and 12 months indicated significant improvements in attention, abstract conceptual reasoning, and impulse control (as measured by improvements in scores from ‘impaired >1.5 standard deviations (SD) from the mean, to ‘within normal range’ <1 SD from the mean) ([112]Fig. 4b). Subjective parent-, teacher-, and self-report questionnaire data at baseline, 6, and 12 months indicated significant improvements in overall executive functioning, inattention, hyperactivity/impulsivity for all raters, as well as self-reported improvements in fatigue ([113]Fig. 4c). The improvements seen on neuropsychological assessment were reflected in school reports taken pre-diet, 6-months on diet, and 12-months on diet by the same classroom teacher. Absenteeism reduced from 8.5 days per semester pre-KD, to 3 days per semester at 12 months on KD. Furthermore, there were decreased ear infections and need for antibiotics (parenteral and ear drops), with 8.5 courses of antibiotics per year before KD (2015–2020), and 3.7 courses per year on KD (2021-ongoing). We hypothesise that the ketogenic diet improved the vulnerability of KS-I's brain response to infection, thereby eliminating the cognitive regression episodes associated with the infections. Due to clear benefits of the KD, KS-I continues on the diet (currently ongoing for 3.2 years), without any further brain fog episodes. Fig. 4. [114]Fig. 4 [115]Open in a new tab Single-cell RNA Sequencing and neuropsychological data from a trial of the ketogenic diet (KD) in patient KS-I. (a) Plasma beta-hydroxybutyrate (BOHB) levels over 24 months on KD in patient KS-I. Dotted line represents normal reference (<0.5 mmol/l). The dashed lines indicate the ideal blood ketone ranges for the KD in paediatric epilepsy (2–5 mmol/L). The patient achieved good levels of BOHB on the ketogenic diet. (b) Objective neuropsychological assessment results at baseline (pre-diet) and 12 months on KD in patient KS-I. A higher negative z-score indicates more impairment, except for those measures denoted with ∗, for which a higher positive z-score indicates more impairment. Green indicates within normal range (<1 SD from the mean), orange indicates borderline impaired (>1 SD < 1.5 SD from the mean), red indicates impaired (>1.5 SD from the mean). Overall, KS-I showed improvements on formal testing most notably in attention and impulse control. (c) Subjective questionnaire data at baseline (pre-diet), 6 months on KD and 12 months on KD in patient KS-I, with reports from parent, teacher, and self. Higher positive z-score indicates more impairment, except for those measures denoted with ∗, for which a higher negative z-score indicates more impairment. Green indicates within normal range (<1 SD from the mean), orange indicates borderline impaired (>1 SD < 1.5 SD from the mean), red indicates impaired (>1.5 SD from the mean). There are increasing scores in the normal range on the diet (increasing green compared to red). (d) Bar chart showing top 5 enriched upregulated Reactome pathways in CD8+ T-cells for KS-I (12 months on diet) versus KS-I (baseline), including mRNA translation and translation termination. (e) CNET plot linking the top DEGs from enriched upregulated Reactome pathways in CD8+ T-cells for KS-I (12 months on diet) versus KS-I (baseline), which are predominantly RPL and RPS ribosomal protein genes. (f) Bar chart showing top 5 enriched upregulated Reactome pathways in natural killer cells for KS-I (12 months on diet) versus KS-I (baseline), including mRNA translation and translation termination. (g) CNET plot linking the top DEGs from enriched upregulated Reactome pathways in natural killer cells for KS-I (12 months on diet) versus KS-I (baseline), which are predominantly RPL and RPS ribosomal protein genes. The ketogenic diet partially reversed ribosomal protein dysregulation in KS-I To determine the effects of the KD on gene regulation, we performed scRNA-Seq of patient KS-I at 12 months on KD compared to KS-I at baseline. The scRNA-Seq at 12 months exhibited upregulation of Reactome pathways in CD8+ and NK cells related to mRNA translation and selenocysteine synthesis ([116]Fig. 4d and f), the same pathways that were downregulated at baseline in KS-I compared to sibling control ([117]Fig. 1c and g). The Reactome CNET plots ([118]Fig. 4e and g) further revealed that the following RPS and RPL ribosomal protein genes were upregulated (RPS4X, RPS14, RPS27A, RPL13, RPL18A) after 12 months on KD compared to KS-I baseline. These same ribosomal protein genes were downregulated at baseline in KS-I compared to sibling control. Discussion MDEMs are genetic disorders caused by mutations in epigenetic machinery, leading to multi-organ abnormalities including intellectual disability and neurodevelopmental disorders.[119]17, [120]18, [121]19 These disorders affect chromatin arrangement and have been implicated in the pathogenesis of neurodevelopmental disorders.[122]20, [123]21, [124]22 Our study provided a unique opportunity to conduct an integrated comparison of single-cell transcriptomes and bulk proteomes in individuals with KS compared to controls. These analyses have provided insight into the pathways of genes and proteins that are dysregulated in the PBMCs of individuals with KS. Specifically, our findings suggest significant dysregulation of gene translation processes in KS, in particular ribosomal protein genes (and their proteins). Although cytoplasmic (RPS and RPL) and mitochondrial ribosomal protein genes (MRPS and MRPL) were generally reduced at an RNA and protein level across different cell types in the patients with KS, patient KS-II exhibited upregulated ribosomal protein expression in CD8+ T-cells, suggesting a more complex interplay of epigenetic control on ribosomal function according to cell type. This highlights the strength and importance of single cell (and therefore cell type) analysis, as opposed to ‘bulk’ analysis which may mask these important signals. As supported by our proteomic data, we found a combination of upregulated (interferon immune signalling) and downregulated (azure granule lumen) immune pathways, highlighting immune dysregulation (rather than solely immune deficiency) in KS. We propose that our scRNA-Seq and proteomic data demonstrates the role of epigenetic machinery on ribosomes and ribosomal proteins, with secondary immune dysregulation in individuals with KS. The effects of epigenetic control on ribosomal protein function Given KS is a monogenic disorder, understanding the epigenetic implications of loss-of-function KM2TD may help us understand the downstream effects on cellular function. Lysine methylation of histone proteins (via KMTs) is an integral post-translational modification that regulates gene expression, whereby the location of the methyl lysine residue and degree of methylation influences the accessibility of chromatin to RNA polymerase.[125]^23 In our cohort, it is intuitive that a loss-of-function mutation in KMT2D induced a reduction in lysine methyltransferase activity and subsequent dysregulated gene expression. There is increasing evidence in multiple organisms regarding the importance of controlling ribosomal function during stress.[126]^24^,[127]^25 Estimated to use ∼60% of cellular energy,[128]^26 ribosomal biogenesis is an energy-intensive process that encompasses transcription and processing of ribosomal RNA (rRNA), in addition to transcription, translation, and transport of ribosomal proteins within the cell.[129]^27 As such, ribosomal biogenesis is intricately tied to the energy status of the cell. During periods of stress (including proteostasis imbalance, nutrient deprivation, oxidative stress), evolutionary mechanisms to conserve ribosomal function are integral for cell survival.[130]^25 Termed the ‘environmental stress response (ESR),’ this diversion of cellular energy from growth-related processes such as transcription, translation and ribosomal biogenesis, to autophagy, DNA damage repair and protein folding and degradation is associated with increased resistance to stress.[131]^28^,[132]^29 For example, yeast cells respond to stress by initiating the ESR, resulting in repression of approximately 130 ribosomal protein genes and ∼450 protein synthesis genes.[133]^30 Epigenetic enzymes such as lysine methyltransferase (KMT) and lysine demethylase (KDM) are intimately involved in ribosomal gene transcription and translational machinery, such that depletion of these histone modifier enzymes results in dysregulated ribosomal protein biogenesis.[134]31, [135]32, [136]33, [137]34 Notably, ribosomopathies, or conditions associated with aberrant ribosomal protein levels, have been implicated in the pathogenesis of various neurodevelopmental disorders. For example, downregulation of ribosomal protein genes RPL10 and RPS23 has been associated with congenital microcephaly, immature neurogenesis, autism, intellectual disability and progressive grey matter degeneration in humans.[138]35, [139]36, [140]37 As such, this ribosomal dysregulation profile we describe in KMT2D associated KS may serve as an epigenetic biomarker of environmental stress. As described in yeast cells above,[141]28, [142]29, [143]30 future studies are needed to explore whether ribosomal protein dysregulation is associated with conditions of environmental or ‘epigenetic’ stress in humans with neurodevelopmental disorders, perhaps by measuring longitudinal samples of scRNA-Seq and proteomics within individuals during periods of infection and when well. Compensatory mechanisms in KMT-KDM homeostasis Interestingly, the MLL1 complex (also known as KMT2A) pathway was significantly upregulated in the GO proteomic data in patients with KS compared to controls ([144]Fig. 3f and h). MLL1/KMT2A is another histone methyltransferase that trimethylates H3K4 to form H3K4me3 - a post-translational modification found at promoter regions of transcriptionally active genes.[145]^38 Specifically, the lysine methyltransferase proteins (KMT2A and KMT2C) and lysine acetylase proteins (KAT8 and KAT6B) exhibited increased abundance compared to controls ([146]Fig. 3h), and these proteins typically induce an ‘open’ chromatin state and promote gene expression.[147]^39^,[148]^40 Given the loss-of-function mutations in KMT2D in our KS cohort and subsequent ‘closed’ chromatin state, we hypothesise that the increased abundance of other KMTs and KATs may represent a compensatory mechanism to ‘reopen’ the chromatin and promote gene expression. Furthermore, reduced expression of lysine methyltransferases (KMTs) will result in a relative excess of its opposing lysine demethylases (KDMs), and subsequent imbalance of KMT-KDM homeostasis. We observed this via increased abundance of KDM2B and KDM5A in the DNA binding pathway of GO proteomic data ([149]Fig. 3h). Previous studies have reported an association between KDMs and ribosomal protein genes, such that deletion of KDM4 (also known as Rph1) increases expression of ribosomal genes by increasing RNA Pol II accessibility and RNA Pol I transcription in yeast cells under nutrient stress.[150]^41 The importance of H3K4 methylation in memory and learning Precise control of histone 3 lysine 4 (H3K4) methylation is integral for brain development and cognitive function, and dysregulation of these processes has been implicated in the pathogenesis of neurodevelopmental disorders.[151]20, [152]21, [153]22 Loss of function mutations in KMT2D result in a reduction in methyl marks required to stimulate expression of key neuronal differentiation genes, thereby reducing transformation of stem cells to neuronal cells.[154]^42 The importance of trimethylation of histone 3 lysine 4 (H3K4me3) marks in relation to learning and memory has been shown in animal models. In mice subjected to contextual fear conditioning, global levels of H3K4me3 in the CA1 region of hippocampus were increased when primed with cues associated with electrical shock, as compared to shocked mice with no learning cues.[155]^43 This increase in H3K4me3 occurred with memory retrieval, and knock-out of lysine methyltransferase (Kmt2a) in the CA1 after training reduced H3K4 me3 levels and performance of learning tasks.[156]^43 Furthermore, mice with Kmt2a knockout in the prefrontal cortex exhibit spatial working memory deficits compared to control mice,[157]^44 suggesting an integral role in lysine methyltransferases in cognition and learning. Ketone bodies are HDAC inhibitors which can counteract the ‘closed’ chromatin state and promote gene expression Notably, our n = 1 trial of the KD in a child with KS and debilitating brain fog episodes combined clinical data with comprehensive -omic analyses to unravel mechanisms of action of the KD in ameliorating these episodes. Given loss-of-function histone lysine methyltransferase activity results in a ‘closed’ chromatin state and subsequent gene repression, the production of ketone bodies which are known to be HDAC inhibitors[158]^6 may ‘reopen’ chromatin and encourage expression of genes. This hypothesis is supported by our scRNA-Seq data for KS-I, whereby 12 months on the KD increased the expression of previously downregulated ribosomal protein genes. This change in transcriptomic profile was associated with the resolution of brain fog episodes and improvement on objective neuropsychological assessment. A key strength of our study was the ability to explore comprehensive differences in the biological profiles in individuals with a monogenic disorder; and integrate this with detailed neuropsychological clinical data. The KD is considered a restrictive diet to adhere to, however, KS-I was able to demonstrate good compliance to the diet via urinary acetoacetate and serum BOHB levels in ranges recommended for paediatric epilepsy ([159]Fig. 4a). The parents also maintained a strict daily food diary via electronic application for the 3-year period, and the availability of this dietary data may prove useful in future studies investigating the effects of the KD on gut microbiome.[160]^45 Despite our small cohort, we were able to demonstrate statistically significant differences in gene expression and protein abundance between patients with KS and controls. This highlights the statistical power of scRNA-Seq in particular; and its potential to explore molecular function in individual patients with monogenic disease. The limitations of our study include not exploring chromatin arrangement via CHIP-seq[161]^46 or CUT&RUN sequencing,[162]^47 which allows for genome-wide profiling including histone modifications, and therefore evaluation of whether loss-of-function KMT2D directly affects ribosomal protein gene transcription. Interestingly, Jung et al.[163]^48 recently utilised CHIP-seq and bulk RNA sequencing to identify differences in the chromatin profile of 33 individuals with KS versus 36 controls. The top Gene Ontology terms enriched for differential H3K4me2 peaks in KS versus control included ‘rRNA metabolic process’, ‘rRNA processing’, ‘ribosome biogenesis’ and ‘ribonucleoprotein complex biogenesis,’ thereby suggesting that the ribosomal protein dysfunction we observed was associated with disruption to chromatin arrangement in KS. These differences in H3K4me2 peaks were reasonably correlated with transcriptional changes, with ∼70% of genes indicating changes in expression and H3K4 me2 in the same direction (Pearson's r = 0.4). Whilst animal models have been able to demonstrate the direct effects of the KD on neuronal cells through hippocampal tissue, we chose to use a less-invasive and more feasible method for human trials as PBMCs can be accessed at the same time as routine clinical blood tests. Subsequently one of the inherent limitations is that the effects observed in PBMCs may not directly reflect the effects on brain tissue. Nevertheless, PBMCs have been widely used in single-cell RNA sequencing of neurological disorders, including Alzheimer's disease[164]^49 and ischaemic stroke.[165]^50 Moreover, there is increasing interest in the role of the peripheral immune and CNS ‘cross-talk’ and therefore whether modifying peripheral immune cells can result in brain modification,[166]^51^,[167]^52 in addition to the potential direct effects of ketones on brain cells. Furthermore, we compared scRNA-Seq data with bulk proteomics, the latter has limitations due to ‘averaging’ the protein expression profile which loses potential cell type differences as shown in the scRNA-Seq. Single-cell proteomics is increasingly feasible and would present a powerful tool when combined with scRNA-Seq.[168]^53 Finally, ketone bodies have been shown to exert a multitude of systemic effects on immune function,[169]^54 inflammation,[170]^55 in addition to dietary changes to the gut microbiome,[171]^56 and therefore the HDAC inhibitory effects of the KD may not be the only therapeutic mechanism in this scenario. Whilst it is well-established that lysine methyltransferases have a direct effect on histones, KMTs and KDMs can also exert post-translational modifications outside the nucleus and regulate protein translation.[172]^31 Given that we found evidence of ribosomal protein dysregulation at both an RNA and protein level, we believe our findings support the direct effects of KMT2D on ribosomal protein transcription, however it is conceivable that loss-of-function KMT2D may have also induced post-translational modifications that directly modulate ribosomal proteins themselves, as previously described.[173]^57^,[174]^58 Further in vitro studies such as incubating KMT2D-deficient patient cells with butyrate and measuring activity of ribosomal proteins would strengthen our observations, however were outside the scope of the current study. Our study shows the potential of ‘-omic’ driven technologies to help inform mechanisms of action with ketogenic diet. We recognise that due to the small sample size this can not yet be generalised to all patients with KS, and acknowledge the differences in gene variants and other variables that may influence therapeutic effects. Conclusion Using scRNA-Seq and proteomic data, we demonstrate an association between epigenetic machinery (lysine methyltransferases) and the ribosome, and this ribosomal protein dysregulation is observed at a transcriptomic and proteomic level in individuals with KS. The ketogenic diet, perhaps via the HDAC inhibitory effects of ketone bodies, partially reversed ribosomal protein dysregulation in a child with KMT2D KS. These changes in transcriptomic signature were associated with concomitant resolution of brain fog episodes and improvement in cognitive function. We present an ex vivo replication of animal models of the KD in KS, and the utility of a diet in reversing some of the impacts of acquired intellectual disability and neurodevelopmental delay in individuals with KS. Single-cell RNA sequencing and proteomics presents a promising avenue for exploring other MDEMs and non-monogenic neurodevelopmental disorders and may provide insights into future potential therapies. Contributors Author Contributions: E. Tsang: Conceptualisation, Patient recruitment, Ketogenic diet administration and expertise, Methodology, Formal analysis, Funding acquisition, Writing—original draft, Writing—reviewing and editing. V. Han: Patient recruitment, Methodology, Formal analysis, Writing—reviewing and editing. C. Flutter: Conceptualisation, Patient recruitment, Writing—reviewing and editing. S. Alshammery: Formal analysis, Writing—reviewing and editing. B. A. Keating: Methodology (sample processing), Writing—reviewing and editing. T. Williams: Methodology (neuropsychological assessment and questionnaires), Formal analysis, Writing—reviewing and editing. B. Gloss: Methodology (bioinformatics of single-cell RNA sequencing), Writing—reviewing and editing. M. Graham: Methodology (proteomics), Writing—reviewing and editing, N. Aryamanesh: Methodology (bioinformatics of proteomics), Writing—original draft, Writing—reviewing and editing, I. Pang: Methodology (bioinformatics of proteomics), Writing—original draft, Writing—reviewing and editing, M. Wong: Immunology expertise, Writing—reviewing and editing, D. Winlaw: Patient recruitment, Writing—reviewing and editing, M. Cardamone: Patient recruitment, Writing—reviewing and editing, S. Mohammad: Writing—reviewing and editing, W. Gold: Writing—reviewing and editing. S. Patel: Conceptualisation, Patient recruitment, Methodology (sample processing), Supervision, Writing—reviewing and editing. R. C. Dale: Conceptualisation, Patient recruitment, Methodology, Verification of underlying data, Supervision, Funding acquisition, Writing—original draft, Writing—reviewing and editing. E.T and R.C.D accessed and verified the underlying data reported in the manuscript. All authors have read and approved the final version of the manuscript. Data sharing statement The scRNA-Seq data has been uploaded to Gene Expression Omnibus (GEO) under accession [175]GSE261686. The mass spectrometry proteomics data has been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD050547 and 10.6019/PXD050547. Declaration of interests The authors report no competing interests. Acknowledgements