Conceived and designed the experiments: NS. Performed the experiments: AS. Analyzed the data: AJB AS. Wrote the paper: NS ′AJB AS.
The authors have declared that no competing interests exist.
Sporadic CreutzfeldtJakobdisease (sCJD) is a fatal neurodegenerative condition that escapes detection until autopsy. Recently, brain iron dyshomeostasis accompanied by increased transferrin (Tf) was reported in sCJD cases. The consequence of this abnormality on cerebrospinalfluid (CSF) levels of Tf is uncertain. We evaluated the accuracy of CSF Tf, a ‘new’ biomarker, as a premortem diagnostic test for sCJD when used alone or in combination with the ‘current’ biomarker totaltau (Ttau). Levels of totalTf (TTf), isoforms of Tf (Tf1 and Tfβ2), and iron saturation of Tf were quantified in CSF collected 0.3–36 months before death (duration) from 99 autopsy confirmed sCJD (CJD+) and 75 confirmed cases of dementia of nonCJD origin (CJD). Diagnostic accuracy was estimated by nonparametric tests, logistic regression, and receiver operating characteristic (ROC) analysis. Area under the ROC curve (AUC), sensitivity, specificity, positive and negative predictive values (PV), and likelihood ratios (LR) of each biomarker and biomarker combination were calculated. We report that relative to CJD, CJD+ cases had lower median CSF TTf (125,7093 vs. 217,7893) and higher Ttau (11530 vs. 1266) values. AUC was 0.90 (95% confidence interval (CI), 0.85–0.94) for TTf, and 0.93 (95% CI, 0.89–0.97) for TTf combined with Ttau. With cutoffs defined to achieve a sensitivity of ∼85%, TTf identified CJD+ cases with a specificity of 71.6% (95% CI, 59.1–81.7), positive LR of 3.0 (95% CI, 2.1–4.5), negative LR of 0.2 (95% CI, 0.1–0.3), and accuracy of 80.1%. The effect of patient age and duration was insignificant. TTf combined with Ttau identified CJD+ with improved specificity of 87.5% (95%CI, 76.3–94.1), positive LR of 6.8 (95% CI, 3.5–13.1), negative LR of 0.2 (95% CI, 0.1–0.3), positivePV of 91.0%, negativePV of 80.0%, and accuracy of 86.2%. Thus, CSF TTf, a new biomarker, when combined with the current biomarker Ttau, is a reliable premortem diagnostic test for sCJD.
Sporadic CreutzfeldtJakob disease is a uniformly fatal prion disorder of humans that results from refolding of the normal prion protein (PrP^{C}) to a βsheet rich pathogenic conformation called PrPscrapie (PrP^{Sc})
Consequently, significant effort has gone into the identification of sensitive and specific biomarkers for sCJD
Major factors contributing to the inconsistent performance of 1433 and Ttau combination include: 1) variable false positives since 1433 and Ttau are elevated in several other dementias besides sCJD
One such potential biomarker is Tf that is increased in sCJD and prion disease affected animal brains
Human tissue and CSF samples were obtained from the National Prion Disease Pathology Surveillance Center (NPDPSC) at Case Western Reserve University, Ohio. All samples are from deceased subjects, and personal information is limited to age, sex, symptoms, neuropathology, and classification of prion disease. The use of these samples has been granted waiver of informed consent by the Case Western Reserve University Institutional Review Board since the protocol meets criteria for exemption under Federal regulations 45 CFR 46.101 (b).
To avoid sample bias, premortem CSF from autopsyconfirmed cases of sCJD (CJD+) and dementia of nonCJD origin (CJD) received between 2006 and 2008 were included in this study. The samples consisted of 99 CJD+ and 75 CJD cases, amounting to a total of 174 cases. Specific diagnosis of CJD cases included Alzheimer's Disease (AD) (17), encephalitis (3), meningoencephalitis (2), cortical dysplasia (1), cortical angiopathy (1), angiitis with microinfarcts (1), infarct (1), cerebral vasculopathy (1), Lewy body dementia (1), Lewy body variant of Alzheimer's disease (1), anoxic and Wernicke's encephalopathies (1), meningeal carcinomatosis (1), Alexander disease (1), leukoencephalopathy (1), hippocampal sclerosis (1), frontotemporal dementia (1), coccidioidomycosis meningoencephalitis (1), CNS lymphoma (1), chronic meningoencephalitis (1), epilepsy or encephalopathy (1), glioma (1), perivenous encephalomyelitis (1), CNS lymphoma (1), military metastases (1), leptomeningeal lymphomatosis/leukemia (1), mitochondrial encephalopathy (1), lymphoma (1), and nonCJD dementia of uncertain diagnosis (29). The age at autopsy ranged from 37–85 years for CJD+, and 47–84 years for CJD cases. All CSF samples were stored at −80°C until use. Human brain tissue from the frontal cortex of CJD+ and agematched cases of dementia (CJD−) was obtained from the NPDPSC and stored at −80°C (brain tissue and CSF samples are from different cases). Tissue homogenates (10%) were prepared in lysis buffer and analyzed by Western blotting following standard protocols.
Equal volume of CSF from CJD+ and CJD cases was fractionated by SDSPAGE and proteins transferred to PVDF membranes were probed with antitransferrin antibody (Genetex Inc., Cat # GTX 21223) followed by HRPconjugated secondary antibody and visualization of reactive bands with ECL (Amersham). Several procedural and statistical precautions were taken to reduce error in comparing multiple samples. Procedurally, a similar protocol was followed for all Western blots, including exposure times. In addition, strips of PVDF membrane representing specific molecular weights were exposed to Xray film simultaneously in large cassettes to obtain similar exposure. Quantification of immunoreactive bands was performed with UNSCANIT software (version 6.1, Silk Scientific Inc., Utah, USA) using three exposures from a single membrane showing exponential increase in intensity. Statistically, results from different Western blots were analyzed simultaneously in logistic regression by treating these as clustered observations. The statistical software Stata allows estimation of regression coefficients after controlling for clustering to produce unbiased standard errors. Two Stata procedures were used to do this: the ‘vce cluster’ option and svy option. Each logistic regression model was first tested without adjustment for clustering and then tested with each of the 2 clustering options. Both had the effect, in general, of increasing standard errors and pvalues for the coefficient estimates as expected. However, the significance of the coefficients of the biomarkers did not change; all their coefficients remained significant at the p<.05 level. TTf (ELISA) was estimated with Human Transferrin ELISA kit (Alpha diagnostic international Inc., Cat # 1210) following the manufacturer's instructions. Ttau was determined by Human Tau (total) ELISA Kit (Invitrogen, Camarillo, CA; KBH0042) as directed by the manufacturer. Iron saturation of CSF Tf was determined by radiolabeling with ^{59}Fe followed by fractionation on native gradient gels and autoradiography.
The following new biomarkers were evaluated for their diagnostic accuracy: total transferrin by Western blot (TTf (WB)) and ELISA (TTf (ELISA)) techniques, Tf1, Tfβ2, and iron saturation of Tf (^{59}FeTf). Of these, TTf, Tf1, and Tfβ2 offered greater promise, and were analyzed further either alone or in combination with Ttau, a current biomarker that is quantifiable and is superior to other CSF biomarkers used for the diagnosis of sCJD
A combination of univariate and multivariable statistical methods were used to evaluate the relative diagnostic accuracy of individual biomarkers. Since the biomarkers were expected to evidence skewed distributions, nonparametric statistical procedures were used as these are subject to fewer distributional assumptions and provide less biased estimators when data are nonnormal. For relative analysis of individual biomarkers, the nonparametric MannWhitney
Using logistic regression results, an analytic expression for the risk of CJD was derived for each individual biomarker. Area under the receiver operating characteristic (ROC) curve and the Aikake Information Criterion (AIC) were obtained. A ROC curve graphically shows the tradeoffs between sensitivity and specificity for different cutoffs used to discriminate between positive and negative cases (i.e., CJD+ and CJD cases). The area under the ROC curve (AUC) can be understood as an estimate of the probability that the biomarker being tested correctly ranks a CJD+ case higher than a CJD case and, therefore, indexes the discriminating power of the biomarker. The AIC is used to compare different logistic regression models for different biomarkers (individually or in combination). A model with a relatively lower AIC is considered superior due to its better fit to data and its parsimony. For each model, estimates of specificity and positive and negative likelihood ratios (LR) were obtained given a baseline sensitivity of 85%, which has precedence in the literature as a reasonable cutoff level for biomarker comparison of Alzheimer's disease (AD)
To identify an optimal combination of biomarkers, all two way combinations of new and current biomarkers were also entered into a logistic regression model with controls for age, duration, and clustered data (as above). Combining new and current biomarker Ttau was intended to take advantage of the differential pathogenic mechanisms represented by these biomarkers. We focused on twoway combinations for reasons of practicality. However, a specific combination involving TTf and 1433 could not be evaluated because the readouts for 1433 are not precisely quantifiable, and this biomarker has demonstrated relatively low specificity in previous studies
All models were compared using three criteria: AUC, specificity, and AIC
Evaluation of CSF Tf by Western blot (WB) reveals two distinct bands representing Tf1 and Tfβ2, the latter representing deglycosylated Tf specific to the brain and CSF (
(
Quantitative comparison shows a decrease in CSF TTf (WB) by 49% (z = 8.73, p<.001) and an increase in brain TTf (WB) by 39% (z = 2.76, p<.008) in CJD+ relative to CJD cases (
(
Comparison of individual subunits of Tf shows a decrease in CSF Tf1 by 35% (z = 7.59, p<.001) and Tfβ2 by 61% (z = 6.96, p<.001) in CJD+ relative to CJD samples (
To compare the iron content of CSF Tf between CJD and CJD+ cases, first the iron saturation of CSF Tf from a relatively normal CJD case was determined by competing radiolabeled ^{59}FeCl_{3} with decreasing concentrations of unlabeled FeCl_{3} and fractionating ^{59}FeTf on a native gel for quantification (
Comparison of different biomarkers using MannWhitney
(
Comparison test  New Biomarkers  Current Biomarker  
TTf (WB)  Tf1  Tfβ2  TTf (ELISA) (µg/ml)  Ttau (pg/ml)  


CJD+ median (n)  1257093 (99)  1017031 (87)  236243 (87)  26.46 (99)  11530.87 (95) 
CJD median (n)  2177893 (67)  1555881 (67)  606359 (67)  38.90 (64)  1266.97 (64) 
M–W U Test  

8.73  7.59  6.96  6.53  −6.37 

<.001  <.001  <.001  <.001  <.001 


CJD+ median (n)  1257093 (99)  1017031 (87)  236243 (87)  26.46 (99)  11530.87 (95) 
AD median (n)  2009259 (17)  1413527 (17)  577539 (17)  36.89 (15)  876.40 (15) 
M–W U Test  

4.92  4.44  3.96  3.02  −5.17 

<.001  <.001  <.001  0.003  <.001 
After controlling for patient age and duration, Ttau showed significant correlation with duration in CJD+ (ρ = −0.35, p<.001), but not in CJD cases (ρ = −0.14, p = 0.28). None of the new biomarkers showed a significant correlation with duration in the CJD+ or CJD group. (Correlation of TTf (WB), Tf1, and Tfβ2 in CJD+ group ρ = −0.08, p = 0.41; ρ = −0.02, p = 0.89; ρ = 0.11, p = 0.29, respectively, and in CJD group ρ = 0.06, p = 0.62; ρ = 0.01, p = 0.96; ρ = 0.06, p = 0.64, respectively). These results indicate that Ttau changes as sCJD progresses, while the new biomarkers remain fairly stable in CJD+ and CJD cases. None of the biomarkers showed any correlation with age.
For comparative analysis of new and the current biomarker Ttau, logistic regression analysis was performed. Only TTf (WB), not TTf (ELISA) results are considered for reasons mentioned above. When individual biomarkers are tested for their diagnostic accuracy, the new biomarkers TTf (WB), Tf1, and Tfβ2 are superior to the current biomarker Ttau (
Statistic  New Biomarkers (n)  Current biomarker (n)  New + Current  
TTf (WB) (99)  Tf1 (87)  Tfβ2 (87)  TTf ELISA (99)  Ttau (95)  TTf (WB)+Ttau  
Area under ROC (95%CI)  0.90 (0.85–0.94)  0.86 (0.80–0.92)  0.82 (0.76–0.89)  0.80 (0.73–0.88)  0.78 (0.71–0.85)  0.93 (0.89–0.97) 
Sensitivity (95%CI)  85.9 (77.1–91.8)  85.1 (75.4–91.5)  85.1 (75.4–91.5)  85.9 (77.1–91.8)  85.3 (76.2–91.4)  85.3 (76.2–91.4) 
Specificity (95%CI)  71.6 (59.1–81.7)  65.7 (53.0–76.6)  64.2 (51.5–75.3)  64.1 (51.0–75.4)  48.4 (35.9–61.2)  87.5 (76.3–94.1) 
Positive LR (95%CI)  3.0 (2.1–4.5)  2.5 (1.8–3.5)  2.4 (1.7–3.3)  2.4 (1.7–3.3)  1.7 (1.3–2.1)  6.8 (3.5–13.1) 
Negative LR (95%CI)  0.2 (0.1–0.3)  0.2 (0.1–0.4)  0.2 (0.1–0.4)  0.2 (0.1–0.4)  0.3 (0.2–0.5)  0.2 (0.1–0.3) 
PPV (%) (95%CI)  81.7 (72.7–88.4)  76.3 (66.4–84.1)  75.5 (65.6–83.4)  78.7 (69.6–85.8)  71.1 (61.7–79.0)  91.0 (82.6–95.8) 
NPV (%) (95%CI)  77.4 (64.7–86.7)  77.2 (63.8–86.8)  76.8 (63.3–86.6)  74.5 (60.7–84.9)  68.9 (53.2–81.4)  80.0 (68.4–88.3) 
AIC  0.81  0.96  1.03  1.06  1.23  0.70 
Accuracy  80.1  76.6  76  77.3  70.4  86.2 
Among the new group of biomarkers, TTf (WB) has by far the highest diagnostic accuracy relative to other individual biomarkers. Thus, TTf (WB) yields an AUC of 0.90 that is significantly higher than corresponding values for other biomarkers with the exception of Tf1. The AIC associated with the logistic regression model using TTf (WB) is lowest across all other models. Finally, the specificity of TTf (WB) is significantly higher than that of TTau and equivalent to that of Tf1 and Tfβ2. However, TTf (ELISA) does not have the same superior characteristics: the model has a lower AUC (0.81), higher AIC (1.06), and poorer specificity of 64.1% for reasons mentioned above.
Past research shows that relative to any single biomarker, using two biomarkers in combination generally enhances the prediction of individuals likely to suffer from AD or sCJD
(
This report describes the accuracy of new biomarkers CSF TTf (WB) including its two isoforms Tf1 and Tfβ2 in identifying CJD+ from dementia of other causes. Using rigorous statistical modeling and analysis of new and currently used biomarker Ttau, we demonstrate that each of the new biomarkers is superior to Ttau, and the combination of TTf (WB) and Ttau is more accurate in identifying sCJD than the currently used combination of 1433 and Ttau. Below, we discuss distinctive features of the methodology used for comparing new and current biomarkers, possible reasons for high specificity of new biomarkers when used singly and significant improvement in diagnostic accuracy when combined with Ttau, and limitations of our study leading to directed avenues in the identification of accurate CSF biomarker(s) for sCJD.
A rigorous statistical approach was used to examine the significance (nonparametric M–W test), quantitative comparison using uniform criteria (logistic regression), and sensitivityspecificity interdependence (ROC) of different biomarkers. In contrast to specific cutoff values for individual biomarkers, a preset sensitivity of 85% was applied to all biomarkers to provide a common quantitative baseline for comparison
However, instead of the compensatory increase in CSF Tf in response to brain iron deficiency as observed in cases of Restless Leg Syndrome
Regardless of the underlying cause, the difference in CSF Tf between CJD+ and CJD cases is noted much before endstage disease, providing a useful premortem diagnostic biomarker for sCJD. When combined with the surrogate biomarker Ttau, the diagnostic accuracy of TTf (WB) and Ttau improves significantly, and is superior to the reported accuracy of 1433 and TTau combination
Contrary to the norm, a specific cutoff value for TTf or Ttau was not identified to calculate the specificity and sensitivity of these biomarkers. For TTf we thought it premature to decide on such a value due to the limited sample size. For Ttau we did not use the conventional cutoff of 1200 or 1300 pg/ml for three reasons: 1) a consistent parameter was essential for comparing TTf and Ttau singly or in combination, 2) majority of CJD samples were collected in the last month before death when Ttau is likely to be released into the CSF from damaged neurons, and 3) inclusion of conditions such as AD and brain inflammation that are associated with increased levels of Ttau in the CSF
In addition to its superior sensitivity and specificity, CSF Tf by itself and in combination with Ttau offers several additional advantages as a biomarker for sCJD: 1) CSF TTf reflects prion disease associated brain iron imbalance and is therefore likely to be more specific
In conclusion, reduced levels of CSF Tf reinforce previous reports indicating the association of brain iron dyshomeostasis with prion disease pathology, and offer promise as a premortem diagnostic test for sCJD. Although alteration of CSF Tf in sCJD cases has been described previously
(TIF)
(TIF)
We thank Jagdip Singh (Case Western Reserve University) for help and guidance with statistical analysis and critical evaluation of the manuscript, Shreya Nayak for assistance, and Janis Blevins and Kay Edmonds from the NPDPSC for providing CSF samples and patient data.