Prodromal Parkinson disease signs are predicted by a whole-blood inflammatory transcriptional signature in young Pink1−/− rats

Background Parkinson disease (PD) is the fastest growing neurodegenerative disease. The molecular pathology of PD in the prodromal phase is poorly understood; as such, there are no specific prognostic or diagnostic tests. A validated Pink1 genetic knockout rat was used to model early-onset and progressive PD. Male Pink1−/− rats exhibit progressive declines in ultrasonic vocalizations as well as hindlimb and forelimb motor deficits by mid-to-late adulthood. Previous RNA-sequencing work identified upregulation of genes involved in disease pathways and inflammation within the brainstem and vocal fold muscle. The purpose of this study was to identify gene pathways within the whole blood of young Pink1−/− rats (3 months of age) and to link gene expression to early acoustical changes. To accomplish this, limb motor testing (open field and cylinder tests) and ultrasonic vocalization data were collected, immediately followed by the collection of whole blood and RNA extraction. Illumina® Total RNA-Seq TruSeq platform was used to profile differential expression of genes. Statistically significant genes were identified and Weighted Gene Co-expression Network Analysis was used to construct co-expression networks and modules from the whole blood gene expression dataset as well as the open field, cylinder, and USV acoustical dataset. ENRICHR was used to identify the top up-regulated biological pathways. Results The data suggest that inflammation and interferon signaling upregulation in the whole blood is present during early PD. We also identified genes involved in the dysregulation of ribosomal protein and RNA processing gene expression as well as prion protein gene expression. Conclusions These data identified several potential blood biomarkers and pathways that may be linked to anxiety and vocalization acoustic parameters and are key candidates for future drug-repurposing work and comparison to human datasets.


Background
Parkinson disease (PD) is the fastest growing neurological disorder and affects approximately 1% of the world's population over the age of 60. 1 The neuropathological hallmark of PD is death of dopaminergic neurons in the substantia nigra which leads to the diagnostic motor signs of the disease (eg., bradykinesia, rigidity, and resting tremor).3][4] Prodromal and early-stage PD is associated with a multitude of heterogeneous non-motor signs and symptoms such as sleep and vision disturbances, olfactory and gastrointestinal dysfunction, anxiety, and early-onset cranial sensorimotor impairments that likely have a variety of systemic pathologies. 5Moreover, there is no speci c biomarker test to diagnose PD in the early-stage, before the cardinal motor features of the disease appear, and consequently diagnosis and treatment are delayed.
Early-stage PD is di cult to investigate in humans due to inconsistencies in symptom manifestation, age of onset, and environmental factors.Biomarker identi cation research has generally focused on cerebrospinal uid (CSF), yet CSF collection via lumbar puncture is invasive, expensive, painful, and requires skilled healthcare providers to collect specimens.Alternatively, blood collection via phlebotomy is a ubiquitous, relatively non-invasive source for potential biomarkers that is inexpensive, routine, and still has a high clinical application.][10] These studies validate use of peripheral whole blood gene expression in biomarker discovery and potential development of diagnostics and prospective therapeutics.Despite these promising ndings, identi cation and understanding of early-stage PD genetic biomarkers within whole blood samples remains limited.
2][13] Large genome-wide association studies have shown these speci c genes are concomitantly implicated in idiopathic PD pathophysiology, 14 and are involved in a set of molecular pathways that trigger an early-onset pathology sequence that is indistinguishable from sporadic forms. 12Mitochondria within the CNS are subject to the immense metabolic demands of neuronal activity.Mitochondrial stress, abnormal mitophagy, and lysosomal dysfunction leads to the release of damage-associated molecules that can activate an innate immune response, as seen in genetic murine models of PD.For example, PINK1 (PARK6, phosphatase and tensin homolog (PTEN)-induced putative kinase 1) is involved in mitochondrial quality control and protects cells from stress-induced mitochondrial dysfunction.Loss of Pink1 is also involved in increased generation of pro-in ammatory cytokines and chemokines within plasma, sera, CSF, and blood linked to neuron death 15 as well as increased cytosolic mitochondrial mtDNA and induction of type-I interferon responses and apoptosis. 16netic rodent models of early-stage PD provide insight into the underlying genetics of idiopathic disease, provide experimental control to link genes to behavioral dysfunction associated with disease progression, and identify targets for the development of treatments.17 The Pink1-/rat parallels human idiopathic PD progression including early-stage behavioral changes due to sensorimotor and cranial motor dysfunction (eg, vocal communication). 18,19 ork over the past decade has demonstrated that Pink1-/rats develop early motor and non-motor de cits as soon as two months of age. 20,21 2][23][24][25][26][27][28][29] Rats communicate during social interactions by producing ultrasonic vocalizations (USVs) in part by contraction and adduction of the thyroarytenoid (TA) muscle. 30,31 ecent research has shown that loss of functional Pink1 in the TA muscle leads to increased in ammatory and cell death pathways including the TNF-α/NF-kB signaling pathway. 27The identi cation of key disease-related genes and biological pathways are important to develop to identify treatment targets for early signs including communication dysfunction.Yet, there is a need to establish translatability between biomarker identi cation within tissue types and an easily accessible, comprehensive sample type.
While previous cranial motor behavior and tissue-speci c genetic studies have been done in the Pink1-/rat model, whole blood gene expression has not been evaluated. 24,27,28,32 Byidentifying whole blood gene expression pro les in early-stage PD, we can develop a transcriptomic signature capable of detecting PD in prodromal stages.The purpose of the current study was to identify dysregulated gene pathways within the blood of young Pink1-/rats (3 months of age), develop genetic biomarkers or signatures that appear during the early-stage of disease, prior to the onset of hallmark limb motor signs, and evaluate whether they can predict early-stage, cranial motor-based vocalization outcomes.Here, we tested the speci c hypothesis that loss of Pink1 alters in ammation gene expression in whole blood, resulting in the upregulation of genetic pathways that begin in early-stage disease and are bioinformatically correlated to vocal communication acoustic parameters.
When the top 1000 up-and downregulated genes were put in STRING, 96 interacting genes were identi ed (Supplementary File 2).
These are replotted and shown in Fig. 1 and demonstrate enrichment for PD and prion disease.
WGCNA Supplementary File 3 includes the sortable output les, P -values, correlations, and list of genes in the top modules.There were 4 signi cant modules (ME): Red, Yellow, Midnightblue, and Purple.The top module was Red and included Pink1.To determine the genes and their functions that interact with Pink1 within this whole blood RNA dataset, the 248 genes that were in Red were put into the gene enrichment analysis tool to evaluate this speci c gene list against preexisting data sets.Several areas of enrichment were identi ed including protein catabolic process, ion homeostasis, and protein destabilization.The Yellow module signi cantly correlated to frequency modulated USV duration (length) and bandwidth (frequency range) and Yellow (335 genes) was enriched for mitochondrial gene expression.Both the Red and Yellow modules correlated to open eld number of entries (movement into the open eld indicating less anxiety; increased exploration).Other signi cant modules, Midnightblue (89 genes) demonstrated enrichment in iron ion homeostasis, as well as macrophage activation, and immune processes.Purple (152 genes) showed enrichment in multiple cellular processes.

Discussion
The general understanding of inherited, early-onset monogenic forms of PD is limited, yet necessary to provide insight into the polygenic nature of idiopathic PD as well as the development of candidate biomarkers which may be useful in early-stage diagnosis.Whole blood collection is a relatively non-invasive source of potential biomarkers that is inexpensive, easy to obtain, and has translatable clinical relevance.We hypothesized that loss of Pink1 alters in ammation gene expression in whole blood, resulting in the upregulation of genetic pathways that begin in early-stage disease.Further, we hypothesized that these dysregulated genetic pathways are bioinformatically correlated to behavioral outcomes including motor, anxiety, and vocal communication (cranial motor) acoustic parameters.The present study identi ed several dysregulated genes and biological pathways within the blood of young Pink1-/rats.These data suggest that the earliest PD signs, independent of nigrostriatal dopamine loss, are bioinformatically correlated to blood pathway data.
Loss of Pink1 results in dysregulation of ribosomal protein and RNA processing gene expression.
Consistent with previous sequencing studies on brain and TA muscle, there were notably more upregulated than downregulated DEG in Pink1-/-The most cantly downregulated genes were Rpl12 (ribosomal protein L12) and Lyzl1 (Lysozyme like protein 1).Lzyl1 is a protein coding gene that has been recently identi ed in a microarray study as a new locus associated with dementia in PD. 33 Some ribosomal proteins, such as ribosomal protein s15, have been linked to neurodegeneration in LRRK2 overexpression human dopamine neuron models. 34The most signi cantly upregulated gene in whole blood was Celf3 (CUGBP Elav-Like Family Member 3).Celf3 codes for an RNA-binding protein (RBP) involved in various aspects of RNA processing including nucleic acid binding and pre-mRNA alternative splicing.Dysregulation of RBPs has been implicated in neurodegenerative diseases including Alzheimer's disease (AD), amyotrophic lateral sclerosis (ALS), and PD.[37] Interferon signaling is altered in whole blood of Pink1-/-rats.
Loss of Pink1 results in increased production of proin ammatory cytokines and chemokines including tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6) and interleukin1-β (IL1-β), as well as interferons (IFNs) IFN-β1 and IFN-γ, within the blood and brain resulting in in ammation and loss of dopaminergic neurons in both a Pink1-/mouse model and PINK1-associated PD patients. 15FNs exhibit antiproliferative, proapoptotic, antiangiogenic and immunomodulatory functions.In cellular models of PD, loss of Pink1 also increased cytosolic dsDNA derived from mitochondria which resulted in elevated type-I IFN responses and correlated with apoptotic markers and cell death. 16pe-I IFNs (IFN-Is; IFN-α and IFN-β) play a critical role in the innate immune responses by activating classical proin ammatory signaling pathways that result in the production of major in ammatory cytokines: TNF-α, IL-6, and IL1-β.IFN-Is have been shown to regulate neuroin ammation in the central nervous system (CNS) and contribute to degeneration and disease progression in several in vivo and in vitro models of PD. 38,39 Mouse models of PD and postmortem studies of PD brains have con rmed that mRNA expression of IFN-Is is upregulated in PD.Whereas IFN-β de ciency causes mitochondrial dysfunction in primary cortical neuron cultures and causes defects in the nigrostriatal dopaminergic pathway as well as widespread α-synuclein accumulation in Ifnb-/mice. 40In addition to Type-I IFNs, Type-II IFNs have also been implicated in PD pathophysiology.For instance, IFN-γ is elevated in the serum and brain of patients with PD and correlates with disease severity. 41,42 eral genes identi ed in this dataset included those related to type I interferon signaling (I t1bl; interferon-induced protein with tetratricopeptide repeats 1B-like), type II interferon signaling (Nos2; nitric oxide synthase 2), and apoptosis (Dedd2; death effector domain containing 2).Interferon stimulated genes involved in chromatin remodeling (Gadd45a; Supt4h1; Esco2; Pelp1; Bap1; Tada2a) and ATP-binding proteins (Lig4; Slc22a5; Entpd4; Prnp; Vps25; Kifc1; Bub1; Uhrf1; Pidd1; Nek2l1; Abca4; Kif18b; Ckb) were upregulated in our dataset.Further, unpublished data from our lab show that at 12 months of age, Pink1-/rats have signi cantly more up-and downregulated genes compared to age-matched WT rats (upregulated, n = 553; downregulated, n = 1561).In this unpublished data set, numerous interferon stimulated genes (ISGs) were upregulated including Rnasel, Fas, Casp4, Irf1, and I tm1.The second most signi cantly upregulated gene is I t1 (interferon-induced protein with tetratricopeptide repeats 1; P = 5.53 x 10 − 27 ).In addition, receptors for IFN-α (Ifnar2) and IFN-γ (Ifngr1) were also upregulated in the dataset.Therefore, these data suggest that in ammation and interferon signaling begins early in the whole blood (3 months of age) and progresses as Pink1-/rats age.Further work will use bioinformatics to correlate these pathways to behavioral data at 12 months of age.
The lack of Pink1 may cause an early disruption in interferon signaling that leads to downstream overproduction of proin ammatory cytokines (TNF-α, IL-1β, IL-6) and we hypothesize to worsen over time.Targeting interferon signaling with drug compounds may be a potential therapeutic intervention to halt or prevent the further production of harmful proin ammatory cytokines that contribute to neuroin ammation and the death of neurons and should be studied in future work.
Major prion protein gene expression is upregulated.
STRING analysis showed enrichment for prion disease and PD (as in Fig. 1).One of the most interesting genes identi ed in this dataset Prnp, which was signi cantly upregulated in Pink1-/whole blood as early as 3 months of age.Prnp encodes major prion protein (PrP) that is primarily active in the brain and associated with several prion and prion-like diseases.This data further supports the prion hypothesis for PD that has been proposed due to the prion-like misfolding and aggregation of α-synuclein. 43- 48In a prion disease cell model, PINK1/Parkin signaling, speci cally PINK1, was required for mitophagy of damaged mitochondria and activation attenuates prion-induced neuronal apoptosis. 48To our knowledge, this is the rst monogenic PD animal model to report signi cant genetic changes in the Prnp gene.In this dataset, Prnp was a signi cant gene in numerous GO Biological Processes identi ed through gene enrichment of the signi cantly upregulated genes including dendritic spine maintenance, apoptotic processes, negative regulation of interleukin-17 production, T-cell receptor signaling, and calcium-mediated signaling.There were only four drug compounds identi ed to reverse upregulated gene transcription and two of them, MCSF-MCF7 and IFNG-MCF7 included Prnp as a signi cant gene.
Tuba1c, a previously identi ed signi cant gene inPink1-/-is upregulated in whole blood.
Tuba1c upregulation has been recently identi ed in several of our RNA-sequencing datasets including the thyroarytenoid (TA) vocal fold muscle and brainstem. 28Previously, Tuba1c was identi ed as key gene correlated to vocalization acoustic parameter at 2 months of age.In this study, Tuba1c was once again signi cantly upregulated in whole blood of Pink1-/rats; it is also an interconnected gene in the STRING analysis.A recent proteomics study, differential expression of Tuba1c protein was identi ed in the plasma of rotenone-exposed rats. 49Here, Tuba1c was also identi ed as a signi cant gene in the gene enrichment KEGG pathway analysis including, apoptosis, pathways of neurodegeneration, ALS, and Parkinson disease (Table 1).
Bioinformatics analysis gene pathways that cantly correlate to behavioral outcomes in Pink1-/-rats.
Another goal of this study was to use bioinformatics to highlight biological gene pathways within the whole blood and determine whether they are signi cantly correlated to anxiety, motor, or ultrasonic vocalization behavioral outcomes in Pink1-/rats.WGCNA enrichment analysis resulted in four signi cant modules, in which two of the four modules (Red and Yellow) were signi cantly correlated to behavioral outcomes.The Red module, which contained the Pink1

Conclusions
Neuroimaging and CSF biomarkers may be useful in research settings, but due to the ease, availability, and low cost of phlebotomy, whole blood biomarkers are among the most promising and practical methods to screen large populations for an occult, yet common and devastating disease with accelerating incidence.Whole blood genetic biomarkers of PD hold promise to screen large populations for PD risk factors.They may also inform prognosis as well as monitor response to future diseasemodifying treatments of PD applied in the early-stage of disease, prior to manifestation of hallmark motor signs that currently form the basis of diagnosis.Using validated, monogenic rat models we can study the in uence of Parkinsonian genes and their networks and provide data that is translatable to humans.PD has many different identi ed genes and pathways -mitochondrial dysfunction, deranged immune responses, oxidative stress, and prion protein.This study demonstrates that we can identify PD signature prior to development of a clinical motor phenotype and predict progression of ultrasonic vocalization parameters.Thus, using bioinformatics and whole blood sampling, it may be possible to identify genetic signatures in humans that correlate to vocalization dysfunction and target these gene signatures therapeutically for the treatment of vocal de cits in PD.

Gene
Abbreviations: ACSL1 = Acyl-CoA Synthetase Long Chain Family Member 1; CDKN2C = Cyclin-dependent kinase 4 inhibitor C; CKB = Creatine Kinase B; KCNN4 = Potassium Calcium-Activated Channel Subfamily N Member 4; PKIA = CAMP-Dependent Protein Kinase Inhibitor Alpha; PRNP = prion protein; TSPAN8 = Tetraspanin 8; TROAP = Trophinin Associated Protein Top 1000 genes All procedures and protocols (M006329) were approved by the University of Wisconsin-Madison School of Medicine and Public Health Animal Care and Use Committee and were conducted in accordance with the NIH Guide for the Care and Use of Laboratory animals (National Institutes of Health, Bethesda, MA, USA).BehaviorCorresponding rat behavioral data used in this study included open eld (time in center ([sec]), number of entries, total movement [(cm])), cylinder limb motor (number of rears and lands, hindlimb and forelimb movements), and ultrasonic vocalizations (total number of calls, duration ([msec]), bandwidth ([kHz]), intensity [(dB]), and peak frequency ([kHz])).These measures were used in the gene statistical correlation analysis, discussed below, and were previously published byLechner et al. 28

Table 1
Gene enrichment KEGG analysis of the DEG dataset.

Table 3
Gene enrichment GO Biological Process analysis of the signi cantly downregulated genes.

Table 5
Gene enrichment GO Biological Process analysis of the signi cantly upregulated genes.

Table 6
Gene enrichment WikiPathways analysis of signi cantly upregulated genes.
gene, correlated to open eld number of entries (movement into the open eld indicating less anxiety; increased exploration) (Table8).The Red module demonstrated enrichment in the most biological processes including cell division, chromatin organization, regulation of autophagy, cellular response to ATP and reactive oxygen species, and regulation of the cell cycle.The Yellow module signi cantly correlated to frequency modulated USV duration (length) and bandwidth (frequency range) as well as open eld number of entries.Enrichment of the Yellow module included mitochondrial gene expression, lipid transport across blood-brain barrier, tRNA processing, and IL-7 signaling.

Table 8
WGCNA modules and correlated behavioral variables.