The present study represents a comprehensive evaluation of cystatin C as a candidate biomarker in ALS, and is unique in its assessment of two distinct biofluids (CSF and blood plasma), two different measurements of the protein of interest (total concentration and percent of total protein), and longitudinally collected CSF samples. Prior studies using SELDI-TOF-MS found significantly lower cystatin C abundance in the CSF of ALS patients relative to healthy controls 
and mixed healthy/neurologic disease controls 
. These findings were validated by ELISA and immunoblot, respectively. In our current ELISA study for diagnostic utility using a much larger number of total subjects, we also found a significant reduction in cystatin C levels in the CSF of ALS patients relative to healthy controls, but the magnitude of this difference was less robust than in the previous reports. This discrepancy may have resulted from the use of different experimental techniques, as SELDI-TOF-MS recognizes discrete mass-to-charge forms of cystatin C, whereas ELISA may recognize multiple modified or cleaved forms of cystatin C depending upon the antibodies used for capture and detection. Regarding the comparison of these two techniques, we found a significant, positive correlation but a low correlation coefficient between our CSF ELISA data and SELDI-TOF-MS data for the same samples (Figure S1
). This finding suggests that these techniques are sensitive to different, but possibly overlapping, ranges of native cystatin C isoforms, and may provide differential, and perhaps complementary, utility in detecting cystatin C for biomarker assessment.
To be clinically useful as a diagnostic biomarker, cystatin C must also be able to differentiate between ALS patients and individuals with neurologic diseases that closely resemble ALS, or ALS “mimic diseases.” A recent study reported a significant reduction in CSF cystatin C levels in ALS patients relative to polyneuropathy patients 
. In our ELISA analysis, cystatin C was reduced in the CSF of ALS patients relative to all DC combined and, to a greater degree, relative to a mimic disease control group that included a variety of ALS mimics (), but neither difference reached statistical significance. Because these between-group differences were smaller than we expected based on previous mass spectrometry data 
, we conducted a new power analysis using our experimentally-derived group means and standard deviations. This analysis revealed that our study was adequately powered for comparing percent cystatin C between ALS and HC (a significant difference was found), and underpowered for comparing total cystatin C between ALS and HC (main group effects missed significance, but the pairwise comparison was significant), and for comparing ALS with DC for both measures of cystatin C (no significant differences were identified). The observed reductions in both total and percent cystatin C in ALS patients relative to DC may reflect actual differences in clinical cystatin C levels, but a total study enrollment of 1020 and 675 patients (for total and percent cystatin C, respectively) would be required to confirm statistical significance with 80% power and 95% confidence. Interestingly, the between-group differences and trend towards significance improved when comparing limb-onset ALS patients or ALS patients with disease course greater than 1 year from symptom onset to the ALS mimics (). Additionally, we found that the total cystatin C concentration measurement generated superior diagnostic accuracy, indicating that this may be the more efficacious measure of cystatin C. An assessment of the diagnostic parameters of CSF cystatin C concentration revealed that the sensitivity of cystatin C for differentiating ALS patients from disease controls is low for all cutoff values but it displays high levels of specificity and, therefore, cystatin C can only identify a small subset of ALS patients. Together, these findings indicate that CSF cystatin C levels may differ between ALS patients and relevant disease control populations but cystatin C, by itself, has limited diagnostic utility. However, this protein could potentially improve the sensitivity and/or specificity of a diagnostic biomarker panel. Due to the heterogeneous nature of the ALS patient population, it is likely that a multiple biomarker panel will be required, as opposed to any single protein biomarker, in order to differentiate ALS from related disorders with adequate diagnostic certainty 
We also assessed the diagnostic utility of plasma cystatin C levels. Plasma cystatin C has been extensively characterized as a peripheral biomarker for kidney function and as a prognostic indicator of the risk of morbidity and mortality relating to cardiovascular disease 
. However, blood-borne levels of cystatin C have not been evaluated as a biomarker candidate for neurologic disorders. We found that plasma cystatin C levels are equivalently elevated in both ALS patients and disease controls relative to healthy controls, indicating that elevated plasma cystatin C is a nonspecific finding associated with neurologic disease states. Therefore, plasma cystatin C levels, as evaluated by ELISA, do not to have diagnostic utility for ALS. Furthermore, the absence of a relationship between cystatin C levels in concurrently-drawn CSF and plasma samples from individual patients in this study (Figure S2
) suggests that this protein is independently regulated in each biofluid. Accordingly, plasma cystatin C levels are unlikely to be directly correlated with motor neuron degeneration in ALS, though elevated levels may correlate to peripheral metabolic or inflammatory abnormalities during ALS.
A recent study examined a single CSF draw per ALS patient, taken at varying times from symptom onset, to indirectly infer the average longitudinal change in cystatin C concentration in the group as a whole, and they reported that cystatin C levels do not change over time 
. We completed a similar analysis and also found no evidence for a patterned directional change in CSF cystatin C levels over time in ALS patients (). However, both heterogeneity in disease progression speed and individual variation in baseline cystatin C levels could mask significant trends in cystatin C change over the course of disease progression and, therefore, single-draw protein levels are unsuitable for a thorough assessment of longitudinal trends in cystatin C abundance.
We also examined longitudinal CSF data from multiple patients to more accurately assess the changes in cystatin C over time. We found that longitudinal cystatin C concentrations were relatively constant in ALS patients as a combined group. In contrast, the subgroup of patients with slow or absent clinical disease progression exhibited longitudinal increases in cystatin C concentration, and the subgroup with more typical, continuous clinical deterioration exhibited longitudinal decreases in total cystatin C. Interestingly, slow progressors often exhibited lower initial levels of CSF cystatin C than fast progressors (). Similar trends were also observed for percent cystatin C measurements, but statistical significance was not reached. These results indicate that CSF cystatin C levels in ALS patients change over time in a clinically-relevant manner and that increasing cystatin C concentration may be associated with slower disease progression. Conversely, rapid disease progression may be associated with a decrease in cystatin C concentration over time.
We also conducted an analysis to determine the relationship between longitudinal changes in CSF cystatin C levels and time-matched changes in three functional clinical measures of disease progression (ALSFRS-R, MMT, and FVC). However, no significant correlations were found (data not shown). This indicates that cystatin C levels may change independently of the clinical parameters used for monitoring disease progression. However, this finding does not eliminate the possibility that changes in CSF cystatin C levels correlate with more subtle biochemical changes associated with disease progression, as these may not be accurately reflected by overt functional measures of clinical disease status 
. Furthermore, the observed trend of increasing cystatin C levels in patients with slow rates of clinical deterioration may prove to be useful as an objective biomarker for monitoring drug effects in clinical trials.
We recently demonstrated a correlation between CSF cystatin C levels and patient survival by SELDI-TOF-MS 
. In this study, we further verified a direct correlation between CSF cystatin C concentration and patient survival time, supporting the potential utility of this protein for prognostic applications. Subsequent Kaplan-Meier survival analyses for patient groups with CSF cystatin C concentrations above and below qualitatively selected cut-off values confirmed significantly longer survival times for patients in the higher cystatin C groups. Additionally, the prognostic capacity of CSF cystatin C was higher for limb-onset patients () than for all patients combined (). This may have resulted from the confounding effects of combining patients with different sites of disease onset, as bulbar-onset ALS patients typically have shorter survival times than limb-onset patients 
. Unfortunately, there were inadequate numbers of bulbar-, trunk-, and/or dementia-onset patients to analyze these individual subgroups in this study, and further analyses are required to determine the prognostic capacity of cystatin C in these subgroups. Nonetheless, these results show that cystatin C is a candidate prognostic indicator of survival in ALS patients. Alternatively, cystatin C levels could contribute to the process of balancing prognostic variables among experimental groups as recommended to equalize drop-out rates and preserve the balancing effects of randomization in clinical trials 
. Further work is required to more fully characterize the relationship between CSF cystatin C concentration and ALS patient survival, and to determine optimal cut-off values and procedures to stratify patients for prognostic purposes.
The results of this comprehensive biomarker assessment also have implications for the potential mechanistic involvement of cystatin C in the pathogenesis of ALS. The function of cystatin C within the CNS has not been extensively studied, but it appears to have both neurotoxic and neuroprotective properties 
, though its effects specifically on motor neurons have not been reported. The majority of cystatin C in the CSF is produced by the choroid plexus 
, but it is unclear whether the apparent reductions in CSF levels in ALS patients are an independent etiological factor contributing to motor neuron degeneration, a downstream result of disease pathogenesis, or a compensatory response to ALS pathology. However, the association of higher cystatin C concentrations with longer patient survival and the association of increasing cystatin C levels with slower clinical progression both suggest that extracellular cystatin C may exhibit neuroprotective properties within the context of ALS. This would implicate any absolute or relative cystatin C deficiency in ALS as both a potential contributor to disease pathogenesis and a potential therapeutic target. Continuing work in our laboratory is focused on determining the effects of altered cystatin C concentration/activity on motor neurons in vitro
, in order to clarify its potential mechanistic role in ALS pathogenesis.
In summary, we have completed a comprehensive evaluation of cystatin C as a candidate ALS biomarker, including assessments of two complementary measures of cystatin C in two distinct biofluids as well as examinations of both longitudinal CSF samples and patient survival data. Our findings indicate that cystatin C levels, as determined by ELISA, are increased in the plasma and decreased in the CSF of ALS patients relative to healthy controls. CSF cystatin C measurements may possess a more limited diagnostic capacity for ALS than previously proposed, but may still have the potential to improve the diagnostic parameters of a biomarker panel. Additionally, longitudinal changes in CSF cystatin C levels may be useful as a biomarker of fast versus slow rates of disease progression. Our data also demonstrate that CSF cystatin C concentration has prognostic utility in estimating patient survival time. Further validation studies are necessary to confirm these findings and ultimately determine if cystatin C measurements can be used to enhance clinical disease management and clinical trial design. Finally, the association of high or increasing cystatin C levels with slower disease progression and increased survival time suggests a potential neuroprotective role for this protein in the pathobiology of ALS.