Differential expression of haptoglobin in individuals at clinical high risk of psychosis and its association with global functioning and clinical symptoms

Background: Immune dysregulation has been observed in patients with schizophrenia or first-episode psychosis, but few have examined dysregulation in those at clinical high-risk (CHR) for psychosis. The aim of this study was to examine whether the peripheral blood-based proteome was dysregulated in those with CHR. Secondly, we examined whether baseline dysregulation was related to current and future functioning and clinical symptoms. Methods: We used data from participants of the North American Prodromal Longitudinal Studies (NAPLS) 2 and 3 (n = 715) who provided blood samples (Unaffected Comparison subjects (UC) n = 223 and CHR n = 483). Baseline proteomic data was quantified from plasma samples using mass spectrometry. Differential expression was examined between CHR and UC using logistic regression. Psychosocial functioning was measured using the Global Assessment of Functioning scale (GAF). Symptoms were measured using the subscale scores from the Scale of Psychosis-risk Symptoms; positive, negative, general, and disorganised. Three measures of each outcome were included: baseline, longest available follow-up (last follow-up) and most severe follow-up (MSF). Associations between the proteomic data, GAF and symptoms were assessed using ordinal regression. Results: Of the 99 proteins quantified, six were differentially expressed between UC and CHR. However, only haptoglobin (HP) survived FDR-correction (OR:1.45, 95 %CI:1.23 – 1.69, p adj = < 0.001). HP was cross-sectionally and longitudinally associated with functioning and symptoms such that higher HP values were associated with poorer functioning and more severe symptoms. Results were evident after stringent adjustment and poorer functioning was observed in both NAPLS cohort separately. Conclusion: We demonstrate that elevated HP is robustly observed in those at CHR for psychosis, irrespective of transition to psychosis. HP is longitudinally associated with poorer functioning and greater symptom severity. These results agree with previous reports of increased HP gene expression in individuals at-risk for psychosis and with the dysfunction of the acute phase inflammatory response seen in psychotic disorders.


Introduction
Psychotic disorders are among the most severe debilitating psychiatric disorders globally.Several studies have noted peripheral inflammatory dysregulation in individuals with both chronic and recent onset schizophrenia (Dickerson et al. (2016)), those with first episode psychosis (FEP) (Fraguas et al. (2019)) and those with FEP who were medication-naive (Upthegrove et al., 2014).A previous systematic review reported that the acute-phase response protein pathway was the most implicated in distinguishing psychiatric patients with psychosis from those without psychosis (Sabherwal et al., 2016).It has also been reported that dysregulation is evident in childhood prior to the onset of disorder (English et al. (2018)) and may be useful in predicting transition to psychosis from those at clinical high risk (CHR, Mongan et al. (2021)).
Beyond transition to psychosis, CHR status has been associated with a broad range of poorer outcomes (Addington et al., 2017;and Addington et al., 2011).Poorer functioning is evident in those with CHR who do not transition to psychosis and even in those whose psychotic symptoms subside (Addington et al., 2011).Few studies have examined inflammatory dysregulation in those with CHR relative to healthy individuals, often including short lists of well-known interleukins (Misiak et al., 2021;Stojanovic et al. (2014)).For example, results to date suggest IL6 may be elevated in those with CHR (Misiak et al., 2021).Little work has been done to examine dysregulation across the blood-based proteome (Karanikas et al., 2017;Kelsven et al., 2020, Perkins et al., 2015).Moreover, there is a dearth of studies that have examined whether inflammatory dysregulation is associated with functional and clinical outcomes in those with CHR beyond transition status, with the current evidence tentatively suggesting that these phenomena are linked (Zeni-Graiff et al.,2016;Susai et al., (2022a); Goldsmith et al. (2019); Perkins et al., 2015).
Utilising samples from two large cohort studies, the first aim of this investigation was to examine whether peripheral blood-based protein levels are differentially expressed between those with CHR relative to healthy individuals.Secondly, we examined whether the levels of differentially expressed proteins were related to concurrent and subsequent global functioning and clinical symptoms.

Participants
Participants involved in this investigation were part of two clinical high risk cohort studies, the North American Prodrome Longitudinal Study (NAPLS) 2 and the later phase of the study with a new, independent cohort, NAPLS 3.These studies have 8 and 9 sites across North America, respectively.Sites were located at Emory University, Harvard University, University of Calgary, University of California at Los Angeles, at San Diego, and at San Francisco, University of North Carolina Chapel Hill, Yale University, and Zucker Hillside Hospital.Individuals with CHR symptoms were referred to these sites from health care providers, educators, or social service agencies, or by self-referral as a result of community outreach.The recruitment procedures have been described by Addington et al. (2012) and Addington et al. (2022), for NAPLS2 and NAPLS3, respectively.
NAPLS2 consists of 720 CHR individuals while NAPLS3 consists of 560 CHR individuals (subjects' ages ranged from 12 to 35).Both studies also recruited participants who, along with first-degree relatives, were unaffected by psychiatric or neurologic disorders but who were demographically similar to those with CHR (UC, NAPLS2 n = 240; NAPLS3 n = 96).A subsample of UC and CHR participants in both studies provided blood samples which were collected at baseline.715 of these samples were used in this investigation including 375 from NAPLS2 (CHR samples = 222; and UC samples = 153) and 340 from NAPLS3 (CHR samples = 261; and UC samples = 79).Thus, 483 CHR samples and 232 UC samples were included in this investigation.
2.1.2Ethics.Appropriate ethical approval was obtained for the NAPLS2 and NAPLS3 studies at each of the study sites individually.Ethical approval for the plasma biomarker analysis in this study was obtained from the Royal College of Surgeons in Ireland research ethics committee (REC number: 202211009).et al., 2010;Woods et al., 2019) to determine if they met Criteria of Prodromal Syndromes.Transition to psychosis was defined as meeting the Presence of Psychotic Syndrome (POPS) criteria (McGlashan et al., 2010): at least one of the five Scale of Prodromal Symptoms positive symptoms rated as level 6 in intensity, for a frequency of > 1 h per day for 4 days per week during the past month; or that symptoms seriously impacted functioning to the level of severely disorganisation; or the participant presented a danger to themselves or others.

Defining Clinical High-Risk Status and
2.2.1.Mass Spectrometry.Detailed description of the sample preparation and mass spectrometry methods used in this investigation are described in a previous publication (Susai et al., (2022b)).Briefly, these samples were randomised using a block randomisation design (Burger et al., 2020).This allowed for the preservation of transition rate and the proportion of samples from each study across the randomisation sequence.Plasma samples were prepared for mass spectrometry (MS) using PreOmics kits (PreOmics GmbH).Samples were then transferred to EVOSEP tips and eluted.Samples were analysed in a Bruker tims TOF Pro mass spectrometer connected to an Evosep One liquid chromatography system which injected the samples.
2.2.2.Bioinformatics.MaxQuant was used to analyse the raw MS files and label-free quantification values were further processed with Perseus software.Proteins that were identified in more than 70 % of samples were brought forward for analysis.All proteins have been labelled in accordance with the National Centre for Biotechnology Information's Official Gene Symbol coding (https://www.ncbi.nlm.nih.gov/gene/).Missing LFQ values were treated as left-censored missing data (missing not-at-random due to quantity being below the limit of detection).That is, in our a left-censored imputation approach, missing values were replaced with values drawn from a normal distribution centred at − 3 standard deviations below the mean for a protein.This distribution allowed for variation in a ratio of 0.3 times the standard deviation of values for that specific protein.Values for each protein were log2 transformed, standardised to z-scores, and winsorised at 4 standard deviations above and below the mean.
2.3.Outcome Measures.Our outcome measures included the Global Assessment of Functioning scale (GAF, American Psychiatric Association, 1987) and the subscales scores from the four sub-scales (positive, negative, disorganized and general) of the Scale of Psychosisrisk Symptoms (McGlashan al., 2010;Woods et al., 2019).Each outcome was measured at baseline, 6 months, 12 months, 18 months, and 24 months.In scenarios where the participant transitioned to psychosis, additional measurements were captured at the time of transition.Not all participants attended all follow-up time points.Thus, for the purpose of this investigation three different approaches to outcome measures were used; Baseline, Last Follow-Up, and Most Severe Follow-Up (MSF).Baseline levels were considered the first measurement for the outcome variables.Last Follow-Up is the last known outcome scores reported by the participant after baseline.MSF is the poorest known outcome score recorded for each participant in the study after baseline.

Global assessment of functioning.
The global assessment of functioning scale (GAF, American Psychiatric Association, 1987) was used as a single measure assessing psychological, social, and occupational functioning in both cohorts.

Scale of Prodromal Symptoms.
The Scale of Psychosis-risk symptoms (SOPS; McGlashan al., 2010) was used as to measure the dimensions of psychosis.The SOPS consists of four subsections enquiring about positive, negative, general, disorganised symptoms of psychosis.

C. Healy et al.
Ratings of each symptom ranged 'Absent' (0) to 'Severe and Psychotic'/ 'Extreme' (6).For this investigation, symptom ratings in each of these sections were combined into a single section score.The sections of the SOPS are described below.

General Symptoms.
General symptoms included rating on four symptoms including sleep disturbances, dysphoric mood, motor disturbances and impaired tolerance to normal stress.

Positive Symptoms.
Positive symptoms included ratings on five symptoms including, unusual thought content, persecutory ideas, grandiosity, hallucinations, and disorganized communication.

Negative Symptoms.
Negative symptoms include ratings on six items including social anhedonia, avolition, reduced expression of emotion, reduced interpretation of own expression of emotion, reduced ideation richness and occupational functioning.

Disorganised Symptoms.
Disorganised symptoms include ratings on four symptoms including odd behaviour and appearance, trouble with focus and attention, bizarre thinking, and personal hygiene.
2.3.3Confounders.Demographic characteristics such as age and sex of the participants were also collected.Lifestyle factors and clinical characteristics included current tobacco use (yes/no), cannabis use (yes/no), alcohol use (yes/no), baseline antidepressant or anti-psychotic medication use (yes/no).Sample preparation confounders included hours to freeze the sample after draw and the time from draw to mass spectroscopy quantification.In the NAPLS3 cohort, body mass index (BMI) was also recorded.
2.4 Statistical Analysis.Demographic and lifestyle characteristics of the samples were examined using t-tests and chi 2 analysis.Comparisons between UC and CHR participants on these characteristics were conducted in the overall sample and in each NAPLS cohort separately.Logistic regression was used to examine the differential expression of each of the identified proteins between UC and CHR participants.These analyses were adjusted for age, sex and cohort.False discovery rate correction was applied to the analysis using the Benjamini-Hochberg correction method.We also ran a bootstrapped rank choice analysis, using 500 iterations.This analysis examined the consistency of the rank of the proteins based on the magnitude of the differential expression between CHR participants and controls (larger effect sizes are ranked higher).We report the median rank of each protein and the corresponding 95th percentile confidence intervals (for further information see Harrell, 2015).Two sensitivity analyses were conducted to examine if 1) any effects were observed in both cohorts separately, and 2) were the effects observed in those who transition to psychosis and those who did not transition to psychosis.
Subsequently, we examined the association between proteins with FDR-corrected differential expression and our three measures (Baseline, Last Follow-Up, and MSF) of each outcome; general, positive, negative, and disorganised symptoms and GAF.These analyses were conducted using ordinal regression and we report the proportional odds ratios.These analyses were adjusted for all confounders and cohort in the overall analysis.Sensitivity analyses were additionally conducted to examine if effects were observed in both cohorts separately.Additional adjustment for BMI was only carried out for NAPLS3 participants, as this information was not available for the NAPLS2 cohort.

Protein expression and Clinical
Outcome.Subsequent analyses investigated the relationship between HP with global functioning, general symptoms, positive symptoms, negative symptoms, and disorganised symptoms (see Table 2 for the proportional odds of the outcomes).Baseline, Last-Follow up and Most Severe Follow-up were investigated for each outcome.These analyses were conducted before and after adjustment for confounders age, sex, tobacco use, cannabis use, alcohol use, baseline psychiatric medication use, hours to freeze the sample, time between the sample freeze and mass-spectrometric protein quantification, cohort (in the overall analysis) and BMI (NAPLS3 only).

Global Functioning.
In the overall sample, HP was significantly associated with global functioning before and after model adjustment (Baseline Adj-Model p =. 004; Last Follow-up: Adj-Model p =.001; and MSF: Adj-Model p =.001).Model adjustment had limited impact on effect size (range in change of pOR: 0.05-0.06).Overall, the results indicate that a one standard deviation higher HP value is associated with a ~20% lower proportional odds on all measures of functioning (see Table 2).

Positive Symptoms.
HP was cross-sectionally (Adj-Model p =.009) and longitudinally (Last Follow-up: Adj-Model p =.004; and MSF Adj-Model p <.001) associated with positive symptoms, before and after model adjustment.Model adjustment had limited impact on effect size (range in change of pOR: 0.05-0.8).The results indicate that a one standard deviation higher HP value is associated with a 19-27% higher proportional odds in positive symptoms on all measures.
3.3.3.Negative Symptoms.HP was longitudinally, but not cross sectionally, associated with negative symptoms, before and after model adjustment (Baseline: Adj-Model p =.111;Last Follow-up: Adj-Model p =.013; and MSF Adj-Model p =.009).Model adjustment had limited impact on effect size (range in change of pOR: 0.06-0.09).The longitudinal results indicated that a one standard deviation higher HP value at baseline was associated with a ~20% higher proportional odds in subsequent negative symptoms.

General Symptoms.
HP was significantly cross-sectionally associated with baseline general symptoms (Adj-Model p =.002).Model adjustment had limited impact on effect size (range in change of pOR: 0.08-0.11).HP was longitudinally associated with both Last Follow-Up general symptoms and MSF general symptoms before and after adjustment (Last Follow-up: Adj-Model p <.001; and MSF Adj-Model p <.001).The results indicate that increasing HP value by one standard deviation is associated with 23-33% higher proportional odds on all measures of general symptoms from the SOPS.
3.3.5.Disorganised Symptoms.HP was cross-sectionally (Adj-Model p =.045) and longitudinally (Last Follow-up: Adj-Model p =.009; and MSF Adj-Model p =.005) associated with disorganised symptoms before and after model adjustment.Model adjustment had limited impact on effect size (range in change of pOR: 0.06-0.09).The results indicate that a one standard deviation higher HP value is associated with a 16-23% higher proportional odds in disorganised symptoms on all measures.
3.4.Cohort Sensitivity analysis.We examined the association between HP and each outcome in NAPLS2 and NAPLS3 separately.Despite the reduction in power, many of the associations between HP with functioning and general symptoms were retained, with broadly similar effect sizes observed in both cohorts.The reduction in power was more consequential to an association between HP with specific symptom subscales; positive, negative, and disorganised symptom scales.In both cohorts, HP was significantly associated with most of the MSF outcomes on each of these symptom scales.Reduced associations were retained for the baseline and last follow up outcomes on these scales.The effect sizes in both cohorts were broadly similar to the overall effects, suggesting that loss of association is a result of the widening of the confidence interval due to a loss of power.

Discussion
Using the largest proteomic comparison between CHR and UC to date, our results indicate that HP levels are elevated in those at CHR compared to UC.Additionally, higher HP levels are associated with poorer global functioning and greater positive, negative, disorganised, and general symptoms.These findings were evident cross-sectionally and longitudinally.Furthermore, the findings relating to global functioning and general symptoms were evident in both cohorts separately, highlighting the robustness of the finding.HP levels were upregulated in those with CHR for psychosis, irrespective of transition status.Coupling this with the knowledge that the vast majority of those with CHR meet criteria for depression or anxiety, the findings suggest that HP may be a non-specific biological marker of poorer functioning and general symptoms.There is increased circulating HP level in individuals with depression (Zorrilla et al., 2001).Our results support the hypothesis that pro-inflammatory acute phase response processes are represented in the extended psychosis phenotype.
A previous review article has indicated that the acute phase immune response is the most dysregulated pathway in patients with schizophrenia (Sabherwal et al., 2016).HP is an acute phase protein which is elevated during inflammation response.It is a primary binding protein for haemoglobin in the blood, which has anti-inflammatory properties (Polticelli et al., 2008).Taken together, the results suggest an overactivation of the acute inflammatory response in those with CHR that is related to clinical presentation.Moreover, our results support a previous smaller study specifically demonstrating that CHR patients had higher HP gene expression assay levels than healthy controls (Yee, et al., 2022).However, while observing a relationship between HP levels and cognition, Lee and colleagues' study did not find a significant association between HP levels with global functioning or symptoms.Thus, our investigations expand on this research by robustly indicating that HP is associated with current and subsequent global functioning and symptoms.It should be noted that the degree of dysregulation observed in the CHR samples was in line with low-level dysregulation, rather than severe inflammatory dysregulation as may occur in the context of acute infection or injury (Müller (2018)).
Other proteins (SERPIND1, SERPINA6, APCS, ITIH2 and ITIH1) involved in immune processes were differentially expressed.However, these did not survive correction for multiple comparisons nor were they consistently ranked highly within a bootstrapped rank order analysis.Few studies have investigated proteomic discrimination between participants with CHR and UC, with most of these concentrating on cytokines and C-reactive protein (Karanikas et al., 2017;Delaney et al., 2019).Meta-analytic evidence suggests that IL6 and IL1B are differentially expressed in those with CHR from healthy controls (Park & Miller, 2020;and Misiak et al., 2021).However due to the discovery nature of our mass spectrometry investigation and the low abundances of cytokines within peripheral blood samples, these proteins were not identified within the current samples, possibly as they were below the limit of detection.Thus, while we are confident in the findings regarding HP and the other measured proteins, our results do not preclude differential expression of other proteins which were not quantified in this investigation.
Apart from the increased HP levels in CHR of psychosis population, our study results showed that HP levels in the plasma could be a potential marker for poor functioning and higher symptomology.In humans, HP is coded by genes located on chromosome 16q22, and allelic combinations of these genetic information produces three major polymorphisms of HP protein (HP1-1, HP1-2 and HP2-2) (Bensi et al., 1985).In general HP is involved in various biological functions which include oxidative stress, immune reactions, angiogenesis and nitrous oxide metabolism (Polticelli et al., 2008).Previous papers provide preliminary evidence that micro and macro level biological properties of HP could be involved in the pathophysiology of psychosis.Particularly important evidence arises from the involvement of inflammatory pathways that link HP with psychosis (Wang et al., 2001;Yang et al., 2000;Arredouani et al., 2003;Guetta et al., 2007).Inflammatory cytokines such as IL6 and TNF are well known for their involvement in pathophysiology of psychosis as well as production of HP by liver (Wang et al., 2001;Yang et al., 2000).This immune relationship appears to be bidirectional as the HP also influences the production of IL6 by T-helper cells (Arredouani et al., 2003;Guetta et al., 2007).
In addition, HP related effects on nitrous oxide cause narrowing of blood vessels that can result in compromised circulation in internal organs including brain.Previous studies have reported higher levels of serum HP in diabetes and neurodegenerative disorders such as Alzheimer's disease and Parkinson's disease which share common neurovascular features such as brain atrophy, reduced cerebral glucose metabolism, and central nervous system insulin resistance (Verdile, Fuller & Martins, 2015;Rosales-Corral et al., 2015).From our results we speculate that HP-associated functional deterioration and clinical symptoms could be due to vascular and metabolic derangements in the brain, as observed in other neurodegenerative disorders.This opens up a future research option in exploring biological mechanisms underpinning the HP associated psychopathological alterations that predispose to psychosis.In support of this hypothesis, elevated HP has been reported in first-episode psychosis and schizophrenia, and was associated with depression and excitement symptoms on the Positive and Negative Syndrome Scale (PANSS) (Seal & Eist, 1966;Bock, Weeke & Rafaelsen, 1971;Maes et al., 1997;Yang et al., 2006;and Wan et al., 2007).Furthermore, a recent study by Yee et al (2022) found gene expression levels in CHR participants were elevated relative to healthy controls.
Since the biological activity varies across different genotypes of HP, identification of specific genotypes in relation to psychosis could be therapeutically valuable for determining preventive measures in CHR psychosis.

Strengths and limitations.
This is a large investigation utilising biological data from two observational multi-site cohort studies both based in North America.Our results were robust to stringent adjustment for demographic, lifestyle, and sample preparation confounding.Our work adds to limited current knowledge on the biological discrepancies between those at CHR and UC.Such investigations of those meeting CHR criteria were requested in a recent review of the immunophenotype of schizophrenia (Miller et al. (2003) & Goldsmith, 2016).Our study has limitations.Given the large number of biological samples and the requirement of quantification in 70% of the observation, our proteomic profile for participants is relatively shallow (k = 99).This leaves open the possible differential expression of other proteins.Conversely, the number of proteins measured required quite a significant multiple comparison correction leaving open the risk of type II error.Not every participant from the studies will have given blood, suggesting room for representation bias.However, our comparison between those with CHR who did and did not provide a blood sample indicates little to no difference between these two groups (see Supplementary Table 3).Samples from NAPLS2 did not have BMI data, thus we are unable to adjust for it in the main analysis.Thus, we did use BMI as an additional confounder in the stratified analysis using the NAPLS3 cohort.Finally, while we observe an association between haptoglobin and functioning within the samples our result should not be interpreted that lower haptoglobin causes a reduction in functioning but more that lower haptoglobin appears to cooccur and precede lower functioning and higher symptom scores.

Conclusion
We found robust evidence that HP is elevated in those with CHR relative to UC. HP levels were associated with current and future global functioning and clinical symptoms.In line with previous research, our results suggest that acute phase inflammatory response dysfunction, seen in psychotic disorders, is also observable in those at risk for psychosis, irrespective of transition status.This suggests that such dysfunction may be a transdiagnostic marker of distress and poorer functioning.

Transition to Psy- chosis.
Participants aged between 12 and 35 years undertook the Structured Interview for Psychosis-risk Syndromes (SIPS, McGlashan

Table 1
Demographic characteristics of UC and participants at CHR in the overall group and stratified by cohort.

Table 2
Association between HP and the different outcome measures in the overall sample and when stratified by cohort.