Non-invasive modeling of TSPO dynamic PET imaging: development and validation of imaging biomarkers of neuroinflammation

Neuroinflammation plays a crucial role in maintaining central nervous system homeostasis by protecting the brain and spinal cord and promoting tissue repair. While neuroinflammation is a natural defense mechanism, chronic neuroinflammation can become harmful, contributing to the development of neurodegenerative, psychiatric, and inflammatory disorders. This has driven significant research interest in the development of neuroinflammatory biomarkers for patient immune phenotyping and the development of immune therapies across various brain disorders. Positron emission tomography (PET) imaging targeting the mitochondrial 18kDa translocator protein (TSPO) represents the most adopted technique for imaging neuroinflammation in vivo, but it suffers from the lack of a standardized and reliable analytical framework for its quantification. Methodological factors represent the first source of complexity: while standard blood-based kinetic modeling faces practical issues related to the definition of the arterial input function, alternative reference-based quantification approaches are hindered by the challenge of identifying a valid reference region devoid of TSPO binding sites. Beyond methodological aspects, a significant inter- and intra-subject variability of parenchymal and plasma TSPO signal strongly hampers the identification of disease effects. Among the causes of this variability are the single nucleotide genetic polymorphism that affects second-generation TSPO tracer affinity, possible alternations of blood-brain barrier (BBB) functioning that affect tracer delivery and thus measures of brain target density, but also factors such as age, sex, stress which can have an impact on the brain TSPO expression. Finally, the complex and dynamic interplay between central and peripheral immune responses underlying brain disorders remains unclear, adding further complexity to TSPO PET imaging data analysis and interpretation. Over the years there has been a significant disagreement within the PET modeling community regarding the best approach for TSPO PET analysis, and this has often resulted in inconsistent and non-replicable results across studies and concerns about the value of TSPO PET imaging. Positioned within this chaotic framework, this doctoral thesis aims to explore alternative quantification approaches to overcome the major limitations of traditional methodologies. Specifically, this thesis presents three different approaches. In the first section, the methodology of the image-derived input function is exploited to develop a non-invasive marker of tracer blood-to-brain delivery. This marker allows the detection of altered BBB functioning during neuroimmune responses in brain disorders and is fundamental to account for possible differences in tracer delivery which could be misunderstood as changes in brain TSPO load. The second section of the thesis focuses on the definition of a new blood-free reference-free analytical framework for the quantification of neuroinflammatory load. This method shares basic concepts of normative modeling approaches in neuroimaging studies and exploits a logistic regression model that, taking as input the regional tracer time-activity courses and a set of individual covariates, provides a measure of individualized regional neuroinflammatory load based on the probability of TSPO overexpression across brain tissues as compared to a healthy reference. Lastly, the third section is dedicated to a total body analysis of inter-organs connectivity in healthy and inflammatory conditions. This proof-of-concept study aims to provide new insights into the relationship between central and peripheral immune systems, exploiting the advantages of animal models and total body PET imaging. Overall, this thesis provides complementary methodological tools that could facilitate TSPO PET imaging studies. These approaches would remain valid in the event of the development of new molecular targets of neuroinflammation.