Filtered Cerebrospinal Fluid From Patients With Amyotrophic Lateral Sclerosis Displays an Altered Proteome and Affects Motor Phenotype in a Mouse Model

Introduction: Cerebrospinal fluid (CSF) has been implicated in amyotrophic lateral sclerosis (ALS) due to its ability to spread inflammatory proteins throughout the nervous system. We hypothesized that filtration of the CSF could remove pathogenic proteins and prevent them from altering motor phenotypes in a mouse model. Methods: We filtered the CSF from 11 ALS patients via 100 kilodaltons (kD) molecular weight cut-off filters. We used mass spectrometry-based discovery proteomics workflows to compare protein abundances before and after filtration. To test the effects of CSF filtration on motor function, we injected groups of mice with saline, filtered ALS-CSF, or unfiltered ALS-CSF (n=12 per group) and assessed motor function via pole descent and open field tests. Results: We identified proteins implicated in ALS pathogenesis and showed that these were removed in significant amounts in our workflow. Key filtered proteins included complement proteins, chitinases, serine protease inhibitors, and neuro-inflammatory proteins such as amyloid precursor protein, chromogranin A, and glial fibrillary acidic protein. Compared to the filtered ALS-CSF mice, unfiltered ALS-CSF mice took longer to descend a pole (10 days post-injection, 11.14 seconds vs 14.25 seconds, p = 0.02) and explored less on an open field (one day post-injection, 21.81 m vs 16.83 m, p = 0.0004). Conclusions:We demonstrated the ability to filter proteins from the CSF of ALS patients and identified potentially pathologic proteins that were reduced in quantity. Additionally, we demonstrated the ability of unfiltered ALS-CSF to induce motor deficits in mice on the pole descent and open field tests and showed that filtration could prevent this deficit. Given the lack of effective treatments for ALS, this could be a novel solution for patients suffering from this deadly and irreversible condition.


Introduction
Studies have explored the role of the cerebrospinal fluid (CSF) in amyotrophic lateral sclerosis (ALS). The CSF proteome in ALS patients contains increased pro-inflammatory proteins and molecules compared to normal [1][2][3], possibly allowing for the spread of pathogenic proteins throughout the nervous system [4][5][6]. Infusion of ALS-CSF into mice over two weeks (at 0.25 ul/hr) led to TDP-43 accumulation, a decline in locomotor function, and alterations of muscle tissue [6].
We hypothesized that CSF filtration could remove pathogenic proteins and reduce motor phenotypes in ALS. Filtration was explored in the 1990s using a method called liquorpheresis to filter and reintroduce CSF into the thecal sac with a syringe pump [7,8]. Small trials of liquorpheresis showed subjective improvement in patient strength, but no difference in lung capacity or nerve conduction studies [7,8]. Studies on Guillain-Barré syndrome found that liquorpheresis restored nerve function and aided patient recovery [9,10]. Researchers speculated filtration of inflammatory proteins such as C5a, IFN-gamma, IL-2, and IL-6, but lacked studies confirming this [7,8]. Recently, another method of filtration of ALS-CSF was shown to prevent motor defects and motor neuron death when injected into mice [11].
One newer method to filter CSF is Neurapheresis™, which extracts CSF from the lumbar subarachnoid space 1 1 1 2 1 1 group at each time point. Two-way multiple comparisons ANOVA was used to any significant differences between experimental groups set at alpha=0.05, with post-hoc analysis done with multiple comparisons testing at alpha=0.05.

Proteomics
A total of 1,648 proteins were identified in the ALS-CSF samples (shown in Appendices). Of these, 58 were identified as key proteins for further analysis. [1,2,6,[23][24][25][26][27]. The average abundances and number of samples containing each protein in the unfiltered and filtered CSF can be found in Figure 1. Key proteins were grouped by functionality: there were 29 complement proteins, four serine protease inhibitors (SERPINs), three chitinases, 11 neurofilaments, tau proteins, and neuroendocrine proteins, and 11 other immune proteins.

FIGURE 1: Abundances of key ALS proteins before and after CSF filtration: complement system proteins
These plots visualize the abundances of key proteins in samples of each experimental group as analyzed by discovery proteomics. Each box-and-whisker barplot shows the median, Q1, Q3, standard deviation as error bars, and mean as a point. The abundance of proteins discovered in the single control sample are included as red dashed lines in relevant graphs. The molecular weight of each protein is included in the title of each plot. The yaxis shows abundance values on a log scale. The number of samples with protein identified is also included with the x-labels for each plot. P-values are shown, indicating results for Welch's t-test comparisons between groups.
UF: unfiltered ALS-CSF; 100kD: 100kD filtered ALS-CSF; ALS: amyotrophic lateral sclerosis; CSF: cerebrospinal fluid All proteins had lower average abundances when filtered, but 13 complement proteins, two SERPINs, one chitinase, three "neuro-proteins", and four other immunological proteins were statistically significantly less abundant when filtered. Comparisons between the groups are shown in Figure 1. Four complement proteins, one chitinase, and three "neuro-proteins" were not found in enough filtered samples to qualify for statistical testing. Thirty-two of the key proteins were identified in the control sample. Statistical comparisons between the control sample and the experimental samples were not done since there was a single control. The abundances of the control sample are shown in Figures 1-6.

Mouse motor tests
The pole descent test assessed maneuverability, coordination, and fine motor control [19]. Mice were expected to descend quicker with repeated trials. Only 11 of the 12 control mice performed the pole test. On day 1, the control mice descended at an average of 14.86 seconds, filtered mice at 16.43 seconds, and unfiltered mice at 16.83 seconds; on day 1.25, at an average of 10.61 seconds, 11.64 seconds, and 12.75 seconds, respectively; on day 10, at 10.06 seconds, 11.14 seconds, and 14.25 seconds, respectively. Two-way ANOVA between groups was significant at p=0.04. Post-hoc analysis showed that at day 10, unfiltered mice took significantly longer to descend the pole (p=0.02).
The open field test assessed gross motor activity [20], and it is said that mice typically explore the environment when placed in the open field [28]. The control mice, filtered mice, and unfiltered mice traveled an average of 19.95m, 21.81m, and 16.83m, respectively, on day 1, and 15.15m, 15.149m, and 14.52m, respectively, on day 10. Two-factor repeated measures ANOVA was significant at p=0.02, thus post-hoc analysis was conducted. On day 1, the filtered group traveled significantly further than the unfiltered group (p=0.0004). The control group also traveled further than the unfiltered group (p=0.01). None of the groups were significantly different at 10 days. Results are shown in Table 2 and Figure 7.

Filtered vs Unfiltered
Pole Test (mean time to dismount +/-SD, s) Two-way ANOVA between groups: p = 0.04

Discussion
We analyzed the proteome of the CSF from patients with ALS, characterized the changes in composition after filtration, and tested the effect of ALS-CSF on mouse motor activity. Proteomics identified over 1,600 proteins in the CSF, of which 58 were selected as key proteins implicated in the disease [1,2,6,[23][24][25][26][27]. This even results in the filtration of other proteins (mentioned in Table in appendices). However, as previously discussed, Neurapheresis is a safe treatment for cryptococcal meningitis and subarachnoid hemorrhage, so this does not appear to be a detriment. Interestingly, even though we used 100kD filters, many proteins under 100kD in size were significantly reduced in abundance, likely due to clustering and aggregation of proteins that occur in body fluids. Here, we discuss the implications of the identified proteins in the disease process of ALS.

Key proteins
The complement proteins mark targets for phagocytosis and elimination [29]. Excessive complement in neurons of ALS patients is associated with neuronal and glial degeneration [30]. C5aR1 inhibitors reduce motor deficits and increase survival time in rodent ALS models [31,32]. If complement inhibition improves ALS phenotypes, filtration of these out of the CSF could be beneficial to patients. The accumulation of misfolded SERPINs has been found in the tissue of brains affected by ALS [33] and in the CSF mouse models [34], but it is unclear if misfolding drives ALS or is a result of ALS. SERPINF1 has been shown to have a protective effect on motor neurons in vitro, leading to speculation that its increase in the CSF in ALS is a protective mechanism [35]. Given the lack of in-vivo studies, we cannot speculate if filtering out SERPINF1 has a negative effect on ALS progression. SERPINI1 is involved in neurogenesis and dendritic maturation, but increases have been associated with neurodegeneration in Alzheimer's [36]. Although no explicit link has been made with ALS, perhaps a similar neurodegenerative process occurs in ALS. Chitinases are produced by astrocytes and microglia and play a role in the nervous system's immune response [37]. CHIT1, CHI3L1, and CHI3L2 have been found to be elevated in the CSF of patients with ALS, with levels correlating with disease progression [37][38][39][40][41][42].
In-vivo studies have directly linked chromogranin A (CHGA) to ALS, demonstrating it promotes secretion of SOD1, leading to neuron degeneration [43,44]. Glial fibrillary acidic protein (GFAP) is known to increase during the process of gliosis and astrocyte-driven inflammation in the brain [45]. Amyloid precursor protein (APP) has been implicated in Alzheimer's, but its role in ALS is less clear. TDP-43, a key protein implicated in ALS, is involved in the splicing of APP [46]. Perhaps in ALS, APP accumulation is a result of neuronal damage rather than a causative agent. The MAPT, or Tau proteins, which have been associated with the formation of neurofibrillary tangles [47], and neurofilament chains (NEFL, NEFM, and NFH) [48] are increased in the CSF of ALS patients and correlate with disease severity, thus could be used as biomarkers to track disease [47][48][49][50][51][52][53]. Reducing levels of pro-inflammatory proteins such as CHGA and GFAP could be therapeutic for ALS and biomarkers such as APP, MAPT, and neurofilaments could confirm if filtration is modifying ALS progression meaningfully.
C-reactive protein (CRP) may contribute to damage in Alzheimer's [54], and a strong negative correlation between CRP level in ALS-CSF and clinical outcomes has been demonstrated [55]. In an ALS mouse model, there was a significant increase in GS+ cells (Glutamate synthase, encoded by GLUL), correlated with microglial activation, indicating possible mechanisms of inflammation [56][57][58]. ILF3 antisense RNA 1 (ILF3-AS1), while not studied in ALS, leads to increased IL-6 and TNF-a in astrocytes, indicating a potential role in inflammation [59]. Transthyretin (TTR) was shown to be dysregulated in ALS compared to control spinal cord samples [60] and Beta 2-microglobulin (B2M) has shown differential RNA expression in the peripheral white blood cells of ALS patients compared to control [61], making them potential biomarkers. S100-A9 mRNA levels were increased at end-stage ALS but deleting it in a mouse model had no impact on motor neuron survival and slightly accelerated symptoms [62]. In one study, S100B was found to be upregulated in ALS spinal cord samples [63], but in another, levels decreased with disease progression [64]. Genotypic frequencies of CST3 were not statistically significant in ALS patients compared to controls [65], suggesting no expressional difference in cystatin C. The roles of S100-A9, S100B, and CST3 are still unclear in ALS, thus they should be studied further.

Mouse motor tests
The open field measures gross locomotor activity and has been previously studied in the context of ALS [6,66]. On day 1, the unfiltered group traveled significantly less than the filtered control groups, suggesting a motor deficit could have resulted from the ALS-CSF to prevent exploration, but all groups traveled similar distances on day 10. In the open field, mice naturally explore a novel environment [67]; thus, we can hypothesize that unfiltered CSF hampered the tendency of mice to move around the open field on day 1. By day 10, the other two groups naturally reduced locomotion due to the lack of novelty, but the unfiltered group was still unable to explore.
The pole descent test has been tested as a measure of Parkinson's disease [19,68], but not ALS. All three mouse groups decreased time to descend between the first and second trials, the control and filtered CSF groups decreased their descent time a little more by day 10. The unfiltered group did show an increase in average descent time between day 1.25 and day 10 and had a significantly longer descent time than the control group at this timepoint. This suggests some change in the fine motor coordination due to ALS-CSF, resulting in decreased ability to quickly maneuver. The differences in fine motor control are only apparent on day 10, while differences in the open field are apparent on day 1, suggesting differential effects on these two aspects of behavior.

Limitations
One key limitation of this study is the underlying assumption that the alteration in mouse behavior might be due to key proteins that we identified here; however, other filtered cytokines and molecules could also play a role that we have not identified in this study. Additionally, by using mass spectrometry, we are restricted to analyzing relative changes in the identified proteins. We would have had to utilize methods such as enzymelinked immunosorbent assay (ELISA) to identify specific quantities and concentrations of proteins, but we believe that our methodology allows for adequate analysis of changes in the protein makeup of the CSF samples. Our samples are also inherently biased as we were only able to procure samples from male patients; thus, we should expand our future studies to include female patient samples. A limitation of our mouse studies is that we did not conduct studies that would allow us to monitor the recovery of the mice injected with ALS-CSF. CSF is recycled multiple times daily [69], thus these mice might have recovered function as the ALS-CSF was cycled out of their systems.

Conclusions
We analyzed the changes in the proteome of ALS-CSF before and after filtration by molecular weight and demonstrated reductions in numerous proteins implicated in ALS pathophysiology and various biomarkers. Additionally, we demonstrated the ability of the unfiltered ALS-CSF to induce gross motor deficits in mice using pole descent and open field tests and showed that filtration could prevent this deficit. Clinical trials will be needed to test the potential of Neurapheresis as a treatment for ALS. It has been shown to be a safe and effective treatment for other neurological conditions, and with the lack of effective and affordable treatments for ALS at the present time, this could be a novel solution for patients suffering from this deadly and irreversible condition. Future studies include the implementation of repetitive filtration in mice with ALS to determine the volume and frequency of CSF filtration that could serve a therapeutic purpose.