Plasma and Cerebrospinal Fluid Inammation and the Blood Brain Barrier in Older Surgical Patients: The Role of Inammation after Surgery for Elders Study

Our understanding of the relationship between plasma and cerebrospinal uid (CSF) remains limited, which poses an obstacle to the identication of blood-based markers of neuroinammatory disorders. To better understand the relationship between peripheral and central nervous system (CNS) markers of inammation before and after surgery, we aimed to: examine whether surgery compromises the blood-brain barrier (BBB), evaluate postoperative changes in inammatory markers, and assess the correlations between plasma and CSF levels of inammation.


Abstract Background
Our understanding of the relationship between plasma and cerebrospinal uid (CSF) remains limited, which poses an obstacle to the identi cation of blood-based markers of neuroin ammatory disorders. To better understand the relationship between peripheral and central nervous system (CNS) markers of in ammation before and after surgery, we aimed to: examine whether surgery compromises the bloodbrain barrier (BBB), evaluate postoperative changes in in ammatory markers, and assess the correlations between plasma and CSF levels of in ammation.

Methods
We examined the Role of In ammation after Surgery for Elders (RISE) study of adults aged ≥ 65 who underwent elective hip or knee surgery under spinal anesthesia who had plasma and CSF samples collected at baseline and postoperative 1 month (PO1MO) (n = 29). Plasma and CSF levels of three in ammatory markers previously identi ed as increasing after surgery were measured using enzymelinked immunosorbent assay: interleukin-6 (IL-6), C-reactive protein (CRP), and chitinase 3-like protein (also known as YKL-40). Integrity of the BBB was computed as the ratio of CSF/plasma albumin levels (Qalb). Mean Qalb and levels of in ammation were compared between baseline and PO1MO. Spearman correlation coe cients were used to determine correlation between bio uids. For the plasma-CSF bio uids with signi cant correlations, we determined whether the markers were associated by using linear regression models.

Results
Mean Qalb did not change between baseline and PO1MO. Plasma and CSF levels of IL-6, CRP, and YKL-40 were higher on PO1MO relative to baseline, with a disproportionally higher increase in CSF levels relative to plasma levels (IL-6 doubled and CRP tripled in CSF). Signi cant plasma-CSF correlations for CRP (baseline r = 0.70 and PO1MO r = 0.89, p < .01 for both) and IL-6 (PO1MO r = 0.48, p < .01) were observed, with higher correlations on PO1MO compared with baseline.

Conclusions
In this elective surgical sample of older adults, BBB integrity was similar between baseline and PO1MO, in ammation levels were higher PO1MO than baseline, and plasma-CSF correlations were observed for CRP and IL-6. Our identi cation of potential promising plasma markers of in ammation in the CNS may facilitate the early identi cation of patients at greatest risk for neuroin ammation and its associated adverse cognitive outcomes.

Page 4/13
Background Insights into the relationship of plasma and cerebrospinal uid (CSF) biomarkers may help to advance our pathophysiologic understanding of how peripheral events like surgery might trigger neuroin ammation. Our understanding of the inter-relationship between plasma and CSF remains incomplete, and as a consequence, our ability to identify blood-based biomarkers of the surgical effect, including in ammation and neuroin ammation, remains limited. Recent data support an emerging model highlighting the potential associated downstream consequences of surgery: individuals predisposed to a heightened in ammatory response when exposed to an acute stressor, such as surgery or infection, are at increased risk for longer-term adverse outcomes (Marcantonio, 2012). Under certain conditions, these systemic in ammatory mediators are hypothesized to activate brain microglia, leading to neuroin ammation, which if sustained, may cause permanent neuronal injury (as illustrated in Marcantonio, 2012).
From this posited neuroin ammatory hypothesis, two prominent questions remain. First, how does an acute stressor such as surgery impact the BBB after a prolonged period of time? Second, can we identify blood-based markers indicative of neuroin ammation? Such questions require understanding the complex nature of molecular dynamics underscoring the protein levels in peripheral blood and the central nervous system (CNS), which can be in uenced by many factors.
To better understand the complex relationship between peripheral and CNS markers of in ammation before and after surgery, we evaluated three key knowledge gaps in a study of older adults undergoing major elective surgery under spinal anesthesia. Our speci c aims were: (1) to examine whether surgery compromises the blood-brain barrier (BBB, measured from CSF-plasma albumin ratio [Qalb]) one month post-surgery; (2) to evaluate whether changes in levels of in ammatory markers following surgery are greater in plasma and CSF at one month post-operation (PO1MO) relative to preoperation (PREOP or baseline); and (3) to examine 3 previously identi ed plasma markers of in ammation associated with surgery for their correlations with CSF levels: interleukin (IL)-6, C reactive protein (CRP) and chitinase-3 Like Protein 1 (CHI3L1, also known as YKL-40). We hypothesized that: 1) surgery would compromise the BBB, resulting in higher permeability of proteins observed one month following surgery; 2a) mean PO1MO levels of all three in ammatory markers would be higher on PO1MO compared to PREOP; 2b) CSF levels would exhibit disproportionately higher increases in these in ammatory markers following surgery compared with plasma; 3a) plasma markers of in ammation would have high correlations with CSF, and 3b) higher plasma-CSF correlations would be observed on PO1MO compared with PREOP.

Study Sample
The Role of In ammation after Surgery in Elders (RISE) study is a cohort study aimed to assess the correlation of blood plasma, CSF, and imaging biomarkers of in ammation in patients aged 65 year or older who underwent elective hip or knee arthroplasty under spinal anesthesia. The overall study design and protocol have been described (Hshieh et al., 2019). Brie y, patients were enrolled if they had planned admission for at least 24 hours and surgery scheduled at least 15 days in advance to allow time for preoperative testing. Exclusion criteria have been previously described, and include safety exclusions for lumbar puncture and magnetic resonance imaging.
The Institutional Review Board of Partners Healthcare System (Massachusetts General Hospital, Brigham and Women's Hospital, Brigham and Women's Faulkner Hospital) approved all study procedures, with ceded review from Beth Israel Deaconess Medical Center and Hebrew SeniorLife, the study coordinating center.

Specimen Collection
Phlebotomy was performed on patients at three time points: baseline (at home or during preadmission testing clinic visit), postoperative day 1 (POD1), and approximately one month postoperatively (PO1MO).
CSF was acquired preoperatively during induction of spinal anesthesia (baseline) and at PO1MO via lumbar puncture. CSF was collected via dropwise collection or aspiration directly into collection tubes. To minimize potential blood contamination of the CSF, samples were centrifuged at 1000 rcf for 10 minutes prior to storage at -80 o C in low absorption polypropylene tubes. This paper focuses on the baseline and PO1MO time points since CSF was only available at these two time points.

Immunoassays
Plasma and CSF levels of three in ammatory markers (IL-6, CRP, and YKL-40) and of albumin were measured using sandwich assays: the fully automated Ella System (ProteinSimple San Jose, CA) for the in ammatory markers, and an enzyme-linked immunosorbent assay from Abcam (Cambridge, MA; ab108788) run on a semiautomatic Tecan Freedom Evo liquid handling platform (Männedorf, Switzerland) for albumin. Qalb was de ned as CSF albumin/plasma albumin x 10 − 3 . Coe cient of variations (CVs) of duplicate measures were generally < 5%. If a CV was > 10%, the assay was repeated.

Statistical Analysis
To determine correlation within bio uids (i.e., plasma-plasma or CSF-CSF) and between bio uids (i.e., plasma-CSF), we examined Spearman correlation coe cients. All analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC). Table 1 reports the characteristics of our study sample, presenting means and standard deviations, as well as proportions, for the sample of patients with complete plasma and CSF biospecimen data at both baseline and PO1MO (n=29). Patients were on average age 75 and mostly female, 10% had ≥2 Charlson comorbidities, and all patients underwent spinal anesthesia alone. Fifty-two percent of patients underwent total knee arthroplasty and 48% underwent total hip arthroplasty. The average hospital length of stay was 3.2 days. Table 2 reports the distributions of albumin, IL-6, CRP, and YKL-40 in the sample with complete plasma and CSF biospecimen data (n=29). Between baseline and PO1MO, we observed three general patterns. First, mean Qalb was similar between the baseline and PO1MO time points (6.42 for both); a small number of patients had Qalb ≥9.0 (4 patients [14%] at baseline and 2 patients (7%) on PO1MO, including the one patient having Qalb ≥9.0 at both time points), indicating compromised BBB (Bowman et al., 2007). Second, IL-6 and CRP levels increased for both plasma and CSF. Average plasma IL-6 levels increased from 3.7 pg/ml (baseline) to 5.4 pg/ml (PO1MO), and average CSF IL-6 doubled from 4.0 pg/ml (baseline) to 8.3 pg/ml (PO1MO). Similarly, increases in CRP levels were observed between the two time points: average plasma CRP levels increased from 7.4 mg/l (baseline) to 8.2 mg/l (PO1MO), and average CSF CRP tripled between baseline (0.01 mg/l) and PO1MO (0.03 mg/l). Third, YKL-40 increased in plasma only (96.8 ng/ml [baseline] to 135.7 ng/ml [PO1MO]), while CSF levels remained near equivalent at baseline (280.1 ng/ml) and PO1MO (280.0 ng/ml). When we conducted the same analyses using the full dataset of RISE study patients with blood and CSF at PREOP (n=57) and at PO1MO (n=42) or without blood and/or CSF at either time point, the overall ndings remained similar (results available upon request). Table 3 reports the Spearman correlation coe cients of plasma and CSF in ammatory markers at baseline (n=29). Between bio uids (plasma and CSF; shown in red), three signi cant correlations were identi ed --all including CSF levels of CRP with plasma levels at baseline of: itself (CRP: r=0.70, p<.01), IL-6 (r=0.58, p<.01), and YKL-40 (r=0.43, p<.05). Table 4 lists the Spearman correlation coe cients of plasma and CSF in ammatory markers at PO1MO (n=29). Similar to baseline, IL-6, YKL-40, and CRP plasma levels were signi cantly correlated with CRP CSF levels, all assessed at 1 month (r=0.47, 0.59, and 0.89, respectively). Additionally, IL-6 plasma and IL-6 CSF levels were moderately correlated at P01MO (r=0.48, p<.01).

Results
We additionally examined the Spearman correlation coe cients of plasma and CSF in ammatory markers between baseline and PO1MO. Within both plasma and CSF, every protein was signi cantly correlated with itself between the two time points. Spearman r's for plasma IL-6, CRP, and YKL-40 measured on baseline and PO1MO were: 0.49, 0.52, and 0.80 (p<.01 for all); Spearman r's for CSF IL-6, CRP, and YKL-40 measured on baseline and PO1MO were: 0.50, 0.58, and 0.93 (p<.01 for all). The only signi cant correlation between bio uids at the two time points was observed between baseline levels of YKL-40 plasma and PO1MO CSF levels of CRP (r=0.51, p<.01).

Discussion
In this study of older adults undergoing major noncardiac surgery, we found evidence in support of two of our three hypotheses (summarized in Fig. 1). Our rst hypothesis was not supported; we need did not observe compromises in the integrity of the BBB between PREOP and PO1MO based on the CSF-albumin ratio (Hypothesis 1). For our second hypothesis, we found that both plasma and CSF mean levels of IL-6, CRP, and YKL-40 were higher on PO1MO relative to PREOP levels (Hypothesis 2a). Additionally, there was a disproportionately higher increase in CSF levels relative to plasma levels between PREOP and PO1MO for: IL-6 (doubled in CSF) and CRP (tripled in CSF) (Hypothesis 2b). For our third hypothesis, we observed signi cant plasma-CSF correlations for CRP (PREOP and PO1MO) and IL-6 (PO1MO only) (Hypothesis 3a), with higher correlations on PO1MO compared to PREOP (Hypothesis 3b).
At both baseline and PO1MO, we found that plasma and CSF levels of CRP were well correlated, and that levels of plasma IL-6 and levels of plasma YKL-40 were correlated with CSF CRP. At PO1MO only, plasma and CSF levels of IL-6 were correlated. Despite these between bio uid correlations, we were ultimately unable to identify promising blood-based in ammatory markers of CNS in ammation, based on our knowledge of the origins of these in ammatory markers. As an example, since CRP is felt to be solely produced in the periphery (predominantly in the liver), it is more probable that our ndings re ect a 'leaking' of CRP from the periphery (i.e., plasma) into the CNS (i.e., CSF) than 'leakage' from the CNS into the periphery. Interestingly, the increased levels of CSF in ammatory proteins at PO1MO occurred despite no change in Qalb, suggesting that BBB integrity for other proteins may not be fully re ected in this measure. Given this, it is unlikely that plasma CRP, IL-6, or YKL-40 are promising blood-based markers of CSF in ammation. We also found that: (1) peripheral markers of in ammation correlated well with other peripheral in ammatory markers within the same time point, and (2) for plasma and CSF, each in ammatory marker was correlated with itself between the baseline and PO1MO time points.
Although not observed in our study, compromises in BBB have been associated with surgery in humans and animal models. In mice, anesthesia and/or surgery may induce age-associated BBB permeability, as determined by immunohistochemistry imaging and spectrophotometric quanti cation (Yung et al., 2017). Among patients undergoing cardiac surgery, postsurgical disruption of the BBB detected using magnetic resonance imaging (MRI) was observed in 47% of the 19 patients, all of whom had no clinical evidence of a stroke or delirium at the time of gadolinium administration or the MRI scan (Merino et al., 2013). The absence of changes in BBB permeability between baseline and PO1MO in our healthy sample of older RISE patients may be due to the fact that post-surgery disruptions in the BBB may have resolved by the 1 month post-surgery time point or may be related to the poor sensitivity of Qalb to detect changes in BBB permeability. Ideally, we would have examined an alternate measure of BBB integrity or Qalb closer to the surgical event (e.g., postoperative day 1 [POD1]); however, we were limited in the measures we could assay and were unable to obtain CSF at the POD1 time point.
Plasma and CSF correlations of in ammatory markers seem to be dependent on the characteristics of the study sample. Among 141 patients with Alzheimer's disease (AD), a good correlation between plasma and CSF levels of IL-6 levels was reported (r = 0.76, p < .001) (Sun et al., 2003). In contrast, no correlation between IL-6 plasma and CSF levels was observed in 173 older adults who were asymptomatic for AD (r = 0.16, p = .05) (Bettcher et al., 2018). Our ndings in older surgical adults (none of whom had known preexisting dementia, yielded a plasma-CSF correlation of IL-6 levels that was between the Sun et al. (2003) and Bettcher et al. (2018) publications (r = 0.48, p < .01 on PO1MO). This suggests that, for IL-6, the relationship between plasma and CSF may be in uenced by factors associated with cognition (e.g., BBB integrity or presence of AD), although we were unable to adequately probe this possibility within our small sample. Although Qalb did not increase between baseline and PO1MO in our sample, the plasma-CSF IL-6 correlation and association ndings suggest the possibility of: (1) increased permeability for small proteins, such as IL-6 (approximately 20 kDa in size versus, for example, the size of albumin [about 60 kDa]), (2) IL-6 may be stimulating CNS IL-6 production indirectly (irrespective of the integrity of the BBB), or (3) stimuli that induce IL-6 expression in peripheral blood mononuclear cells also induces the expression in microglial cells since plasma and CSF levels of IL-6 were observed to be relatively similar at baseline.
Among the in ammatory markers examined, we found preliminary evidence for the possibility of stronger plasma-CSF correlations at the PO1MO time point. For instance, CRP exhibited medium to high correlations at baseline (r = 0.70, p < .01) and extremely high correlation at PO1MO (r = 0.89, p < .01). These promising ndings highlight the importance of further examining these relationships in larger surgical cohorts.
We highlight several study strengths. RISE applied state-of-the art approaches to the collection of biospecimens and detailed clinical data on older patients undergoing major surgery, including collection of plasma and CSF. Our empirically driven analysis examined potential correlations between plasma and CSF, with a focus on the longer-term effects of surgery (one month post-operation). This facilitated further probing of correlations and the integrity of the BBB following surgery over a longer time frame than may not have been previously observed in the literature.
Some study limitations warrant mention. First, this is a relatively small study which will need con rmation; yet it provides important descriptive information to better understand BBB in the context of the longer-term effects of surgery and to identify peripheral markers of neuroin ammation. Given the sample size, we did not adjust for multiple testing, control for covariates, or examine alternate means of considering nonlinearities in the relationship between plasma and CSF levels in our regression models. We ultimately employed a biomarker "correlational" discovery approach to generate hypotheses of peripheral-CNS relationships that will require con rmation in larger cohorts, for which we are currently enrolling and intend to pursue more rigorous statistical analyses in future work. Second, although we did not observe a notably strong correlation between plasma and CSF in ammatory markers in this small, selected sample, we acknowledge that an absence of a correlation does not indicate an inherent lack of peripheral markers of neuroin ammation. It may be that our limited set of in ammatory markers does not appropriately represent indicators of in ammation in the CNS. For instance, macrophage in ammatory protein (MIP-1β, also known as CCL4), previously observed to be moderately correlated in plasma and CSF among older adults without AD (r = 0.55, Bettcher et al., 2018), may be a promising additional in ammatory marker for future examination. Ultimately, the identi cation of a blood-based marker of neuroin ammation will require identi cation of an in ammatory marker that is produced entirely, or nearly entirely in the CNS (and not in the periphery). Third, at baseline, there was more variability in the time interval between blood and plasma acquisition compared to the collection of both bio uids at PO1MO, which was almost always done at the same time. As previously noted, this may explain the generally stronger correlations observed on PO1MO relative to baseline. Fourth, since all patients underwent spinal anesthesia, we cannot rule-out possible protective effects where spinal anesthesia may mediate the relationships observed. Last, we use CSF levels of in ammation as the "gold standard" for neuroin ammation given the absence of an alternative approach, such as brain imaging markers or brain tissue in the RISE study.
In summary, we found that all three in ammatory markers were higher at 1 month post-surgery than at baseline, with a higher increase in CSF levels relative to plasma levels between the two time points. In contrast, the integrity of the BBB was similar between the two time points. The plasma-CSF correlation results suggest that CRP plasma-CSF correlations are high at all time points, likely due to leakage of peripheral CRP into the CNS. Alternatively, IL-6 measured on PO1MO may be a promising blood-based marker of neuroin ammation. Future studies in larger surgical populations that further probe the possibility of IL-6 and other potential blood-based markers of in ammation in the CNS will facilitate the identi cation of blood-based markers of neuroin ammation and understanding of CNS in ammatory disorders. Our ability to identify such a marker would facilitate the early detection of patients with greatest risk for neuroin ammation and its associated adverse cognitive outcomes.   Abbreviations: CSF=cerebrospinal uid, PO1MO=postoperative 1 month, SD=standard deviation Table 3 Spearman correlation coe cients of bio uids measured at baseline (n=29)