In total, we studied 22 patients with RTT and mutations in MECP2 (21 females, 1 male); 15 male patients with MDS; 12 patients with RTT-like phenotypes and mutations in CDKL5 (1 female, 3 males), FOXG1 (1 female, 1 male), NR2F1 (1 female), GRIN2B (1 female), and AHDC1 (1 female), and 3 female patients without known mutation; as well as 13 healthy controls (7 females, 6 males) [Table 1]. We evaluated XCI in all the females. There was a similar proportion of random and skewed XCI in female controls and patients [Table S2].
Table 1
The composition of the studied cohort, which consists of individuals with Rett syndrome (RTT) with mutations in MECP2, MECP2 duplication syndrome (MDS), Rett-like (RTT-like) with mutations in different genes that are not MECP2 and healthy controls.
Individuals
|
Females
|
Males
|
Total
|
RTT
|
21
|
1
|
22
|
MDS
|
-
|
15
|
15
|
RTT-like
|
8
|
4
|
12
|
CDKL5
|
1
|
3
|
4
|
FOXG1
|
1
|
1
|
2
|
NR2F1
|
1
|
-
|
1
|
GRIN2B
|
1
|
-
|
1
|
AHDC1
|
1
|
-
|
1
|
Unknown mutation
|
3
|
-
|
3
|
Healthy controls
|
7
|
6
|
13
|
Characterisation Of RTT-MECP2 Versus Control
There were similar MECP2 mRNA amounts in patients with RTT and controls, whereas MeCP2 protein amount was significantly reduced in patients with RTT [Figure 1A, B]. We found a significant correlation between MeCP2 levels and the Pineda clinical severity score of our patients with RTT [Figure 1C].
Differential expression analysis of patients with RTT carrying MECP2 mutations versus healthy controls showed 3446 DEGs [Figure 2A], whereas proteomics differential expression analysis revealed 224 DEPs [Figure 2C, Table S3a,b]. Enrichment analysis uncovered significant overrepresentation of genes and proteins involved in several cellular functions and processes, some of which may be extrapolated to neuronal tissues and thus are especially interesting when trying to elucidate the pathomechanisms underlying RTT [Table S4a,b]. Although the correlation between transcriptome and proteome differential expression findings was not high [Pearson correlation coefficient = 0.09, p = 1.8e-11, Fig. 2D], we found 75 genes consistently deregulated at both the RNA and protein levels in patients with RTT. The concordant genes constitute strong candidates for deciphering some of the pathomechanisms behind RTT, as well as for establishing biomarkers of this disorder. Different pathways that repeatedly appeared significantly enriched with DEGs and DEPs, and that have some supporting evidence about being involved in RTT, are discussed in the following paragraphs.
Cytoskeletal actin-filament-based processes are significantly enriched in DEGs and DEPs [Figure 3A]. Extensive alteration of cytoskeleton-related proteins has been described in patients with RTT and RTT mouse models [27, 31, 41, 42, 63–65], which has been linked to alterations in dendritic spine morphology and axonal orientation and organisation observed in MeCP2-deficient neurons [63, 66–71]. Actin dynamics regulate morphological changes associated with synaptic plasticity and anchor and traffic membrane channels and signalling proteins needed for the correct maintenance of synaptic function [72]. Abnormal spine morphologies have been described in cognitive disorders such as RTT, autism, or Fragile X [72].
Among the DEPs, scaffolding proteins, actin monomers, and regulatory proteins were upregulated. p21-activated kinase 1 (PAK1, OMIM*602590) is a serine/threonine kinase that acts downstream of CDC42 and RAC; it is essential for regulation of the actin cytoskeleton. It is expressed in neuronal cells and controls dendritic spine morphogenesis and excitatory synapse formation [73]. Moreover, we found an upregulation of proteins related to cytoskeletal motor activities, such as tubulin monomers and kinesins, that could be implicated in axonal transport to the neuronal growth cone. Roux et al. [74] found alterations in the axonal transport of vesicular cargo, including Bdnf, due to dysfunction of the Htt-Hap1 pathway in Mecp2-null mice. The cytoskeletal motor proteins that we found differentially expressed in our patients could underlie these known alterations and lead to a profound dysregulation of axonal transport.
RTT fibroblasts also presented a downregulation of protein levels of Ca2+/calmodulin-activated Ser-Thr kinase (CASK, OMIM*300172), a ubiquitously expressed scaffolding protein that is involved in synaptic transmembrane protein anchoring in the brain. It belongs to the protein complex involved in the trafficking of glutamate ionotropic receptor NMDA type subunit 2B (GRIN2B) to the plasma membrane [75]. Interestingly, these genes have been linked to neurodevelopmental disorders with overlapping phenotypes with RTT [76–78], which makes CASK dysfunction a promising route towards understanding some of the pathomechanisms behind this disorder. In fact, CASK deficiency in both humans and mice has been shown to cause malfunctioning of the brainstem leading to hypoventilation, a common feature among patients with RTT [76].
ARMC9 (OMIM*617612), a known RTT spectrum gene, was significantly downregulated both as mRNA and protein in our RTT cohort. ARMC9 is involved in cilium assembly, signalling, and transport. Bi-allelic ARMC9 deficiency causes Joubert syndrome 30 (OMIM#617622), which presents some phenotypic overlap with classical RTT (delayed psychomotor development, intellectual disability, speech delay, seizures, hypotonia, and breathing abnormalities, among other symptoms) [79]. Given the implications in cytoskeletal dynamics of ARMC9 and the cytoskeletal abnormalities identified in patients with RTT, this could be a potential link between the shared phenotypes in these two disorders.
Another consistently downregulated mRNA and protein is COMT (OMIM + 116790). COMT is a methyltransferase required for the metabolism and degradation of catecholamine neurotransmitters, including epinephrine, norepinephrine, and dopamine [80]. Patients with RTT and RTT mouse models have shown low levels of these biogenic amines, and alteration in dopaminergic metabolism has been associated with the characteristic motor deficits of RTT [81, 82]. Hence, COMT downregulation in patients with RTT could be yet another sign of altered dopaminergic metabolism, either as a direct effect of MeCP2 dysfunction or as a secondary alteration due to the malfunctioning of the pathway.
An upstream analysis with ChEA3 revealed two TFs that are worth highlighting: CREB1 and SRF [Figure2B, Table S5a]. CREB1 (OMIM*123810), which is a known MeCP2 interactor, regulates transcription in processes of relevance for neuronal survival and memory consolidation, among others [83, 84]. In astrocytes, it even regulates genes related to mitochondrial function, vesicle dynamics, and the cytoskeleton [85]. Besides, one third of our DEGs are regulated by CREB1 and CREB1 itself was significantly upregulated in our cohort at the mRNA level. SRF (OMIM*600589), which is an integrator of mitogen-activated protein kinase (MAPK) and Rho-GTPase-mediated signalling, regulates cytoskeletal dynamics. SRF binds to the serum response element (SRE) sequence, present in a subset of cytoskeletal genes such as ACTB and immediate early genes (IEGs) [86]. Besides, SRF regulates neuronal morphology and activity-dependent transcription [87] and suppression of SRF-mediated transcriptional responses has been found to produce a reduction in dendritic complexity in cortical neurons, which could contribute to the neuronal spine dysgenesis phenotype observed in patients with an RTT-spectrum phenotype [88].
We also found a significant enrichment in genes and proteins related to vesicular activity located in the Golgi apparatus and the nuclear outer membrane-endoplasmic reticulum membrane network, as well as secretory vesicles [Figure 3B]. One of the dysregulated genes at the RNA and protein is alsin (ALS2, OMIM*606352), which is a nucleotide exchange factor for the small GTPase RAB5. RAB5 is an essential regulator of endocytosis that enables dynamic positioning of cell surface receptors to stimulate or attenuate signalling processes [89]. Alsin is expressed in several tissues, but predominantly in neurons where it modulates AMPA receptor trafficking, particularly of glutamate receptor subunit 2 (GluR2) [90]. Mecp2-null mice have shown a diminished presence of GluR2 in hippocampal neurons, leading to impaired long-term potentiation (LTP) [91]. In our data, we found an initial downregulation of ALS2 mRNA followed by an upregulation of ALS2 protein that, if extrapolated to neuronal cells, might contribute to alterations in AMPA receptors in the brain.
Another dysregulated gene found at mRNA and protein levels, which is also a known RTT spectrum gene, is prolyl endopeptidase like (PREPL, OMIM*609557). PREPL encodes a cytoplasmic protein with high expression in neuronal tissues that interacts with adaptor complex 1 (AP-1), a protein complex that organises membrane proteins in clathrin-coated vesicles at the trans-Golgi network and endosomes [92]. Loss-of-function mutations in PREPL cause myasthenic syndrome, congenital, 22 (OMIM#616224) in a recessive manner; the disease presents with delayed motor development, waddling gait, and hypotonia, with onset at birth [93, 94]. The interaction of PREPL with AP-1 promotes the dissociation of the protein complex from the plasma membrane. Therefore, PREPL deficiency leads to an increased amount of membrane-bound AP-1, whilst its overexpression results in AP-1 dissociation from the membranes [92]. Hence, the overexpression of PREPL that we observe in fibroblasts from patients with RTT could produce a lack of membrane-bound AP-1, which is necessary for the correct membrane positioning of soluble N-ethylmaleimide-sensitive factor attachment protein receptors (SNAREs) [95]. This alteration could play a role in the altered vesicular activity that we identified through enrichment analysis.
Enrichment analysis revealed downregulation of genes and proteins involved in ribosomal RNA (rRNA) processing and ribosome biogenesis, which would impact general protein translation in the affected cells [Figure 3C]. Previous studies have identified a reduction in mammalian target of rapamycin 1 (mTORC1) activity [96, 97], leading to dysregulation of ribosomal proteins and translational initiation factors [97, 98] and resulting in a global reduction in protein synthesis in RTT neurons [96, 97, 99]. This impairment in protein synthesis may help explain, at least in part, some of the differences observed between the DEGs and DEPs datasets. In addition, three of these DEPs are known MeCP2 interaction partners: MYBBP1A, DDX31, and DDX54 [100]. The downregulation of these proteins is a potential explanation for the alterations in rRNA processing and mRNA splicing that we detected in the enrichment analysis of both transcriptomics and proteomics in patients with RTT. The exact nature of the interaction between MeCP2 and these proteins is still unknown, but it could involve recruiting transcription repressor activity to regulate gene expression or helicase activity to play a part in RNA processing.
We also observed an enrichment of downregulated DEGs involved in mRNA splicing and spliceosomal complexes [Figure 3D]. Several groups, including ours, have reported splicing-related DEGs when analysing transcriptomics results [65, 101–103], which could be expected because MECP2 functions as a splicing regulator because it interacts with splicing factors [101, 104, 105]. Even so, a recent publication questions MECP2’s role as a global regulator of splicing because changing levels of MECP2 result in few differential splicing events compared to WT conditions [106]. Perhaps MECP2 only regulates alternative splicing of specific genes rather than doing it in a global way. Nevertheless, a significant amount of genes involved in mRNA splicing are repeatedly dysregulated in different transcriptomics experiments; thus, additional studies with the aim of clarifying MECP2’s role in splicing are needed.
Patients With RTT Versus Patients With MDS
RTT and MDS present some shared phenotypic traits such as intellectual disability, neurodevelopmental delay, hypotonia, epilepsy, and gastrointestinal problems. We compared the results of the differential expression analysis performed with patients with classical RTT and patients with MDS to identify common gene expression dysregulations that could shed some light into the pathomechanism underlying both syndromes.
Transcriptomics differential expression analysis of male patients with MDS versus male controls, and female patients with RTT versus female controls revealed 2465 and 3716 DEGs, respectively. Proteomics differential expression analysis returned 300 and 238 DEPs, respectively. Of these, 721 DEGs and 12 DEPs are shared between both groups. Because MECP2 expression is decreased in RTT and increased in MDS, we wondered whether they share DEGs that are expressed in opposite directions. In our cohort, 82 DEGs were positively correlated with MeCP2 expression levels (hence, upregulated in MDS and downregulated in RTT), and 100 DEGs were negatively correlated with MeCP2 expression (upregulated in RTT and downregulated in MDS) [Table S6a,b]. Enrichment analysis of those two gene sets revealed that pathways related to cytoskeleton and mRNA processing are altered [Table S4c]. In addition, we found other molecular functions and pathways commonly altered between the RTT and MDS cohorts, some of which could help to understand why these two syndromes share clinical traits.
The 82 DEGs downregulated in patients with RTT and upregulated in patients with MDS are overrepresented in terms related to mRNA processing and cell cycle [Figure 4A, Table S4c].
mRNA-splicing-related genes appear dysregulated in both analyses. Interestingly, 8 of the 82 DEGs are part of spliceosome complexes and another four are related to mRNA stability, processing, and maturation functions. These results seem to indicate that MECP2 is indeed important for mRNA processing, because both increased and decreased expression result in the alteration of genes involved in spliceosomal activity. These findings highlight the necessity of studying the role of MECP2 in mRNA splicing in greater depth.
DNA replication and cell-cycle-related terms have been described by us and by other groups [65, 107]. For example, Moran-Salvador et al. [108] found a downregulated group of genes involved in DNA replication and cell proliferation in hepatic stellate cells of Mecp2-null mice and suggested inhibition of Mecp2 phosphorylation as a liver fibrosis treatment.
When looking at the ChEA3 TF enrichment analysis that regulates the same 82 DEGs, we found several TFs, most of them zinc finger proteins, described as cell cycle regulators (E2F7, MYBL2, E2F1, TP53, ZNF695, ZNF888, ZNF670, and ZNF878), and also SRF, which we found in the RTT ChEA3 analysis [Figure 4C, Table S5b]. These results are consistent with our findings in transcriptomics enrichment.
The 100 DEGs upregulated in RTT and downregulated in MDS enrich processes related to neurogenesis regulation; signalling cascades, such as Wnt, BMP, and TGFß; and the cytoskeleton [Figure 4B, Table S4c]. The Wnt, BMP, and TGFß signalling pathways participate in neurogenesis regulation, synapse formation, myelin synthesis, and neural stem cell renewal [109–112]. The BMP signalling cascade has recently been found upregulated in neural stem/precursor cells from patients with RTT, and treatment with BMP inhibitors partially rescued the abnormal development detected in RTT brain organoids [113]. Besides, the Wnt, BMP, and TGFß signalling pathways are also involved in osteoblast activity and maintenance of cartilage [114–116]. Girls with RTT suffer from scoliosis, their bone mass density is low, and their bone fracture rate is four times that of the general population [117–119]. Scoliosis is the most commonly reported orthopaedic issue in patients with MDS, and osteopenia, contractures of joints, and fractures have also been reported [120, 121].
TF analysis with ChEA3 for the 100 DEGs upregulated in RTT and downregulated in MDS revealed that CREB1 (BH < 0.05 in differential expression analysis) is upregulated and SRF is downregulated, and that they regulate 39% and 22% of the shared 100 DEGs, respectively. Moreover, the following TFs related to neuronal function are also enriched in the ChEA3 analysis [Figure 4C, Table S5c]. HEYL (OMIM*609034) and GLIS2 (OMIM*608539) are two TFs that promote neuronal differentiation [122, 123]. In some tissues, GLIS2 activates the nuclear factor kappa B (NF-κB) pathway [124], and when upregulated, it can lead to neuronal apoptosis [125]. The NF-κB pathway regulates neural development and dendritic complexity and has been found to be upregulated in Mecp2-null mice or in neurons overexpressing Irak1 [126, 127]. The resulting reduced dendritic arborisation found in those mice resembles the phenotype observed in patients with RTT. In addition, attenuation of the aberrant NF-κB signalling ameliorates the dendritic phenotype and expands their lifespan. Another TF is NFATC4 (OMIM*602699) that regulates adult hippocampal neurogenesis and shares a common signalling process with BDNF for neuron maturation [128, 129]. BDNF (OMIM*113505) is a signalling molecule that modulates many aspects of neuronal development, synaptic transmission and plasticity. It is regulated by MECP2 and thus is dysregulated in RTT [130]. BDNF is required by GABAergic interneurons for their maturation. GABA signalling regulates hippocampal neurogenesis during adulthood and plays a role in the pathophysiology of anxiety disorders, even though it is not a core feature of RTT or MDS, it is a present trait in around half of the patients [121, 131, 132]. In mice, GABAaR stimulation decreases innate anxiety via NFATC4 activation, making it a pathway worth studying [133]. Finally, we have JUN (OMIM*165160), which is part of cJUN N-terminal kinase (JNK) pathway that is activated in RTT, and plays a role in neuronal migration and axon-dendritic architecture. Inhibition of the JNK pathway reduces breathing abnormalities in RTT mice and induced Pluripotent Stem Cells (iPSCs) neuronal models, and rescues the dendritic spine alterations [134]. Detecting so many TFs involved in neural processes and dendritic complexity – regulating between 16% and 37% of the DEGs shared between patients with RTT and patients with MDS – highlights the resemblance of both syndromes at a molecular level. Moreover, the therapeutic strategies that seem promising for one syndrome could also benefit the other if the correct gene dosages are reached.
Three out of the 12 shared DEPs, APPL2, CNPY4, and CTSC, regulate immune response and are downregulated in RTT and upregulated in MDS [Figure 4E, Table S6c]. Immune system alterations in MDS have been thoroughly described because patients with MDS suffer recurrent infections [135, 136]. Moreover, the TGFß signalling pathway, which is enriched with the 100 DEGs at transcriptomics level, is also implicated in innate and adaptive immunity [137]. Finding dysregulated pathways related to immunity in patients with RTT and patients with MDS highlights the importance of studying these pathways and genes in order to get a treatment to ameliorate the quality of life and lifespan of these children.
There are two common DEPs between RTT and MDS related to cytoskeleton functions: REPS1 and CNN1. REPS1 (OMIM*614825) is a signalling adaptor protein that mediates cytoskeletal changes as endocytosis, and the protein is upregulated in both syndromes. CNN1 (OMIM*600806) can bind to the cytoskeleton and produce smooth muscle contractions and is upregulated in RTT and downregulated in MDS.
The overlap between common significant DEGs and DEPs highlights two shared genes, MYO1C and HARS2 [Figure 4E, Table S6c]. MYO1C (OMIM*606538) is a myosin involved in cytoskeletal organisation and vesicle trafficking to the plasma membrane. HARS2 (OMIM*600783) is a mitochondrial histidyl-tRNA synthetase 2. At the RNA levels, it is consistently upregulated in patients with MDS and downregulated in patients with RTT. At the protein level, however, it is upregulated in both sets of patients. Ricciardi et al. [96] found the AKT/mTOR signalling pathway downregulated in Mecp2 null models. Their results point towards a combination between a deregulation of transcription followed by a limited ability to generate functional proteins. Our results support the possibility of a deregulation of the correct protein synthesis.
Patients With RTT Versus Patients With RTT-like Phenotypes
Our RTT-like cohort was recruited considering their resemblance to the RTT phenotype. It encompassed nine patients with mutations in five different genes plus three patients without an established molecular diagnosis. The greater heterogeneity of this group complicated the identification of DEGs, as well as the interpretation of the differential expression results. Therefore, we established a significance threshold of BH < 0.1 for transcriptomics to be able to call DEGs despite the data heterogeneity. We interpreted these results in comparison with those obtained in typical patients with RTT, searching for shared molecular alterations that could constitute common grounds in the pathogenesis of overlapping disorders of diverse genetic nature.
DE analysis of transcriptomics data revealed 63 genes consistently altered in patients with RTT and RTT-like phenotypes (25 upregulated and 38 downregulated) [Figure 5A, Table S6d], whilst proteomics data showed 81 proteins consistently dysregulated (39 upregulated and 42 downregulated) [Figure 5C; Table S6e]. Enrichment analysis of these genes revealed terms related to cytoskeletal organisation, mRNA processing, vesicular activity, and translation, which constitute shared molecular alterations common between patients with typical RTT and RTT-like phenotypes [Figure 5B and D].
It is remarkable that 31 out of the 63 common DEGs could be identified as potential SRF targets. This result is in line with the enrichment in processes identified in transcriptomics data, and may indicate the importance of this shared molecular alteration in RTT-spectrum phenotypes [Figure 5; Table S5d]. Actually, almost one third of the common DEGs are involved in cytoskeleton organisation and regulation, and some of these have important functions in neurons. Among these DEGs, we found SPON1 (OMIM*604989), which encodes an extracellular cell adhesion protein implicated in axon guidance by promoting and inhibiting the outgrowth of different kinds of axons [138], and STMN2 (OMIM*600621), a microtubule regulator present in differentiated neurons involved in neurite extension [139]. Decreased dendritic spine density has been reported in patients with RTT and RTT mouse models as well as in other neurodevelopmental disorders including Angelman syndrome, which has considerable phenotypic overlap with RTT [140, 141]. According to our data, the malfunctioning of cytoskeletal genes with prominent functions in neurite development could constitute a shared feature in patients with RTT-spectrum phenotypes. This malfunction could lead to neuronal spine dysgenesis and, consequently, to the emergence of disorders with common traits derived from this structural synaptic dysfunction. In fact, our analysis also detected an overrepresentation of several terms related to nervous system development and structure, supporting that common molecular alterations found in patients with RTT-spectrum phenotypes can impact neuronal phenotypes. Interestingly, the downregulation of ARMC9 observed in patients with typical RTT can also be observed in patients with RTT-like phenotypes. This alteration constitutes a link between RTT-spectrum disorders and the overlapping phenotype caused by loss-of-function variants in this gene.
The patients with RTT-spectrum phenotypes in our study shared a downregulation of SNRPC (OMIM*603522) expression that was also previously found in post-mortem brain tissue and embryonic stem cells of patients with RTT [97, 142]. SNRPC is a spliceosome component involved in 5’ splice-site recognition, so it may affect the splicing of many different targets and could constitute a shared mechanism of splicing dysregulation of patients with RTT-spectrum phenotypes. The dysregulation of splicing factors and regulators has been described in RTT as well as in other monogenic intellectual disabilities and in autism spectrum disorder (ASD) [143]. The biological pathways affected by these splicing defects (neuronal development, vesicular activity, and cytoskeleton organisation) are also shared, which could suggest common pathomechanisms behind these disorders [143].
Several of the rRNA processing and ribosome-biogenesis-related proteins found altered in patients with RTT were also consistently dysregulated in patients with RTT-like phenotypes, indicating that protein translation may be affected in all patients with RTT-spectrum phenotypes [Figure 5, Table S4d]. Among these, we found DDX54 (OMIM*611665), DDX31 (OMIM*616533) and MYBBP1A (OMIM*604885), which are MeCP2 partners and are linked to rRNA expression and pre-processing and could explain, at least to some extent, the shared dysregulation of ribosome biogenesis.
We also found a significant downregulation of transcripts and proteins involved in vesicular-activity-related processes and located in neuronal axons, further evidencing common alterations in synaptic activities in patients with RTT-spectrum phenotypes [Figure 5; Table S4d,e]. ZFPL1 (OMIM*619397) is a cis-Golgi membrane protein that regulates trafficking from the endoplasmic reticulum to the Golgi apparatus and maintains cis-Golgi structural and functional integrity [Chiu 2008]. NCALD (OMIM*606722) is a neuronal calcium sensor protein that is involved in calcium signalling. It interacts with clathrin and actin and is involved in the modulation of endocytosis and synaptic vesicle recycling. NCALD was found to bind clathrin only at low calcium levels, resulting in inhibition and modulation of synaptic vesicle recycling [144]. A study involving patients with spinal muscular atrophy found that NCALD overexpression impairs neurite outgrowth, and that a reduction in its expression was a protective modifier for developing the full disease [144]. They demonstrated that NCALD knockdown improved endocytosis in fibroblast cell lines of affected patients, a therapeutic strategy that could also be considered for patients with RTT-spectrum disorders overexpressing this protein.