Glial Activity Load on PET Reveals Persistent “Smoldering” Inflammation in MS Despite Disease-Modifying Treatment

PET imaging using radioligands that target the 18-kDa translocator protein (TSPO), located on the outer mitochondrial membrane, has been used to study innate immune activation in multiple sclerosis (MS) and various other brain diseases.¹ Pathological studies have shown that TSPO-PET signal in MS largely originates from CD68+ microglia/macrophages with some contribution from astrocytes and that TSPO-PET reports on glial density in MS.^2,3 “Smoldering” inflammation is a recently popularized term used to describe widespread innate immune activation that is present beyond focal inflammatory lesions in MS patients.⁴ This persistent smoldering inflammation has been considered as a critical component of disease progression but is not readily assessed in MS patients.⁴

¹⁸F-PBR06, a phenoxy-aryl-acetamide derivative, is a second-generation TSPO ligand that has been validated as a marker of innate immune activation in several preclinical models and has been evaluated in human studies involving MS patients.^5–12 Our initial studies have revealed that ¹⁸F-PBR06-PET signal is increased in progressive MS (PMS) as compared with relapsing MS (RMS),⁵ is linked with fatigue and depressive symptoms,⁶ and can demonstrate changes prospectively, following treatment with disease-modifying treatment (DMT) in MS patients.¹³ We have also shown that, in a head-to-head comparison with ¹¹C-PBR28, ¹⁸F-PBR06-PET measures had a lower coefficient of variation and correlated more strongly with clinical measures in MS subjects.¹²

¹⁸F-PBR06 has higher affinity (K_i = 0.3 ± 0.08 nM) for its target as compared with ¹¹C-PK11195 (Ki = 3.48 ± 1.26 nM); high brain uptake and a high proportion of specific to nonspecific binding provide high sensitivity to detect small changes in TSPOs in the brain.^7,9,14 Metabolic stability of ¹⁸F-PBR06 relative to the longer half-life of the ¹⁸F-PBR06 allows the extended acquisition time to measure TSPO as a biomarker of innate immune activation in the brain. Longer half-life of ¹⁸F-PBR06 enables widespread use of this ligand for clinical trials and potential clinical use in the future, without the need of an on-site cyclotron. Moreover, routine arterial sampling and metabolite analyses are not clinically feasible,¹⁵ and generally speaking, many assumptions made in tracer kinetic modeling may not be valid and generate parameters that themselves require further validation by other independent measurements, which are of biochemical, clinical, and/or pathological significance.^15–17 Further, it has been stated that mere fitting of a model to kinetic data does not prove the validity of a model.¹⁷ Hence, our aim is to study a novel, noninvasive, clinically feasible, individualized approach using ¹⁸F-PBR06-PET for identifying widespread smoldering inflammation and its relationship with clinical disability, brain volumetric changes (specifically, cortical atrophy), and serum biomarkers in MS patients.

PATIENTS AND METHODS

Subjects

Thirty ¹⁸F-PBR06-PET scans were performed in 22 MS subjects and 8 healthy controls (HCs). Of the 22 MS subjects, 13 patients were on high-efficacy DMT (H-DMT) (9 women; mean age ± SD, 43 ± 13 years [range, 23–65 years]; median Expanded Disability Status Scale [EDSS] score, 4.0 [range, 1.0–7.5]; timed 25-ft walk [T25FW], 7.47 ± 4.61 seconds [range, 3.40–19.40 seconds]), and 9 patients were on low-efficacy DMT or no DMT (L-DMT) (7 women; mean age ± SD, 50 ± 14 years [range, 28–69 years]; median EDSS score, 3.0 [range, 1.5–6.5]; T25FW, 6.77 ± 4.41 seconds [range, 3.91–17.26 seconds]). Eight HCs (3 women; mean age ± SD, 44 ± 17 years [range, 25–69 years]) were also prospectively recruited for this study. Multiple sclerosis subjects in the HT group were on rituximab, natalizumab, and fingolimod (DMTs that have shown >50% efficacy in reducing relapse rates in MS clinical trials) versus those in the LT group were on no DMT or on platform therapies including glatiramer acetate and interferons. Subject characteristics are summarized in Tables 1 and 2. The study was approved by the institutional review board and radioactive drug research committee of our hospital; written informed consent was obtained from all participants. The ClinicalTrials.gov registration number for this study was NCT02649985.

Genotyping

Blood samples drawn on the initial screening visit were used to obtain genomic DNA and assess polymorphism within the TSPO gene on chromosome 22q13.2, using a Taqman assay. High-affinity binder (HAB) and medium-affinity binder (MAB) were included in this study, whereas low-affinity binders were excluded.¹⁸ Of the 13 HT subjects, 8 were HABs and 5 were MABs. Of the 9 LT subjects, 5 were HABs and 4 were MABs. Of the 8 HC subjects, 4 were HABs and 4 were MABs.

Production of Radiopharmaceuticals

¹⁸F-PBR06 was produced in the PET radiochemistry facility at our hospital according to standardized procedures.^7,9,14 These were purified by high-pressure liquid chromatography and sterilized by membrane filtration using a 0.22-μm membrane filter. The final product was dispensed in an isotonic solution that was sterile and pyrogen-free and ready for intravenous (IV) administration. The radiochemical purity of the radiopharmaceuticals was determined using high-pressure liquid chromatography. The organic solvents were determined using gas chromatography. The radiochemical purity of each radiopharmaceutical was greater than 95%.

MRI Acquisition and Analysis

All subjects underwent brain MRI scans on the same scanner (Siemens 3 T Skyra; Siemens Healthineers, Erlangen, Germany) using the same high-resolution acquisition protocol. Whole-brain images were acquired with a 2D T1-weighted spin echo axial series. In addition, patients underwent a 3D magnetization-prepared rapid gradient-echo sequence (MPRAGE; voxel size 1 mm³). Only noncontrast MRI scans were obtained; IV gadolinium contrast was not administered as part of this study. Whole-brain and thalamic normalized volumes and cortical thickness (CoT) were measured based on our previously described methods using the MPRAGE sequence.^5,19–21

Serum glial fibrillary acid protein (GFAP) and neurofilament light chain (NfL) were measured according to our previously described methods.²²

PET Acquisition and Analysis

Radiotracers were injected as a bolus for PET scanning using an IV catheter inserted into an upper extremity vein; images were acquired in a list acquisition mode using a high-resolution, whole-body PET/CT scanner. The mean injected doses of ¹⁸F-PBR06 were 2.51 ± 0.53 mCi (range, 1.39–3.56 mCi). The mean time interval between MRI and ¹⁸F-PBR06-PET scans was 28 ± 42 days (range, 0–182 days). Dynamic images were acquired over 120 minutes, and summed ¹⁸F-PBR06-PET images were derived based on PET data acquired between 60 and 90 minutes after tracer injection and were then coregistered to each individual patient’s T1 MRI (the spin-echo or MPRAGE) scans using the PMOD 3.8/3.9 platform (PMOD Technologies, Zurich, Switzerland; www.pmod.com). This standardized algorithm involved coregistration of the T1-weighted series and PET images of each individual with the Automated Anatomical Template.^5,6 Individualized parametric 3-dimensional z-score maps of brain parenchymal microglial activation (MA) were generated by voxel-by-voxel statistical comparison between each subject’s globally normalized PET images and an HC dataset (Supplementary Fig. S1, https://links.lww.com/CNM/A470).⁸ Logarithmically transformed “glial activity load on PET” (GALP) scores (calculated as the sum of voxel-by-voxel z-scores thresholded at ≥4 in cortical gray matter [CoGM] and white matter [WM] regions), “lnGALP,” were compared between MS subjects on HT and LT, and correlated with clinical measures, serum biomarkers, and CoT.

Statistical Analysis

Student t tests were used to assess mean differences. Pearson correlation coefficient r values were calculated; multivariate regression analyses and mediation analyses were performed. IBM SPSS statistics version 24.0 was used for statistical analyses. P < 0.05 was considered statistically significant.

RESULTS

Group Comparisons

MS Patients Have Increased Cortical and White Matter GALP Than Healthy Controls

CoGM-lnGALP and WM-lnGALP scores were both increased in MS versus HC (10.0 ± 1.5 vs 7.5 ± 1.4, +33% P = 0.0003, Fig. 1A; and 9.8 ± 1.5 vs 6.6 ± 2.3, +48%, P = 0.0002, Fig. 1B, respectively; Table 3). Meanwhile, CoT, normalized whole-brain volume (nBPV), and normalized thalamic volume (ThV) were all decreased in MS versus HC (2.48 ± 0.07 vs 2.55 ± 0.08 mm, −3%, P = 0.041, Fig. 1C; 1404.3 ± 77.6 vs 1491.3 ± 51.7 mL, −6%, P = 0.008, Fig. 1D; and 20.1 ± 1.9 vs 22.8 ± 1.0 mL, −12%, P = 0.001, Fig. 2C, respectively).

MS Patients Treated With High-Efficacy DMTs Have Lower GALP Than Those Treated With Low-Efficacy DMTs but Are Still Abnormal as Compared With HCs

CoGM-lnGALP and WM-lnGALP scores were both decreased in HT versus LT (9.1 ± 0.9 vs 11.3 ± 1.1, −20%, P = 0.00008, Fig. 1E; and 9.1 ± 1.2 vs 10.8 ± 1.3, −16%, P = 0.006, Fig. 1F, respectively; Table 3). On the other hand, there were no significant differences in CoT, nBPV, ThV, serum GFAP, and serum NfL between HT and LT groups.

Mean CoGM-lnGALP and WM-lnGALP scores in the HT group of MS patients were both still increased compared with HC (9.1 ± 0.9 vs 7.5 ± 1.4, +21%, P = 0.006, Fig. 1E; and 9.1 ± 1.2 vs 6.6 ± 2.3, +37.8%, P = 0.006, respectively, Fig. 1F). Although nBPV and ThV were decreased in HT versus HC (1379.8 ± 73.4 vs 1491.3 ± 55.3 mL, P = 0.002, Fig. 1H; and 19.7 ± 2.2 vs 22.8 ± 1.0 mL, P = 0.002, Fig. 2A, respectively), the difference between mean CoT in HT and HC did not attain statistical significance (2.48 ± 0.08 vs 2.55 ± 0.08 mm, P = 0.068, Fig. 2G).

PMS Patients Have Increased Cortical Glial Activity as Compared With Relapsing-Remitting MS Patients Despite Treatment With H-DMTs

Within the H-DMT group, CoGM-lnGALP scores were increased in PMS versus RMS (9.8 ± 0.7 vs 8.5 ± 0.7, +15%, P = 0.008, Fig. 1I). However, WM-lnGALP was not significantly different in HT-PMS versus HT-RMS (9.3 ± 1.5 vs 8.9 ± 0.9, +4.5%, P = 0.59). In terms of volumetric measures within the HT group, CoT and nBPV were both decreased in PMS versus RMS (2.43 ± 0.05 vs 2.52 ± 0.06 mm, −3.5%, P = 0.017, Fig. 1K; and 1327.7 ± 57.6 vs 1424.4 ± 53.4 mL, −6.8%, P = 0.015, Fig. 1L), but ThV was not statistically significant in HT-PMS versus HT-RMS (18.8 ± 1.9 vs 20.5 ± 1.9 mL, −8.3%, P = 0.16, respectively). Interestingly, GFAP was increased in PMS versus RMS (42.61 ± 12.96 vs 20.50 ± 7.40, +108%, P = 0.024, n = 18, Fig. 2G). However, within the same subset, NfL was not significantly different in PMS compared with RMS (3.78 ± 1.92 vs 2.60 ± 1.48, +45%, P = 0.38, n = 18).

Correlations Across Groups

Increased Cortical and White Matter Glial Activation Is Linked With Cortical Atrophy

CoGM-lnGALP and WM-lnGALP scores were negatively correlated with CoT across MS and HC groups (r = −0.44 and r = −0.48, P < 0.05, Figs. 3A, B). On the other hand, there was a significant inverse relationship between ThV and WM-lnGALP (r = −0.404, P = 0.027, Fig. 3D), but ThV only showed a trend toward an inverse relationship with CoGM-lnGALP (r = −0.310, P = 0.096, Fig. 3C). The relationship between nBPV and CoGM-lnGALP and WM-lnGALP did not achieve statistical significance.

Correlations Within Subgroups

Cortical Glial Activity Is Associated With Clinical Disability, Fatigue Scores, Cortical Degeneration, and Serum GFAP in H-DMT Patients

On subgroup analyses, within the H-DMT group, CoGM-lnGALP correlated positively with EDSS, T25FW, and modified fatigue impact score (MFIS) (r = 0.645, 0.786, and 0.75, all P's < 0.05, Figs. 4A–C, Table 4), and negatively correlated with CoT (r = −0.659, P < 0.05, Fig. 4D) but did not correlate with nBPV or ThV. CoGM-lnGALP was also positively correlated with serum GFAP (r = 0.67, P < 0.05, Fig. 4E). In contrast, there were no significant correlations between WM-lnGALP and the clinical (EDSS, T25FW, MFIS), MRI (CoT, nBPV, ThV), or serum biomarker measures in the HT group. Correlations of other PET indices with clinical and MRI measures are shown in Supplementary Table S1, https://links.lww.com/CNM/A471.

Within L-DMT group, interestingly, none of the relationships between cortical and WM GALP scores and clinical and morphometric scores in the LT group attained statistical significance. These findings emphasize the specificity of the relationship between cortical MA with cortical thinning and clinical disability in the HT group (Figs. 5–7).

DMT Efficacy Modulates the Relationship Between Cortical Glial Activity and Clinical Impairment and Serum Biomarkers in MS Patients

On multivariate analyses, CoGM-GALP predicted EDSS, T25FW, and MFIS scores, and GFAP and NfL levels independent of DMT efficacy, but each of the relationships was moderated by the interaction between the DMT efficacy and cortical GALP. Further, CoGM-lnGALP predicted a diagnosis of PMS after adjusting for DMT efficacy, and this relationship was moderated by the interaction between the DMT efficacy and CoGM-GALP.

Receiver Operating Characteristic Analysis

Cortical GALP Outperforms Cortical SUVR for Differentiating Progressive From Relapsing-Remitting Patients in H-DMT Group

On receiver operating characteristic (ROC) analysis across modalities, CoGM-lnGALP, nBPV, CoT, and GFAP all provided significant differentiation for distinguishing PMS from RMS within the HT group (areas under the curve [AUCs] = 0.91, 0.90, 0.88, and 0.88, respectively; Supplementary Table S2, https://links.lww.com/CNM/A471). Meanwhile, NfL, WM-lnGALP, and ThV had lower AUCs on ROC analysis (0.71, 0.67, and 0.61, respectively). On further ROC analysis across additional PET indices, SUVg, SUVTh, and SUVRgTh were also significant in distinguishing PMS from RMS within HT (AUC = 0.81, 0.81, and 0.76, respectively; Supplementary Table S3, https://links.lww.com/CNM/A471). SUVRgCo and WM SUV had lower AUCs within HT (0.62 and 0.60, respectively).

Mediation Analysis

Cortical Neuroinflammation Mediates the Relationship Between PMS and Cortical Atrophy and Serum GFAP Levels

Although both lower CoT and higher serum GFAP predicted a diagnosis of PMS in the HT group, this relationship was mediated by cortical glial activity (CoGM-GALP) (Sobel test P value <0.05) on exploratory analyses, supporting the central role of glial activation in the pathogenesis of PMS.

DISCUSSION

There are several key findings in our article. First, we describe a novel, unbiased, individualized approach to analyzing ¹⁸F-PBR06-PET in MS. Second, we demonstrate that, in MS patients, high-efficacy disease-modifying therapies may reduce but do not normalize glial activity in the cerebral cortex and WM. Further, we demonstrate that among MS patients treated with H-DMT, cortical smoldering inflammation measured by ¹⁸F-PBR06-PET is linked with cortical atrophy (lower CoT), worsening physical disability, higher depression and fatigue scores, higher serum GFAP levels, and is increased in PMS as compared with RMS.

Voxel-wise, z-score mapping (VZM) has been extensively applied in brain PET imaging and recommended for enhanced evaluation of individual patients’ FDG-PET scans.²³ This approach has led to a standardized, unbiased approach for interpreting brain FDG-PET images in the evaluation of dementias.²⁴ It has also yielded high sensitivity and specificity in distinguishing patients with Alzheimer disease from patients with other dementias and from HCs.²⁴ A key advantage of the VZM approach is that it can be applied to static PET scans acquired at later time points following radiotracer injection when the radiotracer is more likely to have reached an equilibrium state. This approach is widely used clinically because it is less time-consuming as compared with dynamic imaging and improves patient comfort. From an image quality standpoint, ¹⁸F-PBR06 is a good candidate radiotracer for the application of the VZM approach on static imaging as it is a ¹⁸F-labeled PET tracer with a longer (~110 minutes) half-life. Owing to its longer half-life, static imaging acquired 60–90 minutes after radiotracer injection provides an improved signal-to-noise ratio and better image quality as compared with ¹¹C-labeled tracers, which have a significantly shorter half-life (20 minutes) and therefore result in poor count statistics at later time points.¹²

There is an urgent need to standardize image analysis and interpretation in TSPO-PET imaging. We have found that individualized, VZM mapping is feasible and provides meaningful results in MS patients using ¹⁸F-PBR06. Previous methods in TSPO-PET have focused on evaluating innate immune activation based on average estimates of radiotracer uptake in large brain regions of interest. However, smaller focal areas of the brain parenchyma may have abnormalities that may be missed using the region of interest approach, reducing its sensitivity. Furthermore, using a uniform threshold cutoff to define abnormalities for the entire brain may not be accurate due to the known regional variations in microglial density in healthy brains. On the other hand, a voxel-wise, z-score approach identifies specific abnormalities in a given voxel in a patient’s image as compared with the corresponding set of voxels in the reference population, thereby improving the sensitivity and specificity of the results. Predefined smoothing parameters and z-score thresholds are crucial to minimizing the impact of image noise and false-positives due to multiple statistical comparisons in VZM. In our study, we applied a 3-dimensional smoothing filter during image processing and a z-score threshold of >4 to address these issues.

We have previously reported our preliminary findings using the SUVR approach in a subset of the current sample that showed an inverse relationship between thalamic PET ligand uptake and brain volume.⁵ Using the VZM approach, we are able to demonstrate an inverse relationship between innate immune activation in the cerebral cortex and WM and thalamic and whole-brain volumes. These results reinforce the relationship between innate immune activation and brain parenchymal tissue loss in a larger dataset, and using our new, individualized, VZM methodology.

We demonstrate that patients treated with H-DMTs have lower innate immune activation than those who are either untreated or treated with low-efficacy disease-modifying therapies. Several DMTs have been previously shown to reduce PET MA in MS (largely, relapsing remitting MS [RRMS]) patients in longitudinal studies. A study in a mixed population of RRMS and secondary progressive MS (SPMS) patients showed a reduction in ¹¹C-PK11195-PET distribution volume ratios after 1 year of treatment with natalizumab.²⁵ Similarly, 12% reduction in lesional binding potential (nondisplaceable) was seen after fingolimod treatment for 6 months in an RMS population.²⁶ Further, our preliminary analysis showed a reduction of 19.4% in cortical GALP after 3 months of treatment with ofatumumab, a B-cell therapy.¹³ On the other hand, although it is not straightforward to compare across studies, it is notable that treatment with glatiramer acetate showed only a 3.17% reduction in MA in RMS patients after 1 year.²⁷ Overall, these longitudinal studies suggest that higher efficacy DMTs are associated with significantly reduced MA in at least some brain regions. Our data in this current cross-sectional study are consistent with these longitudinal studies in terms of decreased PET signal in HT versus LT group. Moreover, the magnitude of longitudinal change following initiation of an H-DMT treatment in these studies is similar to the cross-sectional group differences between the HT and LT groups (~20%) we report in our study.

Further, it is well known that the currently approved DMTs have not shown efficacy in inactive PMS patients, and a significant proportion of patients progress despite being on an H-DMT.^28,29 This phenomenon may be explained by our finding that, although H-DMTs are associated with a decreased PET signal, there is still abnormally persistent glial density load (or smoldering inflammation) despite high-efficacy treatment in MS patients. Many of the current H-DMTs target the peripheral immune system, do not cross the blood-brain barrier, and therefore have limited efficacy on the progressive component of the disease process.^29,30 We believe that our approach can enrich clinical trials by identification of patients with significant “residual” MA despite H-DMTs who may benefit from emerging treatments targeting smoldering inflammation. We also believe that the GALP approach can facilitate individualized assessments of treatment response in MS patients, including those who are first started on L-DMTs. Similarly, future longitudinal studies are needed to evaluate the role of TSPO-PET in decision making regarding switching MS subjects from L-DMTs to H-DMTs, on an individual basis.

Cortical MA and neurodegeneration have been associated with MS progression in multiple pathological and imaging studies.^31,32 Within the HT group, we observed higher abnormal cortical MA in SPMS as compared with RMS patients. This is consistent with increased regional cortical TSPO-PET signal, which we have reported previously using ¹⁸F-PBR06 and have also been reported using ¹¹C-PK11195.^5,33 Further, we observed an inverse relationship between cortical GALP scores and CoT. Cortical atrophy is a key feature of MS pathology, and cortical MA associated with synaptic injury has been consistently reported in pathological studies.^31,32,34 Our findings provide novel data reinforcing the link between abnormal innate immune activation and local brain parenchymal injury in the CoGM in MS patients despite high-efficacy treatment. Furthermore, we observed that abnormal cortical GALP scores predicted clinical disability, depression, and fatigue severity, highlighting the clinical relevance and validity of cortical MA and our methodology in these MS patients. Prospective longitudinal studies in MS patients are urgently needed to investigate the prognostic value of abnormal cortical MA detected by ¹⁸F-PBR06-PET. Serum GFAP is an emerging biomarker for PMS,³⁵ and interestingly, our findings suggest that cortical innate immune activation is linked with elevated serum GFAP levels in patients treated with H-DMTs. GFAP is an astrocytic intermediate filament, and astrocytes are known to contribute to the TSPO-PET signal in MS.³⁵ The relationship between TSPO-PET and serum GFAP warrants further detailed evaluation. Moreover, additional studies are needed to evaluate the relationship of TSPO-PET findings with other validated fluid biomarkers of MS.

There are several limitations to our study. Our sample size is small, and further validation of our results in future larger studies is needed. From a technical standpoint, voxel-wise methods may be prone to noisy estimation of parameters, and assessment of test-retest variability is needed, although our preliminary data for evaluating longitudinal changes after therapy are encouraging. Refinement of image processing steps may improve sensitivity for detection of abnormal regional MA.

In summary, we have reported a novel voxel-wise, z-score–based approach to quantitate abnormal innate immune activation using ¹⁸F-PBR06-PET in MS. This study provides PET evidence for clinical and pathobiological importance of this smoldering inflammation that persists in MS patients despite treatment with high-efficacy disease-modifying therapies. This PET approach is clinically feasible, is relevant for multicentric studies, and can provide meaningful insights into MS pathology and response to treatment in individual patients. Further, it can potentially aid the evaluation of progression independent of relapse activity and expedite therapeutics research in this area, which is urgently needed to address the unmet need of targeting innate immune activation/smoldering inflammation in MS.

Supplemental Digital Content

• CNM_2024_04_13_SINGHAL_CNM-D-23-01320_SDC1.docx; [Word] (25 KB)
• CNM_2024_04_13_SINGHAL_CNM-D-23-01320_SDC2.docx; [Word] (15 KB)