodia Ultrasensitive C-peptide ELISA kit, reference standards, Mercodia, Uppsala, Sweden). There does not appear to be an increase in C-peptide after BCG treatment in these six open-label subjects with advanced disease and a pancreas that at the start of the experimental therapy secreted very little or no C-peptide or insulin. Our findings suggest that repeat BCG vaccination restores glucose uptake by correcting an underlying deficiency in aerobic glycolysis.

Both in vitro and in vivo BCG exposure increase glucose uptake in T1D lymphoid cells

It has been reported that BCG treatment increases glucose uptake in isolated white blood cells (T cells, monocytes, PBLs). While the 2-NBDG assay used here is commonly performed on monocytes, all lineages of type 1 diabetic white blood cells show the same metabolic effect^16,17. Here we studied cultured monocytes harvested from BCG-naïve (n = 11) and BCG-vaccinated T1D subjects (n = 11) to determine whether BCG exposure in vitro can also switch peripheral blood monocytes to higher glucose uptake. For the patients in the in vivo BCG group, the 2-NBDG assays were performed at 3.5 (n = 9) or 4.5 years (n = 2) years after first BCG vaccination. The monocytes from both BCG-naïve (unvaccinated) and BCG-treated (vaccinated) T1Ds showed a significant increase in glucose uptake after BCG culture using the 2-NBDG assay (Fig. 1C); p-values using a paired t-test resulted in 0.0026 and 0.0013, respectively. When comparing BCG-naïve and BCG-treated T1Ds not cultured in the presence of BCG, the BCG-treated group also had a significantly higher glucose uptake, reaching a p-value of 0.0062 using an unpaired t-test. The increase was also significant for the BCG-treated subjects after culture in the presence of BCG, attaining a p-value of 0.0032 with an unpaired t-test. Therefore, whether monocytes are treated with BCG in vivo or in vitro, the exposure induces increased glucose uptake as measured by the uptake of fluorescent deoxy-glucose molecules. This increase in glucose uptake nearly corrects the underlying metabolic effect characteristic of the underlying disease. BCG vaccinated individuals also showed increased uptake compared to BCG naïve members (p-value = 0.0062 unpaired t test).

Introduction to ¹⁸F-FDG PET/CT mapping results

The schematic in Fig. 2A provides a representation of how ¹⁸F-FDG PET/CT assists in identifying metabolic niches in the body that might show increased glucose uptake post-BCG treatment. Figure 2B provides representative scans of the spleen in the same subject at baseline, at year 1, 2. The raw SUVR values are shown. Using SUVR measures, significantly increased sugar uptake was observed over the 2 years of study in individual subjects for the spleen (Fig. 2C), (Repeated Measure one-way ANOVA, p-value = 0.0415).

Figure 2.

The Spleen shows significant increased glucose uptake after BCG vaccination. (A) Schematic of organs evaluated for glucose uptake during PET/CT scans (B) ¹⁸F-FDG PET/CT images of the spleen over 2 years depicting the increase in SUVR. (C) Dot plot of the spleen presenting the increase in glucose uptake for the open-label patients on an individual subject. (p-value 0.0415, Repeated Measure one-way ANOVA).

Longitudinal PET/CT standardized uptake value ratio (SUVR) after vs before BCG treatment by human organ

Figure 3A shows the gating regions of various whole organs that were used to study increased glucose uptake. The organs studied for augmented glucose uptake were spleen, aorta, liver, bone marrow and lungs. The green dots indicate where the tissues were gated to measure glucose uptake with the exception of the lungs, for which the whole organ was integrated. We were interested in whether BCG vaccines could cause measurable alterations in ¹⁸F-FDG glucose uptake enabled through in vivo detection organ by organ with evaluations over a 2 year time course. We were also interested in identifying which whole organs in humans might show this accelerated transport. Earlier we showed that isolated in vitro monocytes from BCG-treated T1D subjects displayed a shift in metabolism toward more glucose uptake (Fig. 1C) and past data has shown the accelerated glucose uptake in type 1 diabetic humans year by year is present in all lymphoid lineages (T cells, CD4 T cells, PBLs, etc.)^{12,16,17,20,21}. Here we ask, what human organs show BCG’s functional impact of augmented sugar uptake? Using the PET/CT measures of ¹⁸F-FDG metabolism we looked at 3 lymphoid niches organs (spleen, bone marrow and descending aorta). We also evaluated the lungs and other non-lymphoid organs, such as the liver and muscle, because they are known to contribute to glucose metabolism. No prior studies asked whether BCG could affect non-lymphoid tissues. By evaluating the percentage change/organ uptake of glucose over three longitudinal sampling times, the functional data show ¹⁸F-FDG metabolism can identify organ-to-organ differences in glucose uptake after BCG exposures (Fig. 3). The data are presented for the 3 scans over a 2-year time period during which HbA1c values declined from BCG treatment in the same subjects that were discussed in Fig. 1A,B. Of the three lymphoid organs studied—spleen, descending aorta and bone marrow—only the spleen exhibited significantly increased glucose uptake (p-value = 0.03) using mixed linear effects model statistics (Fig. 3). The time course shows a steady increase in the spleen’s glucose uptake over the first two years. Overall, the spleen showed a 47% increase in glucose uptake compared to baseline, making it the dominant organ for BCG-induced sugar uptake rise. The lymphoid cells within the descending aorta and the lymphoid cells within the bone marrow trended upward but were not statistically significant by year 2 compared to baseline. Also, the lungs, the common site of Mtb residence, and the liver displayed slight increases, albeit only a transient rise. The bone marrow showed a non-statistically significant trend of augmented glucose uptake, but uptake leveled off by year 1. Supplementary Figure S2 shows scans of non-lymphoid tissues that did not show any increased glucose metabolic trends and SUVR %change values for these tissues, indicating that no statistically significant increases were observed in brain (lateral frontal cortex), bone marrow, spine, pectorals, biceps, thighs, and kidneys.

Figure 3.

The ¹⁸F-FDG PET/CT images and scans performed in BCG-treated T1Ds (baseline, year 1, and year 2) to identify the human organs with accelerated glucose uptake. (A) ¹⁸F-FDG CT/PET images of tissues showing an accelerated glucose uptake signal with the gating regions used to monitor uptake. Except for the lung, circles represent the areas where the mean SUV was measured. Right image is the 18F-FDG PET scan alone, with a darker color indicating greater glucose uptake. The left image is the CT superimposed onto the 18F-FDG PET image (red/yellow), with the brighter the color indicating the greater the glucose uptake. For the lungs, segmentations were generated automatically using AI Rad Companion Chest CT Solution software (Siemens). The 18F-FDG PET scan image of the lungs shows lung segmentation superimposed, in different views. (B) Glucose uptake monitored by ¹⁸F-FDG methods at baseline, at year 1 and year 2 after BCG vaccinations, and then compared to baseline glucose uptake in each BCG-treated subject on an organ-per-organ basis (n = 6). The graphs show 5 tissues/organs with augmented glucose uptake (out of 10 studied): spleen, aorta, bone marrow, lungs, and liver. The spleen was the only organ to continue through year 2 versus baseline to show significantly more glucose uptake, according to a linear mixed effects model (p = 0.03).

BCG-treated mice exhibit organ-specific colonies

BALB/c mice (n = 17) were used to test the direct presence of BCG organisms in particular tissue niches after vaccinations (See Fig. 4A for a schematic representation). The purpose of the mouse study was to confirm the functional imaging data in humans showing BCG’s influence on glucose metabolism. Mice were injected with BCG via footpad and mice sacrificed and the organs harvested at various time points thereafter up to 20 weeks. This injection method in mice is both an intradermal injection (ID) and subcutaneous injection (SC) delivery, this is because as the footpad of the mouse is so thick the methodology falls between the two. The spleen, lungs, bone marrow, liver and muscle were harvested at various time points after BCG injection to determine colonization by BCG bacteria. Tissue homogenates were prepared, cultured, and then analyzed for BCG colonies. Middlebrook 7h11 tissue culture plates were used to grow the BCG bacteria. The results for individual mice were binned in periods of 4 weeks each. Although the Tokyo strain of BCG displays a distinct slimy white cauliflower morphology (Fig. 4B), we also confirmed that the slow-growing colonies were actual BCG using the Capilia TB-Neo anti-MPB64 chromatography strips. BCG in vaccinated mice was significantly present in the spleen for almost all the mice harvested. The colonies appeared in the spleen with harvests from week 1–4 and persisted there up to 20 weeks. The data also indicated that BCG bacteria may transiently appear in the bone marrow but only up to week 8 after vaccination. BCG bacteria were scarce in other harvested organs like the lungs and muscle (Fig. 4C). Colonies appear in cultures of the bone marrow and the liver early on, but their presence soon diminishes. The microbe was significantly present in the liver but only up until the period spanning week 9–12.

Figure 4.

BCG vaccinations in mice promote long term persistence of the BCG organism; evaluations of various organ sites and the density of BCG organism/tissue type. (A) Location and time course of BCG persistence in mice vaccinated with BCG. BCG-treated BALB/c mice were serially sacrificed to look at the tissue niches where the BCG organisms might persist as a live organism. The mice were asphyxiated with CO₂ and the organs harvested during time intervals grouped in intervals of 4 weeks at 1–4 (n = 7), 5–8 (n = 3), 9–12 (n = 3), 13–16 (n = 3) and 17–20 (n = 1) weeks after BCG treatment. Harvested tissues were tested for the presence of BCG by using the tissue homogenate and culturing in a media specific for BCG growth (see “Materials and methods”). Tissue from the spleen, lungs, bone marrow, liver, and muscle were observed for persistence of the BCG organism through colony-forming units in culture. (B) Images showing BCG colonies grown in culture and labeled according to the number of colonies visible; plates with greater than 10 colonies were counted as 10. (C) BCG culture experiments using tissues from periodically harvested mice (n = 17) show that BCG was significantly detected in the spleen for mice harvested up until week 20 (p-value =  < 0.0001). There are also significant signs of BCG being present in the liver for mice harvested until week 12 (p-value 0.0002) and in the bone marrow of mice harvested until week 8 (p-value = 0.0312). (One-tailed Wilcoxon signed ranked test).

The colony density per mouse tissue from each mouse was also assessed, with scores assigned based on counting the number of colonies visible on the plate (Fig. 4B). Because it quickly became impossible to distinguish individual colonies upon continued weekly monitoring of culture plates, plates with more than 10 colonies were assigned a score of 10 by default. We also studied the average number of BCG colonies per mouse (data not shown). In the lungs and muscle, there were minimal colonies, whereas the spleen showed higher colony levels over multiple weeks. At 13–16 weeks the overall presence of BCG over the 20 weeks was significant in the spleen. The liver was also observed to have a higher colony number up to 12 weeks post-BCG treatment, and the bone marrow suggested the presence of colonies for the first 8 weeks. The presence of BCG colonies was also significant in the lungs up to 8 weeks.

The human metabolic niche mapping was based on functional sites of increased uptake through a metabolic switch to aerobic glycolysis. The mouse niche studies were based on actual BCG within the organ based on counts of colony-forming units. Remarkably, whether defining the niche by metabolic function in a human or by presence of viable BCG in a mouse; both methods point to the spleen as the dominant metabolic niche in humans and bacterial niche in mice. The mouse time course was faster than the human time course, but the sites and sequence of appearance were identical (Fig. 5).

Figure 5.

Increased glucose uptake in humans measured by metabolic changes induced by BCG compared to the physical presence of the BCG microbe in mice. Scoring on a defined scale of 0–5 (see Table), based on the amount of BCG microbe present in mice tissue, and the degree of glucose uptake observed through the ¹⁸F-FDG PET scans, the spleen, bone marrow, muscle, liver and lung tissues correlate with each other. A score of 5 reflects the best signal for the presence of BCG or for the best glucose uptake observed.

Discussion

In this study of T1D patients, we map organ-specific glucose uptake before versus after BCG treatment using serial multi-year ¹⁸F-FDG PET/CT imaging. Originally developed to map high glucose-consuming metastatic cancer lesions, the ¹⁸F-FDG PET/CT technology shows organs with increased glucose uptake²⁷. Our clinical trial data showed that the spleen, with its large mass of diverse white blood cell populations and lymphoid cells, is the main human organ for augmented glucose metabolism after BCG therapy. The human clinical trial data show steadily increased glucose uptake over a period that correlated in time with the systemic drop in blood sugar. Remarkably, mice injected with BCG show microorganism residence predominantly within the spleen and a similar sequential organ preference as seen in humans.

Our human and mouse data concur that the spleen provides the most robust and long-term metabolic niche in humans and organ niche in mice for higher glucose uptake or live organism persistence, respectively, after BCG vaccination. The spleen’s activity steadily and significantly increased over the course of 2-years of ¹⁸F-FDG mapping, and the microbe persisted there over a period of 20 weeks in the mice. The data also shows a plateau in other organs, such as the bone marrow, while the spleen continues to increase its glucose consumption over time.

The human spleen’s 47% increase in glucose uptake relative to baseline is mechanistically significant because it may help explain BCG’s systemic lowering of HbA1c. Until now it has been hard to argue that the lymphoid compartment of the blood, where we originally documented the BCG-induced metabolic shift to aerobic glycolysis, was sufficiently large to account for such a striking benefit as a clinically significant drop in systemic HbA1c levels. The identification of the spleen—likely with contributions from other lymphoid organs—as a site of the greatest increase in glucose uptake over the 2-year time course reported here suggests that glucose removal from the blood is occurring on a large enough scale to achieve lower systemic HbA1c levels. On a pure cellular mass issue, the spleen alone is the second largest organ in the human. The human spleen, a collection of monocytes, B cells and also T cells, is large enough and has a long-lived population of stem cells with lifelong multi-generation regenerative potential. Thus, the spleen may be an ideal site for M. bovis to recruit abundant stem cells for reprogramming glycolysis^28–30.

A BCG effect restricted to the spleen’s monocytes is too limited to explain systemic lowering of glucose levels by BCG. The spleen’s monocytes represent only 2–8% of its lymphoid population while roughly 80% are lymphocytes, assuming that the spleen’s composition is similar to the blood’s. Lymphocyte participation must be significant, considering that ¹⁸F-FDG labeling is uniformly distributed across the entire spleen and not a punctate distribution of a few scattered monocytes (Fig. 2). By contrast, the spleen’s monocyte population is limited to the subcapsular red pulp³¹. Lymphocyte participation is also consistent with our earlier research showing that the BCG-induced shift to aerobic glycolysis is seen across all lymphoid cell types, including lymphocytes, monocytes and PBLs^{12,16,17,20,21}.

How does the spleen contribute to regulation of glucose levels in the blood, as reflected by nearly normal HbA1c. The spleen’s lymphoid cells have glucose uptake and glucose is the preferred nutrient over fatty acid metabolism³². The human spleen’s regulatory role in glucose metabolism is also supported by the surgical literature where it is well known that non-diabetic humans undergoing a splenectomy, usually for trauma, have a > 50% chance of developing diabetes after the surgery^33–47. The splenectomy literature thus fits well with our ¹⁸F-FDG PET/CT maps showing the spleen to be a hot-spot for glucose uptake after BCG treatment in T1D. We also note that when spontaneously diabetic NOD mice become diabetic their spleens shrink to a tiny size, whereas the spleen size remains normal in BCG-treated NOD⁴³.

It should be recognized that previous research with shorter sampling times after BCG vaccination in the mouse and human can locate BCG or its functional effects in the bone marrow. We too saw bone marrow metabolic niches for BCG, or BCG functional effects, but they were only at early monitoring points. One prior study in healthy humans showed by direct bone marrow sampling at 13 weeks after a single BCG vaccine, functional effects of the BCG organism in the human bone marrow⁴⁸. In mice administered subcutaneous (SC) or intravenous (IV), the IV BCG by culture documentation reached the bone marrow at week 1, persisted to 21 weeks, and then disappeared by 30 weeks⁴⁹. Using our methods of intradermal (ID) BCG, we detected BCG colonies in the bone marrow but only from week 1 through week 8 in the mouse.

A primate study of intravenous (IV) BCG administration provides surprisingly similar results to human and mouse findings reported here for where accelerated glucose utilizations are occurring. Primates were studied by the identical method of PET/CT scans but at short-term 4-week harvests instead of humans in this study monitored for 2 years. Mapping showed increased glucose uptake in lung and spleen, but no signal in the bone marrow. These findings were confirmed with direct culture methods in the primate like we performed in the mouse. While we do not see these effects as early, presumably because the primate study used an intravenous route for BCG, the findings provide further evidence of BCG’s preference for the lung and the splenic niche⁵⁰.

There is mechanistic appeal of the bone marrow as the niche for the BCG microroganism or for BCG systemic reprogramming effects because the bone marrow is the traditional source of hematopoeitic stem cells. The literature supports BCG’s reset of so many different lineages of lymphoid cells (e.g., T cells, monocytes, B cells, PBLs), the data supports perhaps BCG in vivo could reset stem cells for such multi-lineage effects. But the spleen is often overlooked as a human and murine organ source of stem cells with mutli-lineage potential for parenchymal as well as lymphoid tissues^29,30,43,51.

Although not statistically significant, we showed increased uptake trends in other human lymphoid organs. We found both in the blood in the descending aorta and the lymphoid tissue in the bone marrow increased uptake, but these utilization signals were transient and not as robust as in the spleen. The liver also showed a non-significant trend for increased utilization in our trial. We do not know if the uptake was by lymphoid cells in the liver or by the liver’s parenchymal cells. A metabolic study in patients with type 2 diabetes found that the liver can also contribute to glucose regulation.

Bacterial pathogens have evolved to inhabit unique niches in the human host¹. This study is a serial and systemic evaluation of the niches for the Mycobacterium bovis organism as the BCG vaccine in the human and murine host. Also, the TB microbe after natural inhalation exposure usually prefers the lung and is most often restricted to the lung granuloma. Active TB with its devastating clinical sequelae can be a systemic disease. Although both Mycobacterium tuberculosis (Mtb) and Mycobacterium bovis (MB) are similar organisms, no persistent lung signal was prominent in mice or humans reported here. This suggests a difference in this microorganism’ selection of niches, although the difference in route of exposure is still a consideration that cannot be ruled out.

With systemic intradermal and repeat BCG vaccinations, what is the speed of systemic changes in metabolism, what organs help to regulate blood sugars, and what genes are regulated? From a clinical view point the lowering of blood sugars from BCG vaccines takes 2–3 years and this is again what we observe in this study of blood sugars monitored by HbA1c values after BCG vaccinations in this new clinical trial data from T1D^16,17,26. Our new confirmatory observations of this study are also buttressed by the in vitro analysis of lymphoid subpopulation that begin to show accelerated or restored to near normal glucose uptake in culture. This new data also shows 2-year gradual time course of organ specific increases in ¹⁸F-FDG PET/CT scans over which time BCG-induced changes in glycolysis are regulated by methylation of genes in critical glucose regulating pathways^12,17. A prior study has also identified, after BCG multi-dose treatment of T1D a BCG-driven re-methylation of histones, and the activation of the mTOR pathway for facilitated glucose uptake respectively¹². The BCG effect at the gene level was confirmed by reciprocal mRNA changes and again the 2–3-year time course. The DDIT4 gene with known inhibitory role of mTOR was remethylated after BCG, a step likely to allow improved glucose uptake. BCG-driven methylation of KDM2B’s site halted augmented histone activity, a step known to allow cytokine activation and increased glycolysis. In humans, the effects of BCG from lymphoid cells obtained from peripheral blood draws show metabolic changes in both T cells and monocytes. After 3 years of treatment BCG is associated with increased expression of Myc, a key transcription factor that controls aerobic glycolysis, Myc controlled glutaminolysis and polyamine synthesis genes¹⁶. BCG vaccinations through Myc are also associated with the HIF-1/mTOR pathway but not the Akt pathway to promote increased glucose uptake. The actions of Myc on HIF-1a are also well described in the metabolic effects of Myc in humans in the lung granuloma showing once again the similarities between systemic BCG instructions and localized MtB metabolism. Furthermore, infection with MtB also causes activation of host hypoxia induced factor (HIF) in humans⁵². Much research over the years also implicates HIF-1a as responsible for the Warburg effect and this promotes glycolytic flux by activating various genes encoding the main Warburg effect enzymes. This lymphoid metabolic reprogramming of host lymphoid cells is seen in both T cells and monocytes^17,26. Studies with various gram-positive and gram-negative bacteria show that increased glucose uptake is a general effect of bacterial infections and is not limited to just BCG⁵⁵. However, BCG is particularly suitable to elicit this effect because BCG infections are long lasting and have an established safety profile as a vaccine.

We do not yet know if the metabolic effects of BCG are largely due to BCG resetting the immune system through methylation changes, or if the organism needs to persist in the host for these long duration effects on metabolism. Perhaps both can occur. Routinely obtained blood cultures in the clinic do not detect the BCG organism. BCG organisms require long growth observation periods of months and special types of culture plates. When these special culture conditions are set up, many human examples exist to support the BCG microorganism can persist in humans for decades. It is known when massive doses of BCG are administered as a 6-cycle therapy for bladder cancer, BCG can appear transiently in the bone marrow. For instance granulomatous inflammation was observed in the bone marrow of a 68-year old man 2 years after BCG instillation therapy⁵³. Similar rare instances where the proper culture of Mycobacterium bovis was undertaken, with BCG also having been observed in the liver, aorta, and spleen, often becoming clinically significant with immunosuppression^54–56.

Study limitations

Although this human clinical trial is multi-year and patients had an excellent clinical response to repeat BCG vaccinations on blood sugar regulation, this trial only involved 6 subjects. A larger ¹⁸F-FDG PET/CT study might have revealed the nuances of additional and perhaps lower-level glucose uptake in additional organs. Importantly, the BCG studies of glycolysis showed that the effect takes multiple years to occur, so this two-year study may not have been long enough. Additionally, ¹⁸F-FDG PET/CT scanning has a resolution that cannot detect smaller organ regions such as small blood vessels within organs that might harbor concentrations of lymphoid cells or even lymph nodes within organs. As detailed in the methods section, we tried to rule out this error by taking the mean measurement from the volume of a sphere for organs studied (except the brain and lungs which were segmented automatically) to prevent bias. Lastly BCG has been shown to have beneficial blood sugar effects extending beyond 8 years of time from the first vaccine¹⁷. We do observe, in the first year after vaccinations, that the BCG metabolic niche does change and evolve. We do not yet know if the BCG metabolic niches change over the passage of greater monitoring times, and if the splenic two-year metabolic niche persists over time or moves to another human organ.

The PET scan required irradiation due to the radioactive ¹⁸F-FDG. As a result, the ethics committee did not approve inclusion of a T1D control group that was not treated with the possible benefit of BCG but underwent PET scans since this group would carry the risk of radiation exposure without the benefit of potential improvement in blood sugars. As a work-around, since the first PET/CT scan was before BCG treatment, each patient served as their own control with subsequent scan results compared to the baseline scan. However, since there is no placebo arm, we cannot exclude the possibility of confounding factors that might change glucose uptake, such as infections unrelated to BCG that might cause increased glucose uptake in the spleen.

In this study we perform in vitro glucose uptake studies on white blood cells from humans treated with BCG but did not perform the equivalent studies in the mice. We have reported before on 2-NBDG uptake studies in bone marrow cells from in vivo BCG treated and untreated Balb/c, db/db and NOD mice and therefore did not repeat the same studies again²⁶. We found statistically significant increases after BCG treatment in bone marrow cells of the diabetic NOD and db/db mouse models, but only a non-significant increase in the non-diabetic Balb/c. Also, for the presented studies we used normal mice that did not have the underlying disease defects in baseline low aerobic glycolysis to potentiate the change after treatment.

Another limitation is that we were only able to study in vitro the BCG effects on glucose uptake in monocytes. From past studies the white blood cell defect of T1D is within all lineages of cells. Due to limitations in the amount of blood we were allowed to collect according to our clinical trial protocol, we had to limit ourselves to only isolating monocytes and could not also collect and study in vitro all white blood cell lineages. Past studies of type 1 diabetics show the defect in decreased aerobic glycolysis is not confined to monocytes; CD4 and CD8 T cells and PBLs were studied in other trials. From these other studies we know that similar increases in glucose uptake after BCG treatment also hold for T cells and PBLs, both in vivo and in vitro^12,16,17,26.

Some of the literature on glucose utilization refers to glycolysis studies in macrophages. In a strict definition, macrophages are monocytes that have migrated from the bloodstream into body tissues. Since we have isolated our cells from blood, we technically have monocytes and not macrophages. However, the macrophage literature still is relevant to the concept that increased glucose utilization after BCG treatment in vitro or in vivo is due to upregulation of aerobic glycolysis in the white blood cell lineages.

Human and murine studies implicate the spleen as the predominant metabolic niche for humans and organ niche in the BCG vaccine’s functional improvement of glucose metabolism. The spleen is significant because it is one of the largest collections of lymphocytes to explain BCG’s systemic and therapeutic lowering of blood sugar to near normal levels in type 1 diabetes. The studies reported here suggest that, once triggered by BCG, the lymphoid system is a new regulator of blood sugar with implications beyond type 1 diabetes.

Author contributions

Conceptualization, DLF, WMK; PET/CT data collection/analysis, HFD; PET/CT SUV collection/analysis guidance; JFF, CC, JCP; PET/CT patient appointments, ERH, GEW, NCN; PET/CT brain SUV collection/analysis guidance, JFF; PET/CT lung SUV collection/analysis guidance, CC; Mouse BCG colony data collection, TGL; Mouse BCG colony data analysis, HFD, Writing original draft, HFD; Writing-review, DLF, WMK, HFD, JFF, CC, JCP; statistical analysis, HZ, HFD.

Funding

The Iacocca Foundation.

Data availability

Data can be made available upon request: please email dfaustman@mgh.harvard.edu.

Competing interests

The authors declare that they have no known competing financial interests regarding glucose utilizations localized to the spleen, ownership of the study drug or sponsored research that could have appeared to influence the work reported here.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-67905-x.