|Home | About | Journals | Submit | Contact Us | Français|
Therapies designed to decrease SOD1 are currently in clinical trial for patients with superoxide dismutase (SOD1)-linked Familial Amyotrophic Lateral Sclerosis (ALS),
To determine whether SOD1 protein in cerebral spinal fluid (CSF) may be a pharmacodynamic marker and whether SOD1 protein in CSF is a disease marker for ALS.
Antisense oligonucleotides targeting human SOD1 (hSOD1) were administered to SOD1G93A rats. hSOD1 protein levels were measured in rat brain and CSF. In human CSF, the following proteins were measured: SOD1, tau, p-tau, VILIP-1, and YKL-40. was measured in human CSF.
SOD1G93A ALS model rats. ALS subject CSF (N=93), healthy controls (N=880 and neurological disease controls (NDC, N=89), including subjects with Dementia of the Alzheimer’s Type (DAT) (55), multiple sclerosis (19), and peripheral neuropathy (15).
Antisense oligonucleotide-treated SOD1G93A rats had decreased hSOD1 mRNA (69%+/−4%) and protein levels (48%+/ −14%) in brain. Importantly, rat CSF showed a similar 42+/−14% decrease in hSOD1. In human CSF, SOD1 varied 7.1+/−5.7 % on repeat measurements separated by months. SOD1 CSF levels were higher in ALS (172+/−8ng/ml, p<0.05) and NDC (172+/−6 ng/ml, p<0.05) compared with healthy controls (134+/−4ng/ml). Elevated CSF SOD1 did not correlate with disease characteristics in ALS or DAT subjects, but did correlate with tau, p-tau, VILIP-1 and YKL-40 in DAT subjects and controls.
CSF SOD1 may be an excellent pharmacodynamic marker for SOD1-lowering therapies since antisense oligonucleotide therapy lowers protein levels in both rat brain and rat CSF and since SOD1 CSF in humans is stable upon repeat measurements.
Amyotrophic Lateral Sclerosis (ALS) is an adult onset, neurodegenerative disease characterized by selective death of the upper and lower motor neurons of the brain and spinal cord. Symptoms include muscle atrophy, spasticity, paralysis and eventual death from respiratory failure within 3–5 years of diagnosis. There are no adequate therapies. While ALS mostly affects patients without family histories of the disease, 5–10% of ALS is familial (FALS). Nearly 20% of FALS is caused by Cu/Zn superoxide dismutase (SOD1) gene mutations1. SOD1 is a ubiquitously expressed, cytosolic enzyme involved in removal of superoxide. Although the mechanism is unclear, mutant SOD1 gains a toxic function independent of its normal enzymatic activity2, 3.
The fact that mutant SOD1 causes disease by a toxic gain of function 2–4 suggests that lowering levels of mutant SOD1 could benefit patients with SOD1-linked ALS. Antibody mediated lowering of SOD1 5, siRNA to SOD1 delivered by virus6–8, and antisense oligonucleotides to SOD19 have thus far demonstrated that lowering SOD1 in transgenic SOD1 mouse and rat models delays SOD1 mediated disease10. Smith, Miller and colleagues demonstrated the feasibility of the antisense oligonucleotide approach in animal models by administering antisense oligonucleotides to rats expressing a human SOD1-mutant transgene (SOD1G93A)9. Antisense oligonucleotides are short DNA-like chemicals that bind mRNA in a sequence-specific manner, triggering intranuclear mRNA degradation11. Because antisense oligonucleotides do not cross the blood-brain barrier, they must be directly infused into the cerebrospinal fluid (CSF), where they disperse throughout the central nervous system (CNS), reaching neuronal and non-neuronal cells9.
Antisense oligonucleotides that target SOD1 have recently completed a Phase I Clinical Trial (www.clinicaltrials.gov NCT01041222). A critical part of understanding the effects of antisense oligonucleotide therapy is determining whether the targeted protein has indeed been lowered. We report here our efforts to determine whether SOD1 in the CSF can serve as a pharmacodynamic marker for the efficacy of antisense oligonucleotide therapy in the central nervous system. This strategy is appealing for two reasons. First, while it is neither safe nor practical to biopsy brain or spinal cord, drawing CSF via lumbar puncture is a routine clinical practice. Second, SOD1 is highly abundant in the CSF, making it an easy target to follow. To determine the viability of CSF SOD1 as a pharmacodynamic marker, we tested whether antisense oligonucleotides that decrease SOD1 in rat brain also decrease SOD1 in rat CSF. We then measured CSF SOD1 levels in human subjects over time to ascertain whether, in future trials, we can attribute changes in CSF SOD1 levels to antisense oligonucleotide-therapy, rather than to the innate variability of SOD1 in the CSF.
An overlapping interest in SOD1 CSF levels in ALS patients stems from the growing number of reports implicating SOD1in the pathogenesis of sporadic ALS. Gruzman and colleagues found an SOD1 reactive protein (after chemical crosslinking) in ALS subjects but not in controls12. Antibodies that specifically recognize misfolded SOD1 revealed misfolded SOD1 in vulnerable spinal cord neurons of ALS patients, but not controls13. Most interestingly, lowering SOD1 levels in astrocytes derived from sporadic ALS subjects reversed the toxicity of these same astrocytes when co-cultured with motor neurons, again implying that SOD1 may be part of sporadic ALS 14. Given these findings, we also examined SOD1 protein CSF levels as a potential biomarker for sporadic ALS.
Antisense oligonucleotide #333611, which targets human SOD1-mRNA, was prepared by Isis Pharmaceuticals as described9. Either antisense oligonucleotide (100 µg/day) or saline was administered to the right lateral ventricle of 60-day SOD1G93A rats through an Alzet osmotic pump delivery system (Durect Corporation, Model 2004) as described9. Obtained from Taconic Farms, these rats are considered “low copy” because they survive more than 14 months, much longer than the originally described line, which survives 4 months15. Rats were treated for 30 days (from age 60 to age 90 days) and then pumps were removed. Two weeks after pump removal (104 days old), CSF and brain tissues were harvested, flash-frozen, and stored at −80°C.
ELISA was conducted to confirm dosage of antisense oligonucleotide #333611 in rat brain. Total RNA was extracted (Qiagen 74106) from rat tissues, and human SOD1 and rat SOD1-mRNA expression were measured relative to rat cyclophilin through RT-qPCR on the Applied Biosystems 7500 Fast Real-Time PCR System. Primers and probes were ordered from Integrated DNA Technologies. Primer and probe sequences for human SOD1 were as follows: forward, 5’-TGCATCATTGGCCGCA-3’; reverse, 5’-TTTCTTCATTTCCACCTTTGCC-3’; probe, 5'-/56-FAM/ACTGGTGGTCCATGAAAAAGCAGATGACTT/36-TAMTph/-3'. Primer and probe sequences for rat SOD1 were as follows: forward, 5’-CGGATGAAGAGAGGCATGTTG-3’; reverse, 5’TTGGCCACACCGTCCTTT-3’; probe, 5’-/56-FAM/AGACCTGGGCAATGTGGCTGCTG/36-TAMTph/-3'. Primer and probe sequences for rat cyclophilin were as follows: forward, 5’-CCCACCGTGTTCTTCGACA-3’; reverse, 5’-AAACAGCTCGAAGCAGACGC-3’; probe 5’-/56-FAM/CACGGCTGATGGCGAGCCC-/36-TAMTph/-3’. Right temporal-parietal brain sections from rats underwent dounce homogenization in lysis buffer. Total protein measurements were quantified by BCA assay (Pierce 23227). Human SOD1 protein was quantified through Cu/Zn SOD1 ELISA kit (eBioscience BMS222MST). Hemoglobin was detected through ELISA (Bethyl Laboratory, Inc. E80-135).
Healthy Controls (N=88), Neurological Disease Controls (AD, N = 55, multiple sclerosis, N=19 peripheral neuropathy, N=15) and ALS subject CSF samples (N=93) were donated by the Knight Alzheimer’s Disease Research Center (ADRC) of Washington University in St. Louis and the Northeast ALS Consortium (NEALS). The NEALS Consortium and ADRC, respectively, provided the ALS and AD biomarker and clinical data sets. In 14 patients from University of Pittsburgh, CSF was drawn at repeat time points ranging from 2 to 49 months after the initial lumbar puncture. These samples were drawn by the same physician at the same time of day (afternoon). IRB approved consent was obtained prior to CSF collection at MGH, Washington University and University of Pittsburgh. All CSF was immediately placed on ice, centrifuged at 500G at 4 C, then immediately frozen at −80 C. All CSF was clear and colorless.
Data were analyzed using a two-tailed student’s t-test. P values less than 0.05 were considered significant. The strength of correlation was measured by the Pearson’s coefficient of correlation.
SOD1G93A low-copy, asymptomatic rats were dosed intraventricularly with saline or 100 µg/day of antisense oligonucleotides designed to target and degrade human SOD1 mRNA. Consistent with our prior results using this same oligo9, SOD1 mRNA was decreased by 69+/− 4% (Fig 1A) and SOD1 protein by 48+/−14% (Fig. 1B) in the right frontotemporal cortex. To test whether this SOD1 protein decrease in the brain would be reflected in the CSF, we measured SOD1 protein in the CSF. We found that SOD1 protein in CSF was decreased by 42+/−14% (P = 0.003) (Fig. 1C). Furthermore, brain SOD1 protein correlated well with CSF SOD1 protein (R=0.54, p = 0.02) (Fig. 1D). These results from an animal model strongly suggest that measurements of human CSF may be a viable surrogate for knockdown in human brain and spinal cord.
To use SOD1 as a pharmacodynamic marker, another important consideration is the variability of CSF SOD1 in the same individual over time. In 14 human subjects who had repeat CSF draws at various time points over 7 to 48 months (average time between CSF draws = 8 months, average total time subject studied = 23 months), we measured CSF SOD1 protein levels (Fig. 2). While there was substantial variability among individuals’ absolute levels of CSF SOD1, there was surprisingly little variability within individuals over time, average 7.1+/−5.7%. The maximum variation between any two points for a given subject was 21%. Among the 55 total time points, there were four instances of variability greater than 15%. Given that within the same individual, SOD1 varies little over time, CSF SOD1 measurements are likely to accurately reflect the effects of an SOD1-lowering therapy rather than the natural variability of SOD1 metabolism or of the measurement itself.
To better evaluate SOD1 as a potential biomarker for sporadic ALS, we measured CSF SOD1 and total protein levels in 93 ALS subjects, 88 healthy controls (HC), and 89 neurological disease controls (NDC). Patient characteristics are shown in Table 1. Although all CSF appeared 100% clear of any blood contamination by visual inspection, we wanted to further exclude microscopic blood contamination of CSF since red blood cells are a well-known source of SOD116. To do so, we measured CSF hemoglobin values. The maximum hemoglobin in any patient was 9000 ng/ml in CSF. We next measured SOD1 in human blood. We found 1ng of SOD1 per 100,000 ng of hemoglobin. Therefore, even in our most “contaminated” sample, the contribution to the SOD1 measurement from hemoglobin contamination was approximately 0.1%. Consistent with this, hemoglobin values did not correlate with CSF SOD1 levels (R=−0.016 and p=0.916 for HC, R=−0.159 and p=0.241 for NDC and R=−0.059 and p=0.717 for ALS).
CSF SOD1 levels were significantly elevated in both NDC and ALS subjects (Fig. 3). Total protein was significantly elevated in ALS subjects (Fig. 3), as has been reported previously17, 18. The ratio of SOD1 CSF protein to total protein was significantly elevated in both NDC and ALS subjects (Fig. 3C). Thus, while the elevation of SOD1 is particularly interesting for ALS since SOD1 mutations cause ALS, the similar increase in neurologic disease controls suggests that the increase is either non-specific for ALS or specific for the particular set of diseases studied.
Correlations with subject characteristics have the potential to explain some of these changes. As would be expected from previous studies19, CSF total protein levels correlated moderately (R=0.3, p = < 0.05) with increased age in healthy controls (Table 2). Total protein levels did not correlate with age for NDC or ALS subjects. CSF SOD1 levels did not correlate with age in either controls or ALS subjects, but did correlate modestly with age for NDC (R = 0.41, p < 0.5) (Table 2). Previous work in 11 ALS subjects and 19 controls, has suggested a difference between SOD1 CSF values in males and females 20. In our samples, CSF protein values for males were 142+/−47 ng/ml, 173+/−67 ng/ml, and 167+/−63 ng/ml for HC, NDC, and ALS subjects respectively. For females values were 127+/−39 ng/ml, 174+/−85 ng/ml, and 189+/−78 ng/ml for HC, NDC, and ALS, respectively. None of the differences between males and females were significant (p=0.1, 0.95, and 0.16 for HC, NDC, and ALS respectively).
For ALS subjects, we hypothesized that CSF SOD1 would be elevated in subjects with more severe disease as determined by ALSFRS at the time of the CSF draw or more rapidly progressing ALS subjects as determined by ALSFRS change per month. As shown in Fig. 4, neither of these hypotheses is correct. Total CSF protein also did not correlate with ALSFRS (data not shown).
Fifty-five of the 89 neurological disease controls were participants with very mild (CDR 0.5) or mild (CDR 1) Dementia of Alzheimer’s Type (DAT). Similar to subjects with ALS, SOD1 CSF levels did not correlate with disease severity (clinical dementia rating scale (CDR) = 0.5 or 1). Since biomarkers for AD have been extensively studied, we analyzed previously described markers of disease in participants with mild dementia (CDR 0.5 and 1) and age-matched, neurologically normal controls (Table 3). SOD1 CSF levels were either weakly correlated or not correlated with measures of amyloid plaque load including Aβ-42 CSF levels and mean cortical binding potential of PIB, an imaging compound that specifically assesses the fibrillar amyloid plaque burden21, 22. Markers of neuronal damage (tau, p-tau, VILIP-1)23, 24 and inflammation (YKL-40)25 were more strongly correlated with CSF SOD1 levels (see Table 3), though more so in the controls and CDR 0.5 subjects than in CDR 1, with the exception of YKL-40, which showed a weak correlation in CDR 0 and a stronger correlation in CDR 1.0. These data suggest that increased SOD1 CSF levels in neurodegenerative disease are correlated with increased degree of neuronal damage and inflammation.
We demonstrated, for the first time, a strong correlation between knockdown of a target in the brain and knockdown of the same target protein in CSF. For designing therapeutic trials for SOD1 and other neurodegenerative disease proteins, demonstrating knockdown in the CSF is a key piece of information, since now, based on these animal data, we have confidence that SOD1 CSF will be an appropriate fluid for determining a pharmacodynamic response. The ability to sample CSF, which is a relatively routine procedure, will greatly enhance our ability to determine whether the antisense oligonucleotides indeed decrease SOD1 in the CNS.
Equally important for consideration of SOD1 as a pharmacodynamic marker is the finding that there is little variation in CSF SOD1 levels upon repeat measures, with variation on average 7% between time points. This finding is consistent with the lack of correlation between CSF SOD1 levels and ALS disease severity and progression, suggesting that once patients become sick, SOD1 remains stable throughout the disease course. Based on our finding that CSF SOD1 is decreased by 42+/−14% in antisense oligonucleotide-treated rats and based on the variation of 7.1+/−5.7% from point to point in human subjects, we estimate that we have a >95% chance of seeing a 40% or more reduction of CSF SOD1 levels in as little as six mutant-SOD1, familial ALS subjects. We recognize that we may or may not obtain the same robust knockdown in subjects in a Phase II trial as we see in rodents and therefore will need a larger sample size, for example 20–30 subjects. Since we anticipate successful recruitment of the subjects needed for our Phase I (www.clinicaltrials.gov NCT01041222) in this same population, we consider 20–30 subjects an achievable goal for such a Phase II trial.
One remaining issue for planning a pharmacodynamic study of modulating CSF SOD1 levels is the half-life of the protein. Using hydrogen-deuterium exchange and mass spectrometry, a recent paper suggested that the tissue half-life of a SOD1-YFP fusion protein in a mouse is approximately 22 days26. Whether the SOD1 half-life in CSF is the same as in tissue is an important, ongoing question that may be addressed by using stable-isotope linked kinetics (SILK), a method developed to test protein synthesis and half-life in CSF in research subjects27. In addition, it remains unclear whether the process of YFP fusion could affect protein half-life.
Given the recent data suggesting that SOD1 may contribute to sporadic ALS, we explored the possibility of whether SOD1 is specifically increased in sporadic ALS and not other diseases. Our data do not support this hypothesis. We found increased levels of SOD1 in the CSF of NDC as well as ALS subjects; thus increased SOD1 is not a specific biomarker for ALS. We also considered whether CSF SOD1 levels might vary according to ALS disease progression or disease severity, but found no correlation. In participants with DAT, where biomarkers of disease have been extensively evaluated, we also compared CSF SOD1 values to amyloid beta, tau, p-tau, VILIP, and YKL-40 measurements. As in ALS, SOD1 CSF levels did not correlate well with severity of disease. In addition, SOD1 levels did not correlate strongly with amyloid beta levels, suggesting that SOD1 is not a component of the abnormal amyloid beta cascade of AD. We did, however, find correlations with tau, p-tau, and VILIP-1, which are markers of neuronal damage. We also found strong correlation with YKL-40, likely a marker of inflammation. We conclude that increased SOD1 in the CSF of AD and ALS subjects is likely associated with increased neuronal damage and inflammation. Since our neurological disease controls were dominated by AD, it remains possible, though we believe unlikely, that increased SOD1 in the CSF is specific to AD and ALS and not found in other neurodegenerative diseases.
In our study, we found increased total protein in the CSF of ALS patients. This is similar to previous reports from Guilloff 17(N= 38 cases) and Norris28, (N = 385 cases). On the other hand, Younger et al. in a series of 120 patients did not find increased total CSF protein18. The reason for increased CSF protein in some series remains unclear. Neither our study, nor other reported studies found correlations of CSF protein with other disease measures.
For ALS, specific SOD1 post-translational modifications or aggregation state may be a more important biomarker than total SOD1 levels. Although total SOD1 levels are not likely to be a biomarker for sporadic ALS, we remain enthusiastic about ongoing studies to examine whether different properties of SOD1, e.g. misfolded SOD1 in the CSF, may be a biomarker for a subset of sporadic ALS, as suggested by studies of post-mortem material from ALS subjects 12–14. However, a recent paper from Zetterstrom and colleagues using antibodies that recognize misfolded SOD1 showed no difference in misfolded SOD1 in the CSF in ALS vs. controls, even when comparing ALS patients with known SOD1 mutations.29
The data presented here demonstrate that CSF SOD1 is stable and that knockdown of SOD1 in the brain results in reduced SOD1 in the CSF. We propose that SOD1 in the CSF is likely to be a useful pharmacodynamic marker for therapies designed to lower SOD1 in the brain and spinal cord.
CSF was harvested from SOD1G93A rats at ages 90, 105,120 and 150 days with n= 6, 12, 8, and 6 respectively. CSF hSOD1 protein levels were then determined using hSOD1 specific ELISA. Average +/− SE. (p=0.85)
We thank John Morris [J.C.M.] for critical reading of the manuscript and for providing clinical samples. This work was supported by an ALS Association pilot grant [T.M.M.] and CTSA grant UL1 RR024992 [T.M.M.]. CSF samples were from NEALS and the Washington University Knight ADRC (P50-AG05681 [J.C.M.], P01-AG03991 [J.C.M.], P01-AG26276 [J.C.M.], P30-NS057105 [D.M.H.]), Charles and Joanne Knight Alzheimer Research Initiative [J.C.M.] and the University of Pittsburgh (RO1 NS061867-01 [R.B.], RC1 NS068179-01 [R.B.])