Atrophy-related inflammation CSF markers predict clinical progression in genetic frontotemporal dementia

Neuroinflammation is increasingly recognised as a key contributor to frontotemporal dementia (FTD) pathophysiology, yet how inflammatory processes relate to neurodegeneration and clinical progression remains poorly defined. This study aimed to identify a multivariate cerebrospinal fluid (CSF) inflammatory signature covarying with brain atrophy, and to determine its relevance to clinical and cognitive outcomes in genetic FTD. We recruited 234 participants (non-carrier controls n=78, presymptomatic carriers n=115, symptomatic carriers n=41) with CSF, MRI, and clinical (CDR (r) + NACC FTLD sum of boxes [CDR-SB]) and executive (Wechsler Adult Intelligence Scale-Revised Digit Symbol Substitution Task [DSST] assessments. A regularised canonical correlation analysis (rCCA) was used to derive a shared protein-brain covariance axis. Participants were followed annually for up to 8.6 years (CDR-SB) and 11.5 years (DSST), and linear mixed-effects models were used to estimate individual rates of clinical progression (CDR slope) and executive decline (DSST slope). The predictive value of the inflammation- and MRI-based profiles for cross-sectional CDR-SB, CDR slope, DSST and DSST slope were tested in linear models. Due to floor effects and violations of parametric assumptions, Bias-corrected and accelerated (BCa) confidence intervals (CI) and bootstrap CIs (95%, 5000 replicates) were utilised as the primary inferential criterion in the 2nd-level linear models; effects were considered non-significant if the 95% BCa CI crossed zero. The rCCA identified a robust inflammation-brain axis (r=0.89, p<0.001), driven by CXCL9, FLT3L, CCL3, and frontotemporal-insular atrophy pattern. In symptomatic carriers, the MRI canonical variate was associated with CDR-SB (beta=3.80, CI: [0.64, 6.20]) at baseline, whereas the inflammatory protein canonical variate was predictive of CDR-SB slope (beta=0.24, CI: [0.02, 0.46]). Whereas both MRI (beta= -7.27, CI:[-14.20, -1.00]) and protein (beta= -7.83, CI:[-14.00, -2.16]) canonical variates were linked to DSST cross-sectionally, but not longitudinally. No significant associations were found in pre-symptomatic carriers or non-carrier controls. A data-driven CSF inflammatory signature, covarying with frontotemporal-insular atrophy, tracks with clinical severity and cognitive impairment in people with genetic FTD, highlighting inflammation as a candidate therapeutic target and progression biomarker in the earliest clinically manifest stages of disease.