Evaluating the levels of CSF and serum factors in ALS

Abstract Objectives The aim of this study was to identify CSF and serum factors as biomarkers that may aid in distinguishing ALS patients from control subjects and predicting ALS progression as well as prognosis. Methods Serum and CSF samples from 105 patients with ALS and 56 control subjects were analyzed for 13 factors using ELISA. The revised ALS functional rating scale (ALSFRS‐r) was used to evaluate the overall functional status of ALS patients, and we also followed up with ALS patients either by phone or with clinic visits for five years after enrollment in this study. Finally, we examined the correlations between factor levels and various clinical parameters and evaluated the predictive value for prognosis through a multivariate statistic model. Results A total of eight factors were obviously elevated in CSF, and twelve markers were increased in serum. In the correlation analyses, there were trends toward higher bFGF, VEGF, MIP‐1α levels in ALS with a longer disease duration and slower disease progression in both CSF and serum. Higher MCP‐1 levels were associated with worse disease severity and faster progression, and the IFN‐γ levels were positively associated with disease progression in either CSF or serum. Finally, a better prognosis was observed with higher levels bFGF in CSF and VEGF in CSF and serum; conversely, patients with higher levels of IFN‐γ in the CSF had shorter overall survival. Conclusions We demonstrated that a factor profile of ALS patients is distinct from control subjects and may be useful in clinical practice and therapeutic trials.


| INTRODUCTION
Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder that affects motor neurons in the brain and spinal cord leading to paralysis and eventually death (Hardiman, Van den Berg, & Kiernan, 2011). There have been various ALS biomarker studies published with more ongoing, with special attention focused on identifying biomarkers in CSF or serum. Most of these studies have examined changes of individual biomarkers that might distinguish patients with ALS from healthy control subjects Takahashi et al., 2015), but it is likely more appropriate to identify panels of biomarkers rather than focusing on a single factor to increase their specificity to ALS. Some studies were limited by the number of samples in the analysis , and some chose a different control population into the study (Tateishi et al., 2010). In addition, a panel of biomarkers has been detected in either the CSF or serum in some recent studies, but they were either more concerned with a diagnosis rather than a clinically relevant prognosis Kuhle et al., 2009;Lawton et al., 2012) or only measuring factors either the CSF or serum (Ehrhart et al., 2015;Mitchell et al., 2009).
We previously reported that several cytokines are elevated in the CSF and serum of ALS patients (Gao, Zhou, Cai, Gong, & Zang, 2014;Gong, Gao, Guo, Lu, & Zang, 2015;Liu, Gao, & Zang, 2015;Yang, Gao, Wu, Zhang, & Zang, 2016). Here we expand on our earlier results by measuring a panel of factors in the CSF and serum from a larger group of ALS patients to intuitively and comprehensively show the differences in the factor levels between ALS patients and control subjects. Then, we further analyzed the associations between the factor levels and clinical parameters to identify biomarkers related to disease progression and prognosis.
Finding reliable biomarkers in ALS is valuable to provide an early diagnosis and the preliminary basis for disease pathogenesis in future studies.  (Brooks, Miller, Swash, & Munsat, 2000), which included clinically definite or probable ALS.

| Subjects
ALS patients with any medical condition related to motor neuron dysfunction or who were taking any medicine were excluded. Disease duration was defined as the interval between onset of symptoms and diagnosis. The revised ALS functional rating scale (ALSFRS-r) is a well-established scale to evaluate overall function of ALS patients and is scored from 0 to 48, with lower scores indicating poorer function (Gladman, Dehaan, Pinto, Geerts, & Zinman, 2014). The disease progression rate (DPR) was calculated using the following formula: DPR = (48 -ALSFRS-r score at time of diagnosis)/disease duration (months). We also followed up ALS patients either by phone or with clinic visits for 5 years after enrollment in this study, and the primary endpoint was death. Survival time was defined as the interval between the time of diagnosis and death from confirmed ALS-related complications. Fifty-six control subjects were diagnosed with tension-type headache (n = 34), hypokalemic paralysis (n = 8), cerebrospinal fluid leakage (n = 4), or low intracranial pressure (n = 10).
However, we excluded controls with any life-threatening disease or who were taking any medicine. The demographic features, including age, gender, and body mass index (BMI), did not show significant diversity between the ALS patients and control subjects (p > .05). The demographic and clinical characteristics are summarized in Table 1.

| Serum and CSF samples collection
Blood samples were available from all the patients by venipuncture and centrifuged immediately to isolate serum. The supernatant was stored at −80°C until further use. CSF samples were simultaneously obtained by lumbar puncture from 81 ALS patients and 56 controls and quickly centrifuged prior to storage of the supernatant at −80°C. Blood and CSF were collected between 8 AM and 12 PM to limit circadian effects.

| Statistical analysis
Analyses were conducted using GraphPad Prism Version 5.0 (RRID:SCR_002798) and SPSS Statistics version 17.0 (RRID:SCR_002865). Continuous demographic and clinical data were described as the means ± standard deviation (SD) in the tables. Either a two sample t-test or Mann-Whitney U-test was used to test the differences in factor levels between the groups. Correlations between the factor levels and various clinical parameters were assessed by either Pearson or Spearman correlation analysis. Univariate and

| Factors level in CSF and serum
First, to identify factors that are obviously different between ALS patients and control subjects, we detected 13 factors in serum and CSF samples using ELISA. The analytical results of the factors are detailed in Table 2. In the CSF, IL-15, IL-17, bFGF, VEGF, MIP-1α, MIP-1β, MCP-1, and IFNγ were significantly elevated in ALS patients compared with control subjects. In contrast, there were more factors notably altered in serum; among these, IL-2, IL-6, IL-10, IL-15, IL-17, G-CSF, GM-CSF, bFGF, VEGF, MIP-1α, MCP-1, and IFNγ levels were strongly increased in ALS patients. The levels of other factors showed no significant differences between the two groups.

| Correlations between factors level and various clinical parameters
To further assess whether the changes in these factors had diagnostic value, we performed a series of correlation analyses between the factors and clinical parameters, including disease duration, ALSFRS-r scores and DPR. We found that bFGF, VEGF, and MIP-1α levels in the CSF and serum were positively correlated with disease duration (Table 3 and Figure 1), whereas opposing results were obtained when their correlation with DPR was analyzed (Table 5 and Figure 3). The levels of MCP-1 in the CSF and serum were negatively correlated with the ALSFRS-r scores of ALS patients (Table 4 and Figure 2); however, these levels showed a positive correlation with DPR (Table 5 and

| Predictors of survival in ALS
After the aforementioned analysis, several factors have exhibited obvious connections with disease severity and progression. Thus, The bold values indicates that the results are statistically significant.

| DISCUSSION
In this study, we utilized ELISAs to identify a panel of serum and CSF factors for ALS, with five significant findings as follows: (i   whereas only IL-15 and IL-17 were clearly increased in the CSF. One possible explanation for the difference observed from the CSF and serum is that IL-2, IL-6, and IL-10 might have a limited input in central nervous system (CNS) during the inflammatory process; as one previous study has shown, these cytokines are at similar levels in the CSF of ALS patients and control subjects (Tateishi et al., 2010). Furthermore, even IL-2 and IL-10 could not be detected in the CSF of ALS patients (Holmoy, Roos, & Kvale, 2006). In addition, no correlations were discovered between these interleukin levels and the clinical parameters in our analyses, although IL-10 in serum was reported as a predictor of total disease duration (Su et al., 2013). These results may suggest that these interleukins were released erratically in response to the inflammation caused by ALS as systemic inflammation factors and have low susceptibility and specificity as biomarkers for ALS.
The increase in serum G-CSF and serum GM-CSF discovered in our research also weakly points to changes of neuroinflammation in ALS. G-CSF is a traditional hematopoietic growth factor that was approved for granulocytopenia therapy. In addition to affecting the hematopoietic system, G-CSF can exert anti-inflammatory effects as well as promote neurogenesis in the CNS (Pitzer et al., 2008).
However, GM-CSF is regarded as a proinflammatory cytokine to   (Barbeito, Mesci, & Boillee, 2010). In our study, we could not detect any significant difference in the CSF between ALS patients and control subjects, nor was there any correlation with either disease progression or prognosis. Although elevated G-CSF and GM-CSF in the CSF were noted in several studies (Mitchell et al., 2009;Tateishi et al., 2010), their predictive and therapeutic values in ALS were also preliminarily evaluated (Su et al., 2013;Zhang et al., 2009).
Differences in race or patient selection among the studies may be cause of these discrepancies. Further studies with a large-scale population to investigate the exact changes and effects of G-CSF in ALS are forthcoming.
Compared to unobvious alterations of the G-CSF and GM-CSF, three chemokines in our study emerged with strong differences between ALS patients and control subjects. Chemokines are a group of secreted proteins whose primary function is to induce inflammatory cell migration (Ramesh, MacLean, & Philipp, 2013). Interestingly, MIP-1α and MCP-1 showed inverse patterns with regard to the clinical parameters: MIP-1α was positively associated with disease duration and negatively related with DPR, whereas MCP-1 exhibited a negative correlation with disease severity and was positively correlated with DPR, suggesting that MIP-1α and MCP-1 interact with different chemokine signal pathways in neuroinflammation. MCP-1, which is also known as CCL2, is considered to be a proinflammatory chemokine and believed to exert a detrimental effect by activating the CCR2 receptor in the inflammatory process (Jaerve & Muller, 2012). Nevertheless, CCR1 and CCR5 are receptors of MIP-1α, which differ from those of MCP-1 and may explain the differential effects between MIP-1α and MCP-1, although the complex relationship between chemokine and receptor is still vague. In the light of previous results demonstrating a correlation between MCP-1 levels and disease severity (Tateishi et al., 2010), our study suggests that MCP-1 has a predictive value for ALS progression and is possibly neurotoxic.
In addition, similar conclusions were reached in our previous report that MIP-1α is likely neuroprotective and possibly can monitor disease progression (Yang et al., 2016). Finally, MIP-1β was noted as exerting an unimpressive influence in neuroinflammation as observed by elevation only in the CSF.
IFNγ is also critical to the neuroinflammatory process and has always been considered as a proinflammatory cytokine to promote the proliferation of cytotoxic microglia (Evans, Couch, Sibson, & Turner, 2013 The bold values indicates that the results are statistically significant. T A B L E 6 Analysis of the survival of ALS patients using univariate and multivariate Cox proportional hazards models correlation and survival implied that IFNγ is a potential biomarker that can distinguish ALS patients from control subjects, monitor progression and predict the prognosis of patients with ALS. In brief, the immune response is neither beneficial nor deleterious by definition but rather a double-edged sword with different mediators under different conditions (Philips & Robberecht, 2011). Our study of cytokines above may provide insights into the neuroinflammatory aspect of ALS.
Apart from neuroinflammation, impaired neurotrophic support has also been long recognized as a hallmark of ALS. Trophic factors guide the generation and location of motor neurons; furthermore, they are involved in axon guidance and synapse formation (Tovary-Romo, Ramirez-Jarquin, Lazo-Gomez, & Tapia, 2014). We choose common neurotrophic factors, including bFGF and VEGF, and surveyed their differences between ALS patients and control subjects.
VEGF is associated with angiogenesis and has recently been shown to exert neurotrophic effects on neurons (Keifer, O'Connor, & Boulis, 2014); a similar trophic effect was also found with bFGF (Chen, Cai, Shen, Cai, & Lei, 2014). We measured increased bFGF and VEGF levels in both the CSF and serum and noted positive correlations with disease duration but negative associations with DPR. Additionally, higher levels of bFGF and VEGF indicated better prognosis as shown in our multivariate survival analysis. Our research confirms previous results, elucidating that increased bFGF and VEGF levels is potentially a reflexive neuroprotective mechanism in ALS (Mitchell et al., 2009).
Currently, a diagnosis of ALS is primarily based on medical history, clinical examination, electrophysiological results, and the exclusion of similar disorders. Therefore, finding reliable biomarkers that could provide accurate information regarding the onset and progression of ALS in clinical practice and trials is urgently necessary. The past year has seen progress in a range of areas, multiple candidate biomarkers have emerged, specifically from neurochemical analysis of biofluids (Chen et al., 2016;Xu, Henderson, David, & McCombe, 2016), neuroimaging biomarkers (Cardenas-Blanco et al., 2016); and neurophysiological techniques (Neuwirth et al., 2016). This study reported comprehensive results focusing on a panel of biomarkers both in the CSF and serum, which provided ample screening of potential predictors used a novel multivariate model. However, several limitations of this study must be noted. First, there was a small sample number chosen in the Cox proportional hazard model, so we describe the results as preliminary to set the stage for large-scale prognostic biomarkers analyses.
Moreover, this is a retrospective design that used samples collected at a single time point rather than measures the factor levels during disease course. Therefore, a longitudinal study that obtains serial samples from ALS patients to observe the dynamic alterations of these factors is necessary.

| CONCLUSIONS
In conclusion, we identified a panel of promising biomarker candidates of ALS that may be useful in clinical practice and therapeutic trials and that offer diagnostic, prognostic or monitoring potential to help elucidate the pathogenic mechanism of ALS. Moreover, CSF is more appropriate for biomarker discovery than serum because of its proximity to the affected regions exhibiting ALS-induced motor neuron death.
More large-scale studies with detailed longitudinal follow-up will be necessary to further evaluate the value of these biomarkers.

ACKNOWLEDGMENTS
The study supported by the fund from the Tianjin Health Bureau (2015KZ022 and 2015KZ035). We sincerely thank the doctors of the Department of Neurology who provided patient samples and the patients who participated in this study.

DISCLOSURES
The authors declare no financial or other conflict of interests.