Background
Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease characterized primarily by the progressive loss of spinal motor neurons and cortical pyramidal cells. To date, hundreds of mutations in more than 20 genes have been implicated in the pathology of ALS, whereby mutations in genes coding for superoxide dismutase 1 (SOD1), TAR DNA-binding protein 43 (TDP-43) and fused in sarcoma (FUS) as well as a hexanucleotide expansion on chromosome 9 in open reading frame 72 (
C9ORF72) account for the most cases with a familial background [
1‐
6]. Despite our growing knowledge about genetics, a vast majority of ALS cases (~85-90%) is considered sporadic (SALS), showing no family history and spontaneous mutations in known causative ALS genes in only a low percentage of patients [
7,
8]. Moreover, while an increasing number of genetic causes for ALS and other motor neuron disorders have recently been identified, epigenetic factors remain largely undefined with exceptions.
Pathological protein aggregation and/or neuronal cytoplasmic inclusions of SOD1, TDP-43 or FUS are a hallmark of nearly all ALS cases [
8,
9]. While aggregates of SOD1 and FUS are mostly limited to patients with mutations in the corresponding genes, aberrantly distributed and aggregated TDP-43 is additionally evident in patients with mutations in other genes including
C9ORF72 as well as SALS, further pronouncing the central role of TDP-43 in ALS [
1,
9,
10]. TDP-43 is described as a ubiquitously expressed, multifunctional RNA-binding protein implicated in mRNA transcription and alternative splicing. Although shuffling between the nucleus and the cytoplasm, TDP-43 is located predominantly in the nucleus. When mutated or under conditions of stress, TDP-43 translocates to the cytoplasm where it is hyperphosphorylated and forms insoluble, ubiquitin-positive aggregates [
10,
11]. The nuclear clearance of TDP-43 as well as the aggregate formation is thought to be involved in ALS pathogenesis.
Recently, TDP-43 was identified as part of nuclear Drosha and cytoplasmic Dicer complexes [
12,
13] and thus also implicated in microRNA (miRNA) biogenesis [
14]. Mature 20–24 nucleotide miRNAs are predominantly negative post-transcriptional regulators of gene expression acting primarily by hybridizing with the 3’ untranslated region of its target mRNAs resulting in translational repression or degradation [
15]. Previous studies already showed that miRNA dysregulation can be observed in neurodegenerative disease models [
16] including ALS [
17]. MiRNA biogenesis starts with long primary transcripts (pri-miRNAs) cleaved by nuclear Drosha complex into shorter miRNA-precursors called pre-miRNAs. The pre-miRNAs are then transported into the cytoplasm where they are further processed by the Dicer complex into mature miRNAs [
15]. Before regulating transcriptome plasticity the miRNAs become part of the RNA-induced silencing complex (RISC) facilitating interactions between miRNAs and target mRNAs [
18]. It was shown that TDP-43 as part of Drosha and Dicer complexes binds to and promotes the cleavage of selected pri- and pre-miRNAs during their biogenesis. Knock-down experiments in cell lines could confirm 10 mature miRNAs as being dysregulated upon TDP-43 depletion [
14,
19].
In this study we address the question whether TDP-43 binding miRNAs are actually dysregulated in ALS patients and determined circulating miRNAs in samples of cerebrospinal fluid (CSF) and serum from patients with SALS. We compared miRNA levels between the CSF and the serum compartment and evaluated altered miRNAs as a potential indicator of decreased TDP-43 function in these easy accessible body fluids. Furthermore we could confirm dysregulated TDP-43 binding miRNAs in lymphoblast cell lines (LCLs) derived from SALS patients and genetically defined patients carrying mutations in the genes coding for TDP-43, FUS, SOD1 and C9ORF72, identifying gene specific miRNA alterations.
Methods
Patient cohorts and ethics statements
Appropriate approval and procedures were used concerning human subjects. With informed written consent and approved by the national medical ethical review boards in accordance with the Declaration of Helsinki (WMA, 1964), blood samples as well as CSF samples were drawn. CSF and serum samples from the same individuals were derived from 24 healthy controls and 22 ALS patients fulfilling the El-Escorial criteria for definite ALS. Patients were considered sporadic cases (SALS) due to a negative family history and no mutations in known ALS genes (Table
1). There was no correlation between miRNA levels and age (R
2 ≤ 0,129) or gender.
Table 1
Human CSF and serum samples
Controls | 24 | 12 male; 12 female | 53.2 ± 16.8 |
SALS | 22 | 12 male; 10 female | 54.5 ± 12.4 |
Lymphoblastoid cell lines
Epstein-Barr virus (EBV) transformed lymphoblastoid cell lines (LCLs) were generated from healthy controls as well as sporadic and genetically defined ALS patients. For this study we used LCLs from six healthy controls and eight SALS patients, again with a negative family history and no mutations in known ALS genes. Genetically defined ALS patients carried mutations in the genes coding for TDP-43, FUS, SOD1 or an expanded hexanucleotid repeat in C9ORF72. In detail we used LCLs of three TDP-43 mutant (N352S), seven FUS mutant (four K510R, two G478L, one R514G) and five SOD1 mutant (three R115G, two E100K) patients as well as LCLs of seven patients carrying the hexanucleotid repeat expansion in C9ORF72 with repeat lengths between 620 and 1100 bp (620, 800, 830, 950, 980, 1050 and 1100 bp) as determined by Southern blot analysis.
RNA isolation
RNA isolation was carried out with the miRNeasy Mini Kit (Qiagen) as specified by the manufacturer. RNA from CSF and serum was isolated in each case from 200 μl using Qiagen’s supplementary protocol for serum and plasma including the spike-in of 5 μl (5 nM) of a synthetic miR-39-3p of Caenorhabditis elegans (Cel-miR-39-3p) as a standard for adjusting different RNA isolation and reverse transcription efficiencies.
Reverse transcription and quantitative PCR
For reverse transcription we used the NCode VILO miRNA cDNA Synthesis Kit (life technologies) according to the manufacturer’s instructions. For CSF and serum samples same volumes of equally isolated RNA were applied to the reactions.
Quantitative PCRs (qPCRs) were run on a CFX96 Real-Time System (Bio-Rad) using the EXPRESS SYBR GreenER qPCR Supermix (life technologies), the Universal qPCR reverse primer included in the reverse transcription kit and miRNA-specific forward primers. MiRNA levels in CSF and serum were normalized relative to Cel-miR-39-3p while expression of miRNAs in LCLs was normalized relative to U6 snRNA using 2
-ΔΔCt-method [
20].
Statistical analysis
All the statistical analysis was carried out using the two-tailed Mann–Whitney U test. P-values smaller than 0.05 were considered statistically significant. *: p < 0.05; **: p < 0.01; ***: p < 0.001.
Discussion
MiRNAs present in bodyfluids are still controversially discussed concerning origin, function and mode of transport. Nevertheless, circulating miRNAs have been shown to be remarkably stable and to reflect disease conditions designating them as useful biomarkers [
25]. Several ALS-related proteins, including TDP-43 that is most central for ALS pathogenesis [
8,
11,
26], have been shown to be involved in miRNA biogenesis [
14,
19,
24]. Specifically in ALS, miRNA alterations are thus most plausible targets for biomarker development or even therapeutic approaches. Several miRNAs depend on TDP-43 for their biogenesis and have been identified as TDP-43 binding in
in vitro knockdown experiments [
14,
19]. We now report downregulation of most of these miRNAs in CSF and serum as well as in lymphoblastoid cell lines of sporadic and familial ALS patients.
In conditions of cell stress, in TDP-43 transgenic mouse models or in human ALS post-mortem tissue, wild-type TDP-43 or mutants thereof tend to re-localize from a predominant nuclear localization to the cytoplasm and aggregate in stress granules and/or ubiquitin-positive cytoplasmic inclusions [
8,
11,
27‐
30]. It seems at least plausible that the reductions in those miRNAs which bind to and require TDP-43 for their biogenesis
in vitro
[
14,
19] reflect an impaired TDP-43 function or altered subcellular TDP-43 distribution in ALS patients. In line with this consideration is the higher than expected number of regulated miRNAs amongst the TDP-43 binding miRNAs, compared to an unbiased approach: 5 out of 9 miRNAs (56%) were shown to be altered (mostly downregulated) in the CSF and serum, respectively. This is a much higher proportion than found in previous studies looking for regulated miRNAs in e.g. post-mortem brain tissue of patients with frontotemporal lobar degeneration with TDP-43 inclusions [
31] or in leucocytes from SALS patients [
17] by miRNA arrays. The question whether TDP-43 binding miRNAs are valid surrogate markers for ALS disease progression or useful for diagnosis is beyond the scope of this study and has to be clarified with higher samples numbers and prospective sample collection.
To our knowledge this work is generally the first study comparing CSF and serum miRNAs from samples matched at the level of individual probands. We detected profound “systemic” changes in the serum, which were just as pronounced as miRNA level alterations in the CSF compartment. This surprisingly robust alteration of TDP-43 binding miRNAs outside the brain and CSF compartment is in agreement with the ubiquitous expression and possibly disease-associated “global” functional alteration of TDP-43. Moreover, with the exception of miR-143-3p whose relative abundance showed a clear correlation between the two body fluids both in patients and in controls, the amount of most miRNAs was independently regulated between the two compartments at an individual-to-individual basis. Furthermore, by comparing the miRNA levels in CSF and serum of healthy controls we could show that some miRNAs, like miR-9-5p, miR-132-5p and miR-558-3p are more abundant in the CSF, while others are higher concentrated in serum. Thus, based on the limited number of miRNAs studied in our work, CSF miRNAs do not seem to be a simple mirror of the usually higher abundant serum miRNAs, nor do changes in the serum necessarily reflect alterations of CSF levels, which suggests that transition of miRNAs across the blood-cerebrospinal fluid barrier is not significantly contributing to their abundance in CSF or serum [
32].
Nevertheless, systemically dysregulated miRNAs could reflect relevant pathogenic aspects of ALS. Increasing evidence suggests that ALS is a systemic disease, with a “peripheral” pathophysiology, e.g. at the metabolic level [
33,
34] or with regard to connective tissue [
35]. Furthermore, altered serum miRNAs could partially reflect or contribute to peripheral nerve and motor endplate pathology or the failure of regenerative attempts. The latter possibility is suggested by the downregulation of both strands of serum miR-132, which has repeatedly been implicated in neuronal development and synaptogenesis [
36]. Moreover, this miRNA (and possibly others) might not only reflect peripheral pathophysiological mechanisms leading to neurodegeneration but could also be a basis for biomarker development in the future as specific molecular markers could turn out to be useful surrogate parameters for disease, even if a correlation between the CSF and serum compartments is not observed.
Overall, our findings point towards a very specific regulation at the level of each miRNA. This is in line with the known high target sequence-dependent, individual role of single miRNAs that can also be tissue or even cell-type dependent. Nevertheless, the fact that we observed mostly a downregulation but rarely an upregulation of TDP-43 binding miRNAs could at least partially be the result of a general default in RNA metabolism in ALS [
37].
Transformation of patient-derived lymphocytes by EBV transfection is routinely used to generate immortalized lymphoblastoid cell lines (LCLs) for preservation of DNA. As expected, this transformation process is accompanied and partially depends on epigenetic changes including certain miRNAs, e.g. miR-155 [
22,
23]. However, lymphocyte transformation seems to depend on specific miRNAs and the majority of miRNAs is most likely not contributing to the transformation process. We thus hypothesized that miRNA alteration resulting from specific mutations in ubiquitously expressed genes (as
TARDBP,
FUS,
C9ORF72 or
SOD1) might still be represented in LCLs. We therefore made an attempt to identify gene specific changes in miRNA levels in LCLs derived from sporadic ALS patients or patients with known gene mutations in
C9ORF72,
SOD1,
TARDBP or
FUS, which were compared to LCLs derived from healthy control individuals. We found both strands of miR-143 downregulated in all ALS-derived LCLs. As
SOD1-mutant patients largely lack TDP-43 pathology [
9] the underlying mechanism seems to be rather independent of TDP-43 malfunction. Conversely, robust downregulation of both strands of miR-132 and miR-574 are found only in LCLs derived from patients with likely TDP-43 pathology, but not in LCLs carrying a mutation in
SOD1. We could thus show that the known differences between SOD1 and non-SOD1 ALS regarding TDP-43 pathology [
9] are also reflected at the level of TDP-43 binding miRNAs, at least in LCLs.
As TDP-43 and FUS share striking functional and structural similarities [
6,
8] it is not surprising that mutation of either protein cause similar alterations of miRNA metabolism even in LCLs. Most interestingly, during this study FUS has been implicated in the biogenesis of several miRNAs that overlap with the TDP-43 binding miRNAs found to be downregulated in our study including
FUS mutant LCLs, e.g. miR-132 and miR-143. Moreover, during our study miR-9-5p biogenesis was shown to depend on FUS [
24], providing a plausible explanation why this miRNA, which we had originally chosen as a control miRNA because it does not bind to TDP-43, was drastically reduced in
FUS mutant LCLs.
Both strands of miR-132, which was the miRNA most robustly downregulated in our study in serum, CSF and LCLs, have been shown to exert multiple functions in neuronal development and morphogenesis. Additionally, downregulation of miR-132 has been detected in brains of patients suffering from Huntington’s and Alzheimer’s disease as well as from schizophrenia and bipolar disorders (summarized in [
36]). Importantly, reduced levels of miR-132 have also been found in brains of patients with frontotemporal lobar degeneration with TDP-43 inclusions, a condition closely related to ALS with regard to the molecular pathogenesis and TDP-43 pathology [
31]. Thus downregulation of miR-132 seems to be a common feature of several degenerative nervous system conditions. Moreover, even though very speculative, miR-132 downregulation may underline neurodegenerative facets of the psychiatric diseases schizophrenia and bipolar disorder.
Another miRNA found to be dysregulated in all types of samples studied here were both strands of miR-143. This miRNA has not been directly implicated in neurodegeneration so far. However, validated targets of miR-143 comprise proteins involved in cell proliferation and apoptosis [
38] as well as multiple proteins participating in actin cytoskeleton remodeling [
39]. Recently, experimental and genetic evidence has implicated disturbed actin dynamics in the pathogenesis of motoneuron diseases. For example, mutations in profilin 1, a protein required for actin polymerization, have been shown to cause ALS in a subset of familial cases [
40,
41].
Competing interests
The authors declare that they have no competing interests.
Authors’ contribution
AF and JHW planned the experiments, interpreted the data and prepared the manuscript. KM genotyped ALS patients. AF carried out the experiments. JHW and ACL supervised the project and gave conceptual input. All authors read and aproved the final manuscript.