Autologous adipose-derived mesenchymal stromal cells for chronic traumatic brain injury

Abstract

Secondary injury from traumatic brain injury (TBI) involves an inflammatory process that leads to chronic microglial activation, causing loss of gray and white matter associated with worse functional outcomes. Pre-clinical studies have shown mesenchymal stromal cell (MSC) treatment dampens microglial activation and improves TBI outcomes. Our study aimed to assess the impact of autologous adipose-derived MSC treatment in patients with chronic TBI using advanced imaging, functional, and neurocognitive outcome measures.

Our Phase 1/2a study analyzed the safety and efficacy of autologous Hope Biosciences adipose-derived MSC (HB-adMSC) treatment in 24 patients with chronic TBI. Outcome measures included functional, neuropsychological, and psychometric testing; multimodal diffusion tensor MRI; and PET using the radiotracer ligand [¹¹C]ER176 as a measure of brain immune cell density and activation. There were no treatment-related adverse events. MRI imaging analysis showed at 6 months after treatment, elevated mean diffusivity volumes were significantly reduced bilaterally in the hippocampus (mean decrease 32.35, 95% confidence interval [CI] −54.61 to −10.09; P=0.0072) and amygdala (mean decrease 21.88, 95% CI −41.11 to −2.65; P=0.030). No significant volumetric changes in the hippocampus (P=0.099) or amygdala (P=0.281) were observed. PET results showed that HB-adMSC treatment induced significant reductions in brain immune cell activation in the right caudate (x,y,z: 12,2,7; T₂₀ = 4.2; P < 0.0002) extending into the right ventral anterior thalamus (x,y,z: 13,−2,11; T₂₀ = 3.7; P < 0.00014) and right nucleus accumbens (x,y,z: 8,9,−3; T₂₀ = 12.6; P < 0.009), and right parahippocampal gyrus (x,y,z: 29,−40,−7; T₂₀ = 4.1; P< 0.0002). Glasgow Outcome Scale-Extended scores improved at 6 and 12 months, with more patients functioning as moderate (55%, 65%) than severe (45%, 35%). Additionally, reductions in depression (P=0.026), fatigue (P=0.008), and pain (P=0.007) 6 months after treatment were also observed. Within-participant multimodal MRI analyses indicated significant interactions between changes in neuropsychological assessments of anxiety and depression and changes in elevated mean diffusivity levels in the amygdala and hippocampus, respectively.

Our study demonstrates autologous adipose-derived MSCs for chronic TBI are safe, have clinically relevant treatment effect sizes in functional and neurocognitive outcome measures, yield significant improvements in specified measures of microstructural integrity in brain-based behavior relations, and reduce microglial activation in regions corresponding to pain, fatigue, and depression. These data provide quantitative justification for the sample size of next-phase clinical trials using either functional outcomes or surrogate imaging outcomes.

Introduction

Traumatic brain injury (TBI) results in a secondary inflammatory process that peaks 72-96 hours after the kinetic event and manifests by acute inflammatory cytokine release, inflammatory cell infiltration, and blood-brain barrier dysfunction. These events then lead to vasogenic and cytotoxic edema (with potential hemorrhage expansion) that transitions to a chronic inflammatory response corresponding with loss of gray and white matter over ensuing months.^1–3 Chronic neuroimmune activation is thought to contribute to neurodegeneration/tissue loss and poor outcomes in TBI patients. Ramlackhansingh et al first demonstrated a chronic microglial activation state after severe TBI where degree of chronic microglial activation correlated inversely with functional outcomes.⁴ Furthermore, data from our group and others suggest that microglial activation is a viable therapeutic target to improve TBI outcomes.^5,6

MSCs (and other cell types) are potent immunomodulatory agents, downregulating “pro-inflammatory” activity induced by activated microglia.^7–9 We and others have demonstrated in pre-clinical and clinical TBI studies that intravenously infused mesenchymal stromal cells (MSCs) and bone marrow mononuclear cells (a mixed population containing MSCs) can dampen microglial activation and improve outcomes.^7,10,11 We have also noted that MSCs modulate microglia in stroke, which shares pathophysiological characteristics with TBI.^8,12

Furthermore, we have observed volumetric loss in serial MRI following TBI.^13,14 Enhanced activation of brain immune cells (microglia, astrocytes, etc.) has been known to occur following TBI. Using PET and a radiotracer with specificity to the mitochondrial translocation protein (TSPO), [¹¹C]ER176,^15,16 density of brain immune activation can be estimated. Given the potential for MSCs to mitigate brain immune activation, we hypothesized that MSC treatment would reduce density of brain immune cell activation and potentially alleviate macroscopic volumetric changes and regional microstructural consequences of inflammation (e.g., changes in diffusivity).

We studied the impact of autologous adipose-derived MSC treatment in 24 patients with a history of chronic TBI using advanced imaging and functional/neurocognitive outcome measures to determine effect sizes that could guide future trial sample sizes. The objective of our study was to establish safety and investigate potential treatment effect of intravenous infusion of HB-adMSC on brain structure, neurocognitive/functional outcomes, and neuroinflammation after TBI in adults. The primary outcome measures were (1) determine the safety of HB-adMSCs as determined by infusional toxicity of the cell product; (2) obtain treatment effect estimates of HB-adMSC infusion on global gray and/or white matter as well as measures of microstructural integrity in regions of interest, as measured by fractional anisotropy and mean diffusivity; and (3) evaluate the neuroimmune response using advanced brain neuroimmune imaging techniques with [¹¹C]ER176 PET. Outcomes (including imaging, neuropsychological, and neurocognitive measures) were investigated before and after HB-adMSC treatment.

Patients and methods

Study Design

We conducted a Phase 1/2a study analyzing safety (infusional toxicity and venous thromboembolism [VTE]), biological target (density of brain immune activation estimated by PET and microstructural imaging), and clinical efficacy (impact on accepted outcome measures) of HB-adMSC infusion with a single-arm, non-randomized, pre-test, post-test design with longitudinal follow-up. Ethics approval was obtained from the Committee for the Protection of Human Subjects at The University of Texas Health Science Center at Houston (HSC-MS-19-0685), Memorial Hermann Hospital System, and Houston Methodist. Supplemental Figure 1 shows longitudinal study flow and trial design.

Data and Safety Monitoring and Trial Registry

A Data and Safety Monitoring Board (DSMB) was utilized. A medical safety monitor (MSM) reviewed post-infusion summaries for each participant. Juno Research provided independent medical monitoring and worked with the MSM and DSMB to ensure safety of study participants. This study was conducted under Federal Drug Administration Investigational New Drug (IND) Application 19204 and registered with ClinicalTrials.gov (NCT04063215).

Participants

Participants were recruited from a combination of investigator clinic population, 2 previous IRB-approved studies conducted at our site (HSC-MS-16-0283, HSC-MS-15-0744), and self-referral. Adults 18-55 years with TBIs from closed head trauma with onset or diagnosis of injury > 6 months before study enrollment were included. Participants were required to have Glasgow Outcome Scale-Extended (GOS-E) scores > 2 and ≤ 6 for inclusion. Exclusion criteria included normal brain imaging (CT/MRI), spinal deformity or complete spinal cord injury, and bleeding disorders or contraindications to low molecular weight heparin (enoxaparin). Full eligibility criteria are in Supplemental Table 1.

The principal investigator made final determination of participant eligibility before enrollment. Written informed consent was obtained from participants or legally authorized representative(s) after nature and possible consequences of the study were explained according to the Declaration of Helsinki.

Interventions

MSC Harvest, Processing, and Infusion

Adipose Tissue Extraction and Processing

Adipose tissue extraction and processing were performed under IND 19204. Adipose tissue was extracted from the participant’s abdomen, flank, or hip area under local anesthesia by licensed medical providers trained in liposuction techniques. Adipose tissue (2-9 mL of emulsified fat) was subsequently treated with collagenase to separate stromal vascular fraction (SVF) for production of HB-adMSCs. SVF was plated in Hope Biosciences (HB)-103 medium, with resulting adherent cells being further expanded in HB-101 medium to establish a P0 culture. Cells were cryopreserved at passages 0, 1, and 2 to create a complete bank of each participant. For infusions, passage 2 cells were thawed and cultured to passage 4. Subsequently, 2.0x10⁸ HB-adMSCs were freshly harvested from passage 4 cultures, packaged in 20 mL sterile 0.9% saline for each dose, and administered within 96 hours of packaging. Product manufacturing was repeated for 3 infusions per participant. To ensure safety and efficacy of investigational drug, final product release criteria were set as follows: viability (≥ 70%), appearance (opaque white to faint yellow with no settlement), USP71 sterility (no organism seen), mycoplasma (negative), endotoxin (≤ 10 EU/mL), Gram stain (no organism seen), and identity/purity by MSC-defining surface markers CD73 & CD29 (> 75%) and CD31 & CD45 (< 5%). All products provided successfully met cGMP-compliant quality control standards. Supplemental Table 2 quality control metrics of each dose.

Cell Infusion

Participants or their caregivers administered a subcutaneous dose of 40 mg enoxaparin daily beginning the day before each HB-adMSC infusion and continued doses for 5 days after infusion (7 days total) to mitigate risk of VTE.^17,18 At the infusion visit, participants were given diphenhydramine and methylprednisolone before infusion (25 mg for participants weighing < 100 kg and 50 mg for those ≥ 100 kg for both medications).

Participants received a total of 3 infusions of 2x10⁸ live cells per dose administered over 6 weeks with 14 (+/−3) day intervals between infusions. The total volume for each infusion was approximately 250 mL, infused over 1 hour.

Safety Monitoring for Infusional Toxicity

Participants’ vital signs were continuously monitored throughout the infusion. Participants were also assessed for infusion-related toxicity for the first 4 hours (pulmonary or anaphylactic reactions) following infusion and by telephone 24 hours later.

Multimodal Neuroimaging

PET and MRI brain imaging were completed prior to first infusion (baseline) and 6 months after treatment. All PET and MRI imaging was done within the Translational Imaging core facility at the Houston Methodist Research Institute, Houston, TX.

MRI Acquisition

MRIs were acquired with a 20ch head coil on a 3T Siemens Vida platform (Siemens Healthineers AG). The same multimodal, multi-weighted neuroimaging protocol was utilized throughout the study period. The following 3D structural MRI (sMRI) sequences were acquired in the sagittal plane with FoV:256x256, 1mm³ isotropic voxel dimensions, and generalized autocalibrating partially parallel acquisition acceleration factor of 2: (1) T1w magnetization-prepared rapid gradient echo (repetition time [TR]: 2800 ms; echo time [TE]: 2.98 ms; flip angle: 8°); (2) T2w space (TR: 2500 ms; TE: 202 ms; flip angle: T2 variable); and (3) T2-weighted fluid-attenuated inversion recovery (FLAIR) (TR: 4800 ms; TE: 202 ms; inversion time [TI]: 1650 ms; flip angle: T2 variable). Additionally, a susceptibility weighted imaging sequence was acquired in the axial plane (voxel dimensions: 0.7x0.7x1 mm³; TR: 17 ms; TE: 9.12 ms; flip angle: 10°) to identify and localize bleeds. Finally, an axial 30-direction echo planar imaging (EPI)-based spin-echo diffusion MRI (dMRI) sequence¹⁹ (2.2 mm³ isotropic voxel dimensions; TR: 8800 ms; TE: 78 ms; b=1000 s/mm²; simultaneous multislice [SMS] and acceleration=2) was acquired for diffusion tensor fitting and reconstruction of common scalar maps, including mean diffusivity (MD). A pair of reverse phase-encode b=0 s/mm² (Anterior-to-Posterior and Posterior-to-Anterior) maps with matching geometry were acquired for post-processing correction of eddy currents, susceptibility, and motion distortions. Completing the entire multimodal, multi-weighted structural imaging protocol required 45 minutes of imaging time.

[¹¹C]ER176 PET Acquisition

PET images were acquired via with a Siemens Biograph mCT PET/CT scanner (Siemens Healthineers AG), with in-line CT-based attenuation correction. Output data was acquired in the form of PET standardized uptake volume (SUV) image maps: 1 dynamic TSPO_SUV PET image for each 90-minute PET scan.

Participants received a single bolus intravenous injection of [¹¹C]ER176, a highly selective TSPO radiotracer, as a measure of brain immune cell activation density immediately prior to PET scan.^15,16 [¹¹C]ER176 is relatively insensitive to the common single nucleotide polymorphism rs6971 compared to prior-generation TSPO radiotracers.²⁰ The dose of [¹¹C]ER176 varied up to a maximum of 12.0 mCi depending on the availability and radiotracer production yield from the Cyclotron Core facility within the Houston Methodist Research Institute. Quality control measures ensured that [¹¹C]ER176 radiotracer produced at our facility is of high radiochemical purity (> 95%) and high molar activity (e.g., > 121 GBq/μmol). The chosen PET data processing approach (e.g., SUV ratio) accounts for inter-individual and intraindividual differences in radiotracer dose.

Participants’ blood pressure, heart rate, respiratory rate, and blood oxygen saturation were monitored at specific timepoints for safety. Participants were asked to lie quietly in the scanner, without head movement, for the total 90-minute imaging time. Following completion of the PET scan, participants were monitored for adverse events by nursing staff for 1 hour and contacted by staff 24-48 hours later to assess for clinical complications, with plans to provide appropriate medical intervention or referral as needed.

Processing of MRI Data

High-resolution T2w and T1w images were co-registered (within participant) using FSL’s (v6.0.3)²¹ linear registration tool FLIRT. The resultant image (T1_2_T2) was used to extract the brain from the skull with FSL’s BET tool and then segmented into 3 tissue classes (gray matter, white matter, CSF) using FSL’s FAST tool. The CSF-based tissue class was excluded from analyses beyond volumetric quantification. Additionally, the T2w-FLAIR images were co-registered (within participant) to the T2w images using FSL’s FLIRT tool (FLAIR_2_T2).

Pre-processing steps for dMRI data included FSL’s topup and eddy functions to correct for geometric and motion-related distortions. Subsequently, FSL’s tensor fitting function (DTIfit) was used to derive common scalar maps, including MD. Given the image properties shared by the T2w and b=0 volumes in the dMRI sequence, dMRI scalar maps were readily transformed to T2w space using FSL’s FLIRT tool.

The dMRI sequence data were used to investigate changes in MD, both globally (e.g., whole brain gray matter and white matter) as well as regionally in the corpus callosum, amygdala, and hippocampus. We followed the classification schema developed by Sener²² from over 130 MRIs to quantify volumetric changes in the amount of restricted diffusion (LowMD values < 0.60x10⁻³ mm²/s) and increased diffusion (SupraMD values ranging from 1.28 – 2.20x10⁻³ mm²/s). Based on Sener’s classification schema, MD values for normal-appearing white matter range between 0.60 - 1.05x10⁻³ mm²/s, and MD values for CSF range between 2.40 - 4.40x10⁻³ mm²/s.

Specific structurally defined regions of interest (e.g., hippocampus, amygdala, corpus callosum) commonly impacted by TBI were parcellated from the rest of the brain using Freesurfer version 7.2.0.²³ Thus, binary mask labels from each participant’s Freesurfer processing pipeline were transformed to corresponding native dMRI space for volumetric quantification of LowMD and SupraMD voxels in each structural label for further analyses.

Processing of PET Data

All PET images were co-registered and smoothed using a 2 mm Gaussian filter as previously described.^24,25 Parametric processing of TSPO_SUV images was completed using PMOD software v4.3 (Brucker). Following correction for motion artifact, an average image was calculated for each scan across time where the radiotracer achieved steady state. The average PET image was subsequently co-registered with the participant’s structural T1w MRI image. The T1w MRI was normalized to the Montreal National Institute (MNI) template brain space using a 3mm kernel, with the resulting transformation applied to the average PET image, then transforming the image to MNI template space.²⁴ A SUV ratio (SUVR) data processing approach was applied to each participant’s transformed TSPO_SUV images using SUV values averaged across their entire cerebellum as a pseudo-reference region, yielding distinct TSPOSUVR image maps for each participant and each timepoint.²⁶

Functional, Neuropsychological, and Psychometric Assessments

Neuropsychological assessments were completed in clinic before first infusion (baseline) and 6 and 12 months after treatment. Outcomes were assessed via structured interview, self-report, and direct testing. Measures were selected in accordance with recommendations supported by the National Institutes of Health (NIH) and Department of Defense for common outcome measures for research and clinical trials in adults with TBI.^3,27 Structured functional outcome interviews completed by participants included Glasgow Outcome Scale - Extended (GOS-E)²⁸ and Disability Rating Scale (DRS),²⁹ with input from family when appropriate. Self-report assessments completed by participants included the TBI Quality of Life Questionnaires (TBI-QOL SF)³⁰ and Brief Symptom Inventory 18 (BSI-18).³¹ The Galveston Orientation and Amnesia Test³² was administered to ensure that participants were oriented. Direct testing assessments included select NIH Toolbox cognitive³³ and motor³⁴ measures, Rey Auditory Verbal Learning Test,³⁵ Wechsler Adult Intelligence Scale (WAIS) Processing Speed,³⁶ and the Controlled Oral Word Association Test.³⁷

For examination in relation to PET imaging, participants completed 3 additional psychometric measures at the time of imaging, including both prior to HB-adMSC treatment and 6 months after. Psychometric measures included the Beck Depression Inventory (BDI),^38,39 Brief Fatigue Inventory (BFI),⁴⁰ and Brief Pain Inventory (BPI). Presence of clinically meaningful symptoms of depression, fatigue, and pain at baseline were defined as BDI > 10, BFI ≥ 1, and BPI ≥ 1, respectively.

Statistical Analysis

Sample Size

The primary objective of our pilot study was to estimate treatment effects and variability in outcome measures to inform design and sample size calculations for a Phase 3 trial. This approach aligns with the Consolidated Standards of Reporting Trials extension for pilot and feasibility studies, which recognizes that pilot studies can serve to provide preliminary data on treatment effects without being powered for formal hypothesis testing.³¹ A sample size of 24 participants was selected to balance the need for robust estimation with the practical constraints of resources, time, and feasibility.

Statistical Approach

Volumetric, DT-MRI, neuropsychological, and functional outcome measures were analyzed with general linear mixed models, including time of assessment (baseline and 6 and/or 12 months after treatment) and age at study start as covariates with random intercepts for participants to account for within-participant correlation. GOS-E was analyzed using a proportional odds logistic model. All other outcomes were analyzed with linear regression models. Odds ratios and mean differences with 95% confidence intervals (CI) were determined for all outcomes at 6 and 12 months after treatment. Effect sizes were calculated using Cohen’s d to estimate the magnitude of change from baseline to 12 months after treatment (0.2 =small effect, 0.5 = medium effect, 0.8 = large effect).

We examined the relationships between medial temporal lobe structures (hippocampus and amygdala) and validated TBI behavioral indicators of depression and anxiety. Baseline anxiety and depression levels were categorized using the BSI-18 subscales, with T-scores above 60 indicating high levels and scores of 60 or below indicating low levels. Repeated measures ANOVA models were applied for each subscale, incorporating time, baseline group, and interaction between time and baseline group to assess changes in MRI-derived thresholded MD volumes in the amygdala and hippocampus. Additionally, we also evaluated the relationship between changes in volumes of thresholded MD voxels in the insula and T-scores from the BSI-18 subscale for somatization dichotomized into high or low, using a T-score of 60 as cutpoint. All MRI analyses were conducted in R software v4.4.1⁴¹ and MATLAB 2022b (Mathworks). Mean differences and 95% CIs are reported for each model.

Using PET outcome data (e.g., TSPO_SUVR image maps), we examined whether HB-adMSC treatment reduced regional activated immune cell density by comparing PET data at baseline and 6 months after treatment. A voxel-wise repeated measures analysis of covariance (ANCOVA) model was implemented to include 2 factors: time (baseline and follow-up) and clinical symptoms (6 levels based on the presence or absence of baseline depression (BDI), fatigue (BFI), or pain (BPI)). Post hoc paired t tests assessed reductions from baseline to 6 months after treatment in participants with clinically significant symptoms of depression, fatigue, or pain at baseline. All PET analyses were completed using Statistical Parametric Mapping (SPM12: Wellcome Trust) software and MATLAB 2022b (Mathworks).

Results

Our single-center, non-randomized study was conducted between January 1, 2020, and October 22, 2023. In total, 24 patients were enrolled in the study (Supplemental Figure 2). Patient characteristics are listed in Supplemental Table 3.

Safety

There were no serious adverse events and no infusion-related or [¹¹C]ER176 radiotracer-related adverse events, including no thromboembolic events (Supplemental Table 4).

MRI Structural Measures

In total, 24 participants completed baseline imaging, and 20 participants completed 6-month follow-up imaging. Given the within-participant design, the 4 participants who did not complete follow-up imaging were excluded from analyses. Estimated differences from baseline to 6 months after treatment for volumetric and microstructural measures of global tissue classes, as well as regions of interest commonly investigated in TBI research, are shown in Table 1. Compared to baseline, at 6 months after treatment, SupraMD volumes were significantly reduced bilaterally in the hippocampus (mean decrease 32.35, 95% CI −54.61 to −10.09; P=0.0072) and amygdala (mean decrease 21.88, 95% CI −41.11 to −2.65; P=0.030). No significant volumetric changes in the hippocampus (P=0.099) or amygdala (P=0.281) were observed.

Table 1.

MRI Structural Measures

MRI Outcome Measure N=20	Baseline*	After Treatment (6 months)*	Mean Difference** (95% CI)	P (2-tailed)
SupraMD volume (mm3)
Global
White Matter	114,303 (29,555)	108,002 (25,752)	−1050.25 (−2958.69, 858.19)	0.27
Gray Matter	430,058 (45,653)	427,864 (54,441)	−365.57 (−1835.10, 1103.97)	0.61
Corpus Callosum	836 (274)	757 (258)	−13.17 (−33.32, 6.98)	0.19
Regional (bilateral)
Hippocampus (bilateral)	6,475 (1,071)	6,281 (1,082)	−32.35 (−54.61, −10.09)	0.0072a
Amygdala (bilateral)	1,926 (457)	1,794 (496)	−21.88 (−41.11, −2.65)	0.03a
Thalamus (bilateral)	4,952 (798)	4,700 (778)	−42.03 (−85.66, 1.61)	0.061
Caudate (bilateral)	3,760 (771)	3,523 (636)	−39.56 (−98.67, 19.56)	0.18
Putamen (bilateral)	901 (1,169)	951 (1,396)	8.25 (−26.01, 42.51)	0.63
Pallidum (bilateral)	450 (393)	459 (448)	1.47 (−14.84, 17.79)	0.85
Insula (Bilateral)	4,985 (1,164)	4,772 (967)	−35.42 (−95.21, 24.36)	0.24
Regional (by hemisphere)
Hippocampus (left)	3,180 (453)	3,140 (503)	−6.58 (−19.50, 6.34)	0.31
Hippocampus (right)	3,295 (663)	3,141 (646)	−25.77 (−46.25, −5.29)	0.018a
Amygdala (left)	870 (195)	813 (243)	−9.60 (−19.79, 0.59)	0.067
Amygdala (right)	1,055 (326)	982 (290)	−12.28 (−25.45, 0.88)	0.069
Thalamus (left)	2,522 (439)	2,363 (437)	−26.37 (−49.12, −3.61)	0.027a
Thalamus (right)	2,431 (388)	2,337 (422)	−15.66 (−42.09, 10.77)	0.24
Caudate (left)	1,786 (253)	1,679 (220)	−17.91 (−37.83, 2.01)	0.079
Caudate (right)	1,974 (578)	1,844 (483)	−21.65 (−62.91, 19.61)	0.29
Putamen (left)	224 (135)	199 (106)	−4.08 (−11.01, 2.84)	0.24
Putamen (right)	678 (1,174)	752 (1,401)	12.33 (−19.69, 44.36)	0.44
Pallidum (left)	179 (150)	159 (125)	−3.25 (−8.67, 2.17)	0.23
Pallidum (right)	272 (319)	300 (407)	4.73 (−9.43, 18.88)	0.5
Insula (left)	2,071 (575)	2,016 (422)	−9.23 (−51.62, 33.16)	0.66
Insula (right)	2,914 (748)	2,756 (688)	−26.19 (−70.22, 17.84)	0.24

^*Mean (SD)

^**Mean difference is calculated as the value at 6 months minus the value at baseline.

^aP < 0.05

MRI=magnetic resonance imaging; MD=mean diffusivity; CI=confidence interval

[¹¹C]ER176 PET Imaging

Four participants did not complete 6-month follow-up PET scans and were excluded from analysis. At baseline, 50% of participants (10/20) were clinically depressed, 80% (16/20) reported clinically meaningful symptoms of fatigue, and 35% (7/20) reported pain. At 6 months after treatment, significant reductions were observed from baseline in all 3 psychometric measures (depression: T=2.3, P=0.026; fatigue: T=2.7, P=0.008; and pain: T=3.4, P=0.007, Figure 1).

Figure 1. Depression symptoms at baseline and after treatment.

(A). Boxplot depicting participant Beck Depression Inventory (BDI) scores at baseline and 6 months after treatment. Blue represents participants with BDI scores ≤ 10 (non-clinical depression); red is participants with BDI scores >10 (clinically meaningful depression). Significant reductions in depression symptoms were observed with treatment (T=2.3, P=0.026). Fatigue symptoms at baseline and after treatment (B). Boxplot depicting participant Brief Fatigue Inventory (BFI) scores at baseline and 6 months after treatment. Blue represents participants with BFI scores < 1 (no symptoms); red is participants with BFI scores ≥1 (presence of fatigue symptoms). Significant reductions in fatigue were observed with treatment (T=2.7, P=0.008). Pain symptoms at baseline and after treatment (C). Boxplot depicting participant Brief Pain Inventory (BPI) scores at baseline and 6 months after treatment. Blue represents participants with BPI scores < 1 (no symptoms); red is participants with BPI scores ≥ 1 (presence of pain symptoms). Significant reductions in pain were observed with treatment (T=3.4, P=0.007). The central mark represents the median and the bottom and top edges of the box represent the 25^th and 75^th centiles, respectively, in all panels.

To provide clinical relevance to our baseline measure of brain immune activation density, we used 2-sample T testing to compare density of brain immune activation at baseline in participants reporting clinical symptoms (depression, fatigue, or pain) with participants who did not report clinical symptoms. At baseline, (1) participants with clinically relevant depressive symptoms had enhanced density of brain immune activation in the left and right caudate and left putamen (Figure 2A); (2) participants with fatigue symptoms had enhanced density of brain immune activation in the left and right putamen and left caudate, left anterior cingulate cortex, and both left and right hippocampus (Figure 2B); and (3) participants with pain symptoms had enhanced density of brain immune activation in the right anterior thalamus,left and right caudate, right nucleus accumbens, left anterior cingulate cortex, and left parahippocampal gyrus (Figure 2C).

Figure 2. Brain immune cell activation at baseline as measured by TSPO activity using PET.

Compared to participants without symptoms of depression, fatigue, and pain, at baseline, participants with clinically relevant depressive symptoms had enhanced density of brain immune activation in the left and right caudate and left putamen (A); participants with fatigue symptoms had enhanced density of brain immune activation in the left and right putamen and left caudate, left anterior cingulate cortex, and both left and right hippocampus (B); and participants with pain symptoms had enhanced density of brain immune activation in the right anterior thalamus, left and right caudate, right nucleus accumbens, left anterior cingulate cortex, and left parahippocampal gyrus (C). ACC=anterior cingulate cortex

Within-participant analyses compared density of brain immune cell activation 6 months after treatment to baseline (Figure 3A). Results showed that HB-adMSC treatment induced significant reductions in the right caudate (x,y,z: 12,2,7; T₂₀ = 4.2; P < 0.0002) extending into the right ventral anterior thalamus (x,y,z: 13,−2,11; T₂₀ = 3.7; P < 0.00014) and right nucleus accumbens (x,y,z: 8,9,−3; T₂₀ = 12.6; P < 0.009), and right parahippocampal gyrus (x,y,z: 29,−40,−7; T₂₀ = 4.1; P < 0.0002).

Figure 3. Brain immune cell activation after treatment as measured by TSPO activity using PET.

In participants who were depressed at baseline, 6 months after treatment, significant reductions in density of brain immune cell activation were observed in the right caudate (x,y,z: 11,4,5.5; T₂₀ = 4.0; P < 0.0003), extending into the right ventral anterior thalamus (x,y,z: 13,−2,10; T₂₀ = 3.7; P < 0.0005) (B). In participants who reported fatigue at baseline, 6 months after treatment, significant reductions in density of brain immune cell activation were observed in the right thalamus (x,y,z: 13,−3,11; T₂₀ = 2.3, P < 0.01) (C). In participants who reported pain at baseline, 6 months after treatment, significant reductions in density of brain immune cell activation were observed in right parahippocampus (x,y,z: 29,−41,−7.6; T₂₀ = 4.0; P < 0.0003), right caudate (x,y,z: 12.5,3.5,6; T₂₀ = 4.4; P < 0.00006) intersecting with right ventral anterior thalamus (x,y,z: 12.5,0.5,10; T₂₀ = 4.0; P < 0.0003), and right nucleus accumbens (x,y,z: 8,10,-1.5; T₂₀ = 3.4; P < 0.001) (D).

Impact of Intervention-Induced Reduction of Brain Immune Activation on Psychometric Measures

Post hoc within-participant analyses attached clinical relevance to intervention-induced reduction in density of brain immune activation (Figure 3). In participants who were depressed at baseline, 6 months after treatment, significant reductions in density of brain immune cell activation were observed in the right caudate (x,y,z: 11,4,5.5; T₂₀ = 4.0; P < 0.0003), extending into the right ventral anterior thalamus (x,y,z: 13,−2,10; T₂₀ = 3.7; P < 0.0005). In participants who reported fatigue at baseline, 6 months after treatment, significant reductions in density of brain immune cell activation were observed in the right thalamus (x,y,z: 13,−3,11; T₂₀ = 2.3, P < 0.01). In participants who reported pain at baseline, 6 months after treatment, significant reductions in density of brain immune cell activation were observed in right parahippocampus (x,y,z: 29,−41,−7.6; T₂₀ = 4.0; P < 0.0003), right caudate (x,y,z: 12.5,3.5,6; T₂₀ = 4.4; P < 0.00006) intersecting with right ventral anterior thalamus (x,y,z: 12.5,0.5,10; T₂₀ = 4.0; P < 0.0003), and right nucleus accumbens (x,y,z: 8,10,−1.5; T₂₀ = 3.4; P < 0.001).

Functional Outcomes

Table 2 details functional outcomes. Baseline GOS-E scores ranged from 3 to 6 and were dichotomized, with severe disability being 3 to 4 and moderate disability being 5 to 6. Baseline distribution was evenly split between both groups (50% [12/24] functioning as severe; 50% [12/24] functioning as moderate). More patients functioned as moderate (55% [11/21]) than severe (45% [9/21]) at 6 months, and most patients functioned as moderate (65% [13/21]) compared to severe (35% [7/21]) at 12 months. Odds of a good outcome increased over time, with the greatest improvement seen at 12 months after treatment (OR 1.71; 95% CI = 0.555.45; P=0.36). DRS, an assessment of impairment, decreased significantly at 6 months (mean decrease 0.38, 95% CI = 0.01-9.75; P=0.047) and 12 months (mean decrease 0.749, 95% CI = 0.379-1.12; P=0.00023), indicating functional improvement after treatment.

Table 2.

Functional and Neuropsychological Outcomes

Outcomes	Month 0 (N=24)	Month 6 (N=21)	Month 12 (N=21)	OR/Mean Difference*(95% CI) 6 months	P 6 months	OR/Mean Difference*(95% CI) 12 months	P 12 months	Effect Size
Functional Outcomes
Glasgow Outcome Scale-Expanded (GOS-E)				OR**: 1.33 (0.43, 4.19)	0.63	1.71 (0.55, 5.45)	0.36
3	7 (29%)	6 (30%)	5 (25%)
4	5 (21%)	3 (15%)	2 (10%)
5	2 (8.3%)	2 (10%)	3 (15%)
6	10 (42%)	9 (45%)	10 (50%)
Missing		1	1
Disability Rating Scale (DRS)	4.54 (3.89)	4.65 (4.07)	3.30 (2.00)	−0.38 (−0.75, −0.01)	0.047	−0.749 (−1.12, −0.379)	0.00023	0.19
Missing		1	1
Neuropsychological Outcomes
Galveston Orientation and Amnesia Test (GOAT)	91.74 (7.81)	95.89 (4.93)	96.10 (4.56)	4.86 (2.42, 7.29)	0.00027	4.9 (2.506, 7.294)	0.00019	0.63
Missing	1	2	1
Brief Symptoms Inventory 18 (BSI 18) SOM T Scores	56.52 (9.85)	52.42 (8.93)	51.60 (7.86)	−2.43 (−6.07, 1.18)	0.19	−3.7 (−7.26, −0.14)	0.043	0.38
Missing	1	2	1
Brief Symptoms Inventory 18 (BSI 18) DEP T Scores	57.17 (12.83)	52.58 (11.65)	51.00 (11.68)	−3.63 (−7.55, 0.26)	0.069	−6.15 (−9.983, −2.317)	0.0026	0.48
Missing	1	2	1
Brief Symptoms Inventory 18 (BSI 18) ANX T Scores	54.74 (10.85)	48.37 (9.06)	47.85 (8.71)	−4.43 (−8.12, −0.79)	0.020	−5.7 (−9.297, −2.103)	0.0029	0.53
Missing	1	2	1
Brief Symptoms Inventory 18 (BSI 18) GSI T Scores	57.48 (11.14)	52.00 (10.37)	50.60 (10.69)	−3.64 (−7.33, 0.02)	0.054	−5.85 (−9.456, −2.244)	0.0023	0.53
Missing	1	2	1
TBI - Quality of Life - Short Form: Cognition - General Concerns T Scores	37.29 (8.85)	41.85 (7.01)	42.72 (9.01)	3 (0.79, 5.22)	0.0097	4.51 (2.335, 6.685)	0.00017	0.51
Missing	1	2	1
TBI - Quality of Life - Short Form: Communication T Scores	47.76 (8.91)	52.18 (7.08)	52.39 (10.60)	2.58 (0.04, 5.14)	0.049	3.555 (1.054, 6.056)	0.0069	0.40
Missing	1	2	1
TBI - Quality of Life - Short Form: Independence T Scores	40.10 (8.82)	43.51 (8.82)	44.95 (7.70)	4.03 (1.21, 6.82)	0.0064	5.005 (2.251, 7.759)	0.00076	0.57
Missing	1	2	1
TBI - Quality of Life - Short Form: Mobility T Scores	39.36 (11.85)	40.12 (11.73)	40.73 (11.30)	3.13 (0.86, 5.38)	0.0084	2.72 (0.5, 4.94)	0.018	0.23
Missing	1	2	1
TBI - Quality of Life - Short Form: Satisfaction with SRA T Scores	41.96 (6.50)	44.53 (4.26)	44.35 (5.04)	2.22 (0.33, 4.13)	0.024	2.31 (0.445, 4.175)	0.017	0.36
Missing	1	2	1
TBI - Quality of Life - Short Form: Upper Extremity/ADL T Scores	37.20 (10.98)	36.65 (9.76)	38.74 (9.39)	1.48 (−0.35, 3.31)	0.11	2.69 (0.893, 4.487)	0.0046	0.24
Missing	1	2	1
Neurocognitive Measures
NIH Toolbox - Cognition: Pattern Comparison Processing Speed Test T Scores	30.65 (14.11)	35.42 (17.17)	39.68 (18.35)	5.69 (1.58, 9.77)	0.0081	8.505 (4.42, 12.606)	0.00017	0.60
Missing	1	2	2
NIH Toolbox - Cognition: Picture Vocabulary Test T Scores	41.96 (8.48)	NA (NA)	43.74 (9.56)	NA (NA, NA)		0.846 (−2.767, 4.515)	0.64	0.10
Missing	1	21	2
NIH Toolbox - Cognition: Dimensional Change Card Sort Test T Scores	36.65 (8.95)	38.05 (8.57)	40.47 (11.89)	2.59 (−1.52, 6.65)	0.21	4.318 (0.241, 8.399)	0.040	0.48
Missing	1	2	2
NIH Toolbox - Cognition: Franker Inhibitory Control and Attention Test T Scores	31.68 (10.83)	28.58 (9.81)	31.58 (10.04)	−0.86 (−2.52, 0.78)	0.30	0.789 (−0.83, 2.409)	0.33	0.07
Missing	2	2	2
NIH Toolbox - Cognition: List Sorting Working Memory Test T Scores	40.43 (9.61)	39.37 (9.33)	43.00 (10.96)	0.1 (−3.49, 3.66)	0.96	2.985 (−0.588, 6.562)	0.10	0.31
Missing	1	2	2
NIH Toolbox - Motor: 9-hole Pegboard Dexterity Test T Scores	26.81 (13.21)	27.61 (13.28)	27.94 (15.28)	3.34 (0.13, 6.53)	0.043	2.718 (−0.484, 5.917)	0.10	0.21
Missing	3	3	3
NIH Toolbox - Motor: 9-hole Pegboard Dexterity Test T Scores Non-Dominant Hands	26.21 (18.82)	20.20 (15.06)	25.09 (20.43)	2.39 (−1.73, 6.44)	0.24	3.79 (−0.152, 7.72)	0.061	0.20
Missing	10	11	10
Rey Auditory Verbal Learning Test (RAVLT) Delay T Scores	31.95 (16.24)	30.57 (19.02)	37.49 (19.20)	1.22 (−2.28, 4.71)	0.49	7.001 (3.506, 10.494)	0.00026	0.43
Missing	2	2	2
Rey Auditory Verbal Learning Test (RAVLT) Trials 1-5 T Scores	28.42 (14.76)	31.01 (16.34)	36.74 (16.94)	3.98 (−0.51, 8.46)	0.083	9.033 (4.55, 13.52)	0.00025	0.61
Missing	2	2	2
Wechsler Adult Intelligence Scale - IV: Coding Scaled Score	6.58 (3.29)	5.88 (3.18)	6.71 (3.69)	0.25 (−0.55, 1.05)	0.53	0.665 (−0.117, 1.445)	0.10	0.20
Missing	5	5	4
Wechsler Adult Intelligence Scale - IV: Symbol Search Scaled Score	6.24 (3.73)	6.06 (2.88)	6.83 (3.81)	0.83 (0.07, 1.59)	0.034	1.043 (0.284, 1.805)	0.0090	0.28
Missing	3	3	3
Wechsler Adult Intelligence Scale - IV: Processing Speed Composite Score	82.05 (17.62)	79.44 (15.12)	83.24 (19.23)	2.8 (−0.84, 6.41)	0.13	4.486 (0.938, 8.026)	0.016	0.25
Missing	5	5	4
Controlled Oral Word Association Test (COWAT) Total Scaled Score	8.88 (3.26)	8.69 (2.30)	10.00 (2.65)	−0.11 (−1.23, 0.97)	0.84	1.083 (−0.028, 2.162)	0.053	0.33
Missing	7	5	4

OR=odds ratio; CI=confidence interval; SOM=somatization; DEP=depression; ANX=anxiety; GSI=general symptom index; SRA=social roles and activities; ADL=activities of daily living;

^*Mean difference is calculated as the value at 6 or 12 months minus the value at baseline (0 month).

^**Odds ratio (OR) >1 indicates higher likelihood of being in higher categories comparing 6 or 12 months to baseline.

Neuropsychological and Psychometric Measures

Neuropsychological outcome measures are reported in Table 2. TBI-QOL scale quantified participant-reported physical health, emotional health, cognition, and social participation concerns after TBI. Twelve months after treatment, participants reported significant improvement in all areas, with medium effect sizes for independence (0.57) and cognition (0.51) and small effect sizes for communication and satisfaction with social roles and activities. Improvement was noted in mobility and upper extremity function (effect size range 0.23-0.40). The BSI-18 General Symptom Index indicated significant reduction in somatization, depression, and anxiety symptoms 12 months after treatment with a medium effect size (0.53).

Several neurocognitive measures also demonstrated significant improvement and moderate effect size from baseline to 12 months after treatment. In particular, the NIH Toolbox – Cognition: Pattern Comparison Processing Speed improved significantly over time, with a 12-month mean increase of 8.505 (95% CI = 4.42-12.606; P=0.00017; effect size = 0.60). Rey Auditory Verbal Learning Test scores improved significantly from baseline to 12 months for both initial learning (effect size = 0.61) and delayed recall (effect size = 0.43). Significant improvement and small effect sizes were also found for the NIH Toolbox Dimensional Change score (effect size = 0.48) and WAIS Processing Speed Index (effect size = 0.25).

For the BSI-18 anxiety measure, 12 of 20 participants were below 60, and 8 of 20 were above 60 at baseline (mean [SD] = 53.55 [10.73]; median=51). For the BSI-18 depression measure, 11 of 20 participants were below 60, and 9 of 20 were above 60 before treatment (mean [SD] = 57.15 [12.38]; median=57). For the BSI-18 somatization measure, 13 of 19 participants were below 60 and 6 of 19 were above 60 before treatment (mean [SD] = 54.8 [10.06]; median=58).

A significant time-by-BSI baseline anxiety group interaction was observed on the magnitude of reduction in volume of SupraMD volels in the amygdala after treatment (F=9.139, degrees of freedom [df]=18, P=0.007; least squares [LS] means for high subgroup: baseline = 1868.875, post-treatment = 1559.375; LS means for low subgroup: baseline = 1963.42, post-treatment = 1950.92) (Figure 4). A significant interaction was also observed between time and baseline BSI depression group (Figure 5) on the magnitude of reduction in volume of SupraMD voxels in the hippocampus after treatment (F=9.362, df=18, p=0.007; LS means for high subgroup: baseline = 6301.11, post-treatment = 5919.33; LS means for low subgroup: baseline = 6617.18, post-treatment = 6576.63).

Figure 4. Post-TBI high anxiety group shows reduced SupraMD volume in the amygdala after treatment as measured by MRI.

A significant time-by-BSI baseline anxiety group interaction was observed on the magnitude of reduction in volume of elevated MD in the amygdala after treatment (F=9.139, degrees of freedom [df]=18, P=0.007; least squares [LS] means for high subgroup: baseline = 1868.875, post-treatment = 1559.375; LS means for low subgroup: baseline = 1963.42, post-treatment = 1950.92). Baseline anxiety levels were categorized using the BSI-18 subscale for anxiety symptoms, with T-scores above 60 indicating high levels (blue; n=8) and scores of 60 or below indicating low levels (red; n=12).

Figure 5. Post-TBI high depression group shows reduced SupraMD volume in the hippocampus after treatment as measured by MRI.

A significant interaction was observed between time and baseline BSI depression group on the magnitude of reduction in volume of elevated MD in the hippocampus after treatment (F=9.362, df=18, p=0.007; LS means for high subgroup: baseline = 6301.11, post-treatment = 5919.33; LS means for low subgroup: baseline = 6617.18, post-treatment = 6576.63). Baseline depression levels were categorized using the BSI-18 subscale for depression symptoms, with T-scores above 60 indicating high levels (blue; n=9) and scores of 60 or below indicating low levels (red; n=11).

We also evaluated the relationship between the BSI-18 somatization measure and changes in volume of SupraMD voxels in the insula (Figure 6). For the BSI-18 somatization measure, 13 of 19 participants (1 did not answer) were below 60, and 6 of 19 were above 60 at baseline (mean [SD] = 54.8 [10.06]; median=58). Significant interaction with time and baseline BSI somatization score was observed on the magnitude of reduction in volume of SupraMD voxels in the insula post-treatment (F=5.321, df=17, p=0.034; LS means for high subgroup: baseline = 5876 and post-treatment = 5069.17; LS means for low subgroup: baseline = 4529.31 and post-treatment = 4554.08).

Figure 6. Post-TBI high somatization group shows reduced SupraMD volume in the insula after treatment as measured by MRI.

A significant interaction was observed between time and baseline BSI somatization group on the magnitude of reduction in volume of elevated MD in the insula post-treatment (F=5.321, df=17, p=0.034; LS means for high subgroup: baseline = 5876 and post-treatment = 5069.17; LS means for low subgroup: baseline = 4529.31 and post-treatment = 4554.08). Baseline somatization levels were characterized using the BSI-18 subscale for somatization symptoms, with T-scores above 60 indicating high levels (blue; n=6) and scores of 60 or below indicating low levels (red; n=13).

Discussion

Our data show that autologous, adipose-derived MSCs delivered intravenously in patients with chronic TBI (1) are safe and free of clinical venous thromboembolic events or infusional toxicity; (2) have clinically relevant treatment effect sizes in functional and neurocognitive outcome measures; (3) show improvement in microstructural measures (MD) in brain regions that correlate with anxiety and depression; and (4) reduce density of brain immune activation in regions component to circuitry shown to be dysregulated in depression, fatigue, and pain states.^25,42–46

Scientific Rationale/Biological Plausibility of Treatment with MSCs for Chronic TBI

Our pursuit of MSCs as a potential therapeutic for chronic TBI was first considered after Ramlackhansingh et al⁴ showed inverse correlations between microglial activation and neurocognitive outcomes in chronic TBI patients (up to months after index injury). These findings localized to the thalamus, a brain region frequently injured in TBI.⁴ Furthermore, notably, Loane and colleagues showed in animal models that chronic microglial activation promoted white matter loss.² Additionally, in small animals exposed to closed concussive injury, others have provided evidence of brain microglial activation with subsequent downregulation following intervention with cell-based therapies.^6,9,47 These findings suggested that density of brain immune activation is a potential target in TBI and that it is feasible to “move” the target with HB-adMSCs. While evidence suggests that degree of systemic immune signaling may be insufficient to promote an anti-inflammatory response to MSC intervention, “inflammatory licensing” is not required for MSC-induced modulation of the brain immune response to injury.⁴⁸

Safety Risk and Mitigation

Moll and colleagues alerted the field to the concept of immediate, complement-mediated reaction of blood to MSCs, leading to prothrombotic events and evaluation of factors related to hemocompatibility.⁴⁹ We and others linked this effect to cell surface expression of CD142 or tissue factor.^17,50 Using a range of assays, we have previously demonstrated that the tissue factor load on MSCs (sourced from different tissues) impacted procoagulant activity of the cells as determined by viscoelastic assays, thrombin generation, and other measurements.⁵¹ Importantly, these potentially deleterious effects could be completely blocked using clinically relevant doses of low molecular weight heparin.¹⁸ To that end, our clinical trials using intravenous MSCs incorporate use of enoxaparin in prophylactic, pre-treatment, and follow-on treatment dosing for 7 days. Applying this regimen, we have not encountered any clinical evidence of VTE-related complications or bleeding complications. As such, we believe this is an effective, low-risk tool to reduce likelihood of venous thromboembolic events.

Structural Imaging

We did 2 imaging correlations in our study with structural MRI and PET imaging. Contrary to one of our recent studies of acute, severe TBI treated with bone marrow mononuclear cells,¹³ our structural imaging analysis here showed no alteration in gray or white matter volumetrics. This is not surprising for 2 reasons: chronicity of the injuries and short interval between treatment and follow-up (6 months). In our previous study, we followed patients with serial volumetric imaging and noted that bone marrow mononuclear cell treatment acutely after a severe TBI showed mitigation of the known volume loss that occurs longitudinally after injury.¹³ In our current study, we anticipated that much of the volume loss associated with severe TBI has occurred, and our follow-up window would not detect a lower rate of ongoing loss.

However, there were microstructural metrics associated with functional outcomes that link to the pathophysiological aspect of ongoing neuroinflammation driven by chronically activated brain immune cells. Specifically, MD is a cellular measure of barrier function and global cellular functioning. Improvements in MD in the hippocampus and amygdala after MSC treatment that correlated with functional improvements suggest a structure-function relationship. Interestingly, the PET imaging did not co-localize to this region, suggesting another non-immune-mediated mechanism explains the MRI findings.

Density of Brain Immune Activation

In our PET imaging analysis, density of brain immune activation (enhanced in thalamic regions) was associated with worse outcomes following HB-adMSC treatment. We observed significant reductions in density of brain immune activation from baseline levels despite the presence of chronicity. HB-adMSC-induced reduction of density of brain immune activation overlaps substantially with prior findings from pre-clinical studies,^6,9,47 further suggesting density of brain immune activation may be a legitimate clinical target for TBI therapeutics.

Further, immune cells (e.g., microglia) have complex, dynamic, and varied roles in the human brain, with function extending beyond cellular homeostasis and specific phase of activation being both context and milieu-dependent,⁵² roles that do not readily lend to dichotomization (e.g., “up=bad, down=good”). In individuals with TBI, it seems likely that a so-called “pro-inflammatory” phenotype of microglia in one brain voxel may facilitate recovery (e.g., cleaning up cellular debris), whereas a “pro-inflammatory” phenotype in a contiguous voxel may have a deleterious effect (e.g., impacting synaptic plasticity in a behaviorally relevant manner). As such, estimating density of brain immune cell activation with TSPO imaging is more informative for our purposes than quantification of a specific phase of microglial activation within specific brain regions.

Our data also suggest that TSPO imaging has promise as a measure of disease persistence and/or severity in TBI with potential for monitoring response to therapeutic interventions. Notably, the regional reductions we observed following whole brain, voxel-wise analysis, and within-brain regions impacted by TBI are implicated in clinical presentations including depression, fatigue, and pain (e.g., caudate, ventral anterior thalamus, nucleus accumbens, and parahippocampal gyrus).^{25,42–46,53–56} It is therefore even more compelling that HB-adMSC-induced reduction in density of brain immune activation within these specific regions predicted improvement in depression, fatigue, and pain symptoms in our small cohort. Additionally, the presence of overlapping brain circuitry between HB-adMSC reduction in brain immune activation (e.g., parahippocampus) and concurrent HB-adMSC-induced improved fractional anisotropy in the hippocampus further underscores the potential for multimodal brain imaging in predicting response and treatment delivery in TBI patients.

That density of regional brain immune activation was reduced following HB-adMSC intervention is both novel and impactful, showing evidence of significant movement of a potential new target following treatment with a novel intervention being studied in chronic TBI. The regional specificity of the effects localized to brain regions overlaps with those identified in animal models^16,57 to show localized modulation of microglial activation following treatment with MSCs. However, how these PET and structural MRI results tie together into a specific mechanism of injury in these patients requires further study.

Correlation With TBI Clinical Symptoms

Neuroregulatory dysfunction in brain circuits incorporating these regions has been reported in depression, fatigue, and pain patients with^53–56 and without^{24,25,42,44–46} history of TBI. In our study population, at baseline, 6-12 months after index injury, participants presented with a range of clinical symptoms, including depression, fatigue, and pain, common clinical manifestations of brain injury in the first year following TBI.^58–60 Significant HB-adMSC-induced reductions in depression, fatigue, and pain were found in our participants, suggesting HB-adMSC may be helpful in mitigating chronic symptoms of depression, fatigue, and pain following TBI. However, given the small sample size (and overlapping symptoms in some participants), while these data are encouraging, further investigation with larger samples is necessary.

Our voxel-wise ANOVA testing showed strong linear relationships between HB-adMSC-induced reduction in regional density of brain immune activation and HB-adMSC-induced reduction in symptoms of depression, fatigue, and pain, outlining the potential clinical impact of this novel intervention. That voxel-wise analyses showed regions where significant effects were identified as critical components of brain circuitry dysregulated in depression, fatigue, and pain further underscores the potential clinical translational impact of our study.

Brain immune activation has been widely reported in both animal models and human subjects with a history of TBI. That we evidenced strong correlations between specific clinical symptoms (depression, fatigue, pain) and density of regional brain immune activation takes us a step further in clinical translation of prior research.

Estimation of Treatment Effect

We chose clinically relevant and accepted read-outs of neurocognitive and functional measures. Across many domains, there were significant improvements across the first 12 months after treatment in functional outcome, self-reported quality of life, and psychological health. Based on the small to medium effect sizes for many of these outcomes, gain approximated 11-20 percentile points. The significant reduction in disability and improvement in self-reported cognition, communication, independence, and satisfaction with social and recreational activities are clinically meaningful. Neurocognitive outcome measures showed significant gains in areas often adversely affected by and persisting following severe TBI; these include processing speed, cognitive flexibility, and learning and retention of new information.^61,62

Significant improvements in cognitive and psychological health in patients with chronic TBI are highly relevant. These are target areas for rehabilitation efforts designed to return persons with TBI to full community participation and engagement. As this is not a comparative trial, these data are encouraging but must be cautiously interpreted. However, more accurately, these data allowed us to estimate a treatment effect size to allow accurate sample size calculations for functional outcome measures in future clinical trials. The sample size used in our current study was determined based on the current effect size range for planned outcome variables. However, we plan an enrichment of the population based on responder status and a priori analysis based on this baseline predictor. Using those characteristics to calculate the sample size, we estimate 64-177 patients would be required (Supplemental Table 5).

Limitations

Some limitations have been discussed above. Additionally, in our study, brain regional TSPO density was estimated via PET imaging and the highly selective TSPO radiotracer, [¹¹C]ER176.^15,16 Genetic testing was completed on all study participants to screen for a genetic variant that confers reduced binding affinity of the TSPO tracers. However, given our radiotracer choice, scanning was possible (and comparable) across all binding affinity phenotypes consequent to this single nucleotide polymorphism.¹⁵ Control for variation in binding affinity was achieved during parametric image construction and SUVR calculation.

Over the past 10-20 years, TSPO imaging has been used as a measure of brain microglial activation in animal models and humans. However, recent studies in animal models and humans with multiple sclerosis suggest that TSPO imaging is not sufficiently specific to microglial activation state (e.g., “pro-inflammatory,” “anti-inflammatory”) in humans, but that its use in that regard is limited to animal models. In support, Nutma et al cite the ability to detect specific activation states in animal models but not in humans with multiple sclerosis and suggest that TSPO imaging can only be interpreted as representing the density of brain immune activation.¹⁶ Indeed, this information is informative, with potential to have major implications in the field. However, certain study limitations may prevent the generalization of their findings to chronic TBI in humans. The study used lipopolysaccharide or interferon to induce TSPO expression as the primary outcome. Differences between rodent and human reactivity to lipopolysaccharide have been reported elsewhere,⁵⁷ and inflammatory processes activated in multiple sclerosis have not been shown to overlap those of chronic TBI in humans, which likely has a chronic immune response distinct from that induced by lipopolysaccharide or interferon, and no history of trauma.

Also, we appreciate that this pseudo-reference region approach may underestimate the TSPO non-displaceable binding potential (BP_ND) within the regions of interest. However, given our statistical plan, largely correlative in nature and not involving calculation of binding potential, we believe subsequent analyses in the setting of this underestimation will likely not pose major confounds to the data. Recently, reactive astrocytes have been noted to have upregulated TSPO, with similar responses as those seen in microglia: hypertrophic morphology, release of pro-inflammatory cytokines and chemokines, and localization at the site of damage signals. These have been studied using both glial fibrillary acidic protein co-localization studies and in TSPO-GFAP knockout models.⁶³

Conclusion

Our study demonstrates autologous adipose-derived MSCs for chronic TBI are safe, have clinically relevant treatment effect sizes in functional and neurocognitive outcome measures, yield significant improvements in specified measures of microstructural integrity in brain-based behavior relations, and reduce microglial activation in regions corresponding to pain, fatigue, and depression. These data provide quantitative justification for the sample size of next-phase clinical trials using either functional outcomes or surrogate imaging outcomes.

Supplementary Material

Supplementary material is available at Brain online.

Data availability

The data that support the findings from this study are available from the corresponding author upon reasonable request and in compliance with institutional policy.