BACKGROUND Screening for subclinical TDP43 neuropathology in autoimmune conditions: a pilot study Natasha Clarke, Nidhi Sofat, Lena Assi and Peter GarrardContact: HYPOTHESIS RESULTS Participants Graded Naming Test Max. possible 30/30 NC mean = 20.4 (SD=4.1) Pyramids and Palm Trees Test Max. possible 52/52 NC mean = 90% Category and letter fluency Letter responses > category responses suggests semantic impairment Figure 3 The three tests included in the semantic battery. “What is this?” “Which of the two bottom pictures goes with the top one?” “Tell me all the animals you can think of, starting now.” Pathology of Frontotemporal Lobar Degeneration TDP-43 Proteinopathy and Inflammation Picture Naming Fluency Semantic Association FTLD-tau Aggregates of tau protein in the form of tufted astrocytes or intracellular aggregates (‘Pick bodies’). See below for histopathology. Presenting clinical syndrome partially predicts underlying pathology: Conclusions  TDP-43 pathology is linked to mutations in the gene coding for the progranulin protein (PGRN), a precursor to the granulin (GRN) family of glycosylated peptides (Tang et al., Science 2011).  FTD GWAS showed significant association at the HLA locus on chromosome 6p21.3 (Ferrar et al., Lancet 2014).  Granulins are involved in cytokine regulation by virtue of their inhibitory effects on TNFα.  Up to 43% of patients with autoimmune conditions such as rheumatoid arthritis (RA), but less than 1% of controls have antibodies against PGRN (Thurner et al., Journal of Autoimmunity 2013).  There is a significantly higher prevalence of autoimmune conditions in patients with definite (GRN mutation positive) or highly probable (clinical diagnosis of SD) TDP-43 pathology than in matched controls with other dementia subtypes (Miller et al., Journal of Neurology, Neurosurgery & Psychiatry 2014). If autoimmune disease and TDP-43 proteinopathy are linked by the presence of circulating antibodies resulting in lower levels of GRN and upregulation of TNFα, we might also expect to see evidence for FTLD-TDP in patients with autoimmune conditions. METHOD Figure 2. Hypothetical mechanistic model. FTLD-TDP Cellular inclusions are positive for both ubiquitin and transactive response DNA- binding protein with 43 kDa (TDP-43). See below for histopathology. 1) Semantic test battery 2) Machine learning classification  40 participants, aged within the usual range of clinical presentation of SD (i.e ) and with English as a first language were recruited via NHS clinics.  All were undergoing treatment for an autoimmune mediated condition (Rheumatoid Arthritis or Psoriatic Arthritis) or had similar systemic symptoms without an autoimmune mechanism (Osteoarthritis).  Participants completed a cognitive screening questionnaire, the Mini Mental State Examination (MMSE), an abbreviated semantic test battery (see below).  A blood sample was obtained for later measurement of progranulin and TNFα. Scores were first transformed in to nominal variables based on a median split, before classification using the J48 classifier (Weka (v3.6)). Decision Tree Accuracy Correctly Classified Instances = 70% Incorrectly Classified Instances = 30% Confusion Matrix ab  Classified as 231a=Inflammatory 115b=Osteoarthritis Graded Naming Test Pyramids & Palm Trees Fluency RatioFluency Studies have consistently shown SD to be underpinned by FTLD-TDP pathology, in up to 90% of cases at post-mortem (e.g. Snowden et al., Acta Neuropathologica 2007);. The presence of a semantic deficit in a patient with FTLD is therefore a clinical biomarker for TDP-43 proteinopathy.  Whilst the current sample size may be too small to support definitive conclusions, this pilot study so far suggests that no single neuropsychological measure is sufficient to reveal underlying TDP43 pathology in this patient population.  Machine learning may provide a possible approach to classifying group membership using a combination of neuropsychological features: the unexpected association between high scores on the GNT and inflammatory disease may reflect a more generally semantically impaired subgroup.  Measurement of progranulin and TNFα levels is currently in progress, and will allow us to look for an association between cognitive and biological markers. Figure 1. Underlying pathology in 71 cases. AD = Alzheimer’s Disease, DLDH = Dementia lacking distinctive histology (Alladi et al., Brain 2007). Figure 6. Boxplot with whiskers from minimum to maximum. Figure 4. Bar graph with error bars showing +/- 1 SD from the mean. Figure 7. Boxplot with whiskers from minimum to maximum. Figure 5. Boxplot with whiskers from minimum to maximum. n.s. p =.15 n.s. p =.13 n.s. p =.46 n.s. p =.31 n.s. p =.84 St George’s, University of London, Cranmer Terrace, London Impaired Semantics Normal Semantics