Abstract
Hutchinson-Gilford Progeria Syndrome results from a mutation in Lamin A, and it is characterized by the incorporation of progerin into the nuclear lamina. Progerin expression leads to alterations in genome architecture, nuclear morphology, and epigenetic states, which in turn cause altered phenotypes in all cells of the mesenchymal lineage. Here, we report a comprehensive analysis of the transcriptional status of patient derived HGPS fibroblasts, including nine cell lines not previously reported. We observe that these fibroblasts carry abnormal transcriptional signatures, centering around five main functional hubs: DNA maintenance and epigenetics, bone development and homeostasis, blood vessel maturation and development, fat deposition and lipid management, and processes related to muscle growth. Stratification of patients by age revealed that a cohort of genes related to endochondral ossification and chondrogenic commitment show altered expression patterns in children aged four to seven years old, where this differentiation program starts in earnest, related to the growth of long bones. We further report changes in lamin associated domains and 3D genome organization around a cohort of genes of interest, identified in this study.
Introduction
Hutchinson-Gilford progeria syndrome (HPGS) is a rare disease, characterized by a severe premature aging phenotype (Gilford, 1897; Hutchinson, 1886). Patients appear normal at birth, with a disease onset around 2 years of age when they present with slow growth rate, short stature, and marked lipodystrophy with a characteristic loss of subcutaneous fat. With progression, patients experience arthritis, joint contracture, osteoporosis and stiffening of blood vessels. Average life expectancy for HPGS patients is 13.4 years, with myocardial infarction and stroke being the predominant causes of death. (Foundation, 2019).
HGPS is caused by a thymidine substitution mutation at cytidine 608 (GGC>GGT), within exon 11 of the gene that encodes for Lamin A (LMNA) (De Sandre-Giovannoli et al., 2003; Eriksson et al., 2003). Lamin A is a structural protein of the nuclear envelope, playing an important role in genome structure and nuclear integrity. The mutation does not result in an amino acid substitution, frame shift or early termination. Rather, it induces the usage of a cryptic splice site and the subsequent deletion of 50 amino acids at the C terminus. In turn, this deletion impedes downstream processing of the protein and results in conservation of C terminus farnesylation (Davies et al., 2009). This abnormal protein product is called Progerin. Aggregation of Progerin leads to a disruption of the normal nuclear envelope meshwork, leading to deformed nuclei, nuclear stiffening, and defective mechanotransduction (Apte et al., 2017; Goldman et al., 2004; Lammerding et al., 2004).
Of particular interest, progeria patients present with various defects in all tissues of the mesenchymal lineage. Specifically, regarding bone physiology, patients show reduced stature, generalized osteopenia, thin calvaria, and clavicle regression with absence of medial and lateral ends, as well as resorption of the distal bony phalanges and anterior ribs (Chawla et al., 2017; Cleveland et al., 2012; Gordon et al., 2011; Nazir et al., 2017). These phenotypes are more striking when considering that at birth, bone structure appears normal. Previous observations, where fibroblasts from progeria patients showed abnormal levels of aggrecan (Lemire et al., 2006) a marker of chondrogenic differentiation, led us to hypothesize that fibroblasts derived from HGPS patients - themselves a cell of the mesenchymal lineage – could harbor vestigial transcriptional signatures to abnormal mesenchymal stem cell commitment, which would become apparent by comparing them with different normal age group controls.
In this study, we conduct a comprehensive analysis of previously published RNA-seq datasets for HPGS fibroblast lines harboring the typical C>T mutation in the LMNA gene and provide transcriptomics data for nine previously unreported cell lines. By comparing the transcriptional profile of HPGS fibroblasts with cells derived from age matched controls, healthy adults, and healthy old adults, we provide insight into important defects in repair biology, metabolism (calcium, lipid), and other areas of mesenchymal cell lineage import. Our results show that transcription of genes involved in negative regulation of chondrocyte commitment is compromised at an early age, that temporally correlates with the onset of postnatal endochondral ossification, and onset of symptoms. We report that genes central to differentiation commitment into the chondrogenic-osteogenic lineage identified in this study (PTHLP, RUNX2, BMP4 and WNT5a) show either abnormal genomic architecture or lamin associated domain alterations.
Our results support the hypothesis that defects in the initial steps of chondrogenesis commitment are a potential mechanism for mesenchymal stem cell depletion that later results in abnormal adipogenesis, diminished microvasculature homeostasis, and poor wound repair observed in HGPS patients.
Results
Batch correction is essential for comparison among patient and normal cohorts
To determine gene sets that were consistently misregulated in Progeria fibroblasts, we collected both newly generated and previously published RNA-seq data from all available Progeria patient fibroblast cell samples as well as fibroblasts from control individuals in different age groups (Supplementary Tables 1-3) (Fleischer et al., 2018; Ikegami et al., 2020; Köhler et al., 2020; Mateos et al., 2018). When the resulting transcriptomics data were analyzed by principal component analysis (PCA), datasets clustered according to laboratory of origin regardless of diagnosis, making direct comparisons impossible (Supplementary Figure 1A), as has been previously observed in Progeria transcriptomics analysis (Ikegami et al., 2020) However, after batch effect correction (materials and methods (Zhang et al., 2020)), the first principal component was able to segregate the samples depending on progeria/non progeria origin (Supplementary Figure 1B) which enabled direct comparisons between patients and controls from different age groups which originated from different sources.
Gene ontology analysis of transcriptional changes reveals eight clusters of biological activity affected in HGPS
To derive sets of genes differentially regulated in Progeria, we divided the patient samples into Young (0-8 years old) and Teen (>13 years old) categories based on the idea that different developmental processes take place in these age groups. We then compared each patient group to age matched controls, middle aged adults, and older adults and selected up and down-regulated genes (FDR adjusted p-value < 0.001) for gene ontology analysis. Teenage HGPS patients showed the smallest number of genes changing expression levels when compared to age-matched controls, but the highest number of genes up/down regulated when compared to healthy middle aged or old adults (Table 1). Importantly, comparisons between the normal children cohort and the adult normal controls yield few significant changes, suggesting that HGPS is the driver of the differences observed among cohorts.
The largest number of differentially expressed genes (Nine hundred and seventy) were related to the biological processes of epigenetic programming and DNA maintenance (Fig 1-A). Gene ontology pathways included in this cohort include epigenetic regulation of gene silencing, senescence-associated heterochromatin, DNA repair, DNA recombination, and histone methylation/modifications. This is in line with extensive previous evidence of epigenetic and DNA repair misregulation in Progeria (Aguado et al., 2019; Gonzalo and Coll-Bonfill, 2019; Misteli and Scaffidi, 2005)These genes were predominantly over-expressed in young progeria patients when compared to all controls (Fig 1-B, 1-C) and between teenaged patients compared to their age matched control and adults (Fig 1-B, 1-D).
The next biological process with the most misregulated transcription among age groups was tissue repair. It bears mentioning that most of the nucleosome associated proteins and histones identified in the DNA maintenance cluster, as previously described, overlap with the repair GO terms. In all, 585 genes are differentially expressed between cohorts (Fig 2-A), with HGPS patients predominantly over expressing targets pertaining to the organization of the extracellular matrix (145 genes, Fig 2-B). Interestingly, signatures related to coagulation, Bone Morphogenic Protein signaling, and response to wound healing are all downregulated in young HGPS patients when compared to adults (Fig 2-C).
With regards to biology of mesenchymal tissue, mis-regulation of transcription was observed in bone (165 genes. Fig 3), fat (261 genes. Fig 4), blood vessel homeostasis (131 genes. Fig 5), and muscle (261 genes. Fig 6). All of these lineages have been observed as compromised in HGPS patients, showing phenotypes like osteopenia, early atherosclerosis, and lack of subcutaneous fat deposition. (Gordon et al., 2011; Hamczyk et al., 2018; Xiong et al., 2013)
Bone biology mis-regulation appears to be stratified into two distinct hubs: downregulation of pathways related to calcium homeostasis and transport and upregulation of pathways involved in cell differentiation of chondrocytes and osteoblasts (Fig 3-A,3-B). Comparing young patients’ transcriptional profile to that of old adult normal controls shows upregulation of genes involved in late osteogenic commitment. Such is the case of ontology terms that include osteoblast differentiation, ossification, and chondrocyte differentiation (Fig3-C). In parallel, genes included in pathways related to cation homeostasis, particularly calcium, are downregulated (Fig 3-D). This comparison is similar to that of young patients and normal middle-aged adult controls (Fig 3-E).
Lipid homeostasis and transport pathways are downregulated in all comparisons between samples from young patients. Young HGPS samples show downregulation in pathways involving fat cell differentiation when compared to age-matched controls, in line with observed fat deposition defects in patients (Revechon et al., 2017) (Fig 4 A-C)
Gene ontology terms related to blood vessels show upregulation of blood vessel morphogenesis, but a decrease of blood vessel maturation when comparing samples from young patients to either age-matched or old controls (Fig 5 A-C). This phenotype persists in comparisons between teen aged patients and their age-matched controls (Fig 5D). Further, in a possible contribution to the circulatory defects observed in HGPS patients, both young and teen-aged patient samples show a marked upregulation of genes in pathways related to muscle development and cardiac muscle differentiation and development when comparing both HGPS cohorts against adult controls (Fig 6 A-C).
To further refine our findings related to young age patients and the mesenchymal phenotypes observed, we stratified the young HGPS patients into two age groups: early infancy (0-3 years old) and children (4-7 years old), comparing these cohorts to age matched, middle age and old controls as before. The total numbers of genes up and down regulated between comparisons are described in Table 2.
About forty percent of genes upregulated in early infant patients when compared against their age matched controls are related to epigenetic modifications as described in the previous analysis (Fig 1). Specifically, the GO term “HDACs deacetylate histones” shows the highest enrichment in this group. These same genes are represented in the comparison with the middle age-old adult control groups.
In contrast, the comparison of patients aged 4-7 y/o to their age matched controls show an enrichment for ossification and calcium homeostasis. Out of the cohort of genes identified as downregulated in this comparison (102), about ten percent are related to early chondrogenesis events. Among those, parathyroid hormone related protein (PTHLH), insulin like growth factor (IGF1), bone morphogenic protein receptor 1B (BMPR1B) and collagen 10a1 (COL10a1), are an integral part of early commitment and control necessary for the triggering of endochondral ossification (Bradley and Drissi, 2010; Green et al., 2015; Maruyama et al., 2010). Genes related to ossification and calcium homeostasis were also identified as downregulated in this cohort. Further, upregulated genes in age-matched and adult comparisons include BMP4, WNT5a, and several genes related to WNT5a biology which play an important role in skeletal development (Fig 7A-B).
Genes of interest are impacted by chromatin compartment switches or abnormal lamina associated domains (LADs)
Chromosome conformation capture was performed on skin fibroblasts of parents of HGPS patients (Mother AG03257, Father HGADFN168) and two HGPS fibroblast cell lines (HGADFN167 and AG11513) (Supplementary Table 4). Genome-wide contact maps for all of these cells show globally similar patterns of chromosome conformation, with the exception of a translocation between chr3 and chr11 in AG11513 cells (Figure 8A). At a whole chromosome folding level, we note that progeria fibroblasts show a decrease in telomere interactions and an apparent loss of ‘Rabl’ like structure (Figure 8B). This type of Rabl structure loss was previously observed after DNA damage in fibroblasts (Sanders et al., 2020). We find that topologically associating domain (TAD) structure is preserved in Progeria fibroblasts, with even an increase in TAD boundary strength as compared to healthy parent controls (Figure 8C). This pattern was also previously observed in DNA damaged fibroblasts, suggesting that some of these features of genome structure may relate to the increased constitutive levels of DNA damage observed in these patient fibroblasts.
Using principal component analysis of Hi-C data at 250 kb-resolution, we classified genomic regions into open euchromatin (A) or closed heterochromatin (B) spatial compartments according to positive and negative values of the first eigenvector, respectively. Since accelerated senescence has been previously observed in progeria fibroblast lines (Bridger and Kill, 2004; Wheaton et al., 2017), we performed compartment analysis on cells belonging to a matched father-child pair, both at an early and late passages. Samples were collected for the HGADFN168 cells (Father) at passages 12 and 27, and for the HGADFN167 (Child) at passages 12 and 19. Our results show that in the parent cell line there are small changes in compartment strength (the degree of preference for interactions within the same compartment vs. between different compartments). In contrast, there is a predominant loss of A compartment strength in the late passage HGPS cells (Figure 8E). Interestingly, although the compartmentalization strength appears to be altered, the genome-wide compartment identity remains consistent among cohorts, with small changes in compartment identity appearing sporadically (Figure 8D). The loss of compartment strength at later passages in Progeria cells is consistent with previous observations, though this more deeply sequenced, and less noisy dataset shows that compartments are more preserved in Progeria cells than previously observed (McCord et al., 2013).
To test whether the up or downregulation of genes identified through this study relates to compartment switching, we evaluated the compartment identity of all up or downregulated genes, based on their genomic coordinates, in the healthy and patient datasets. While differences between groups are modest, a subset of the genes that are downregulated in the 0-7 HGPS/normal age-matched comparison shift toward the B compartment. This type of trend is also present when comparing HGPS samples from the 0-7 cohort to middle-aged adults but disappears when comparing them to old controls (Figure 8F-H).
To test whether the up or downregulation of genes identified through this study relates to abnormal distribution of lamin associated domains, we compiled the genomic coordinates of these genes with published Lamin ChIP-seq and DamID-seq data (Dekker et al., 2017; McCord et al., 2013). We observe that differences between all groups are significant, with HGPS intensity values markedly higher than their control counterparts (Supplementary Figure 2A and B, Kruskal-Wallis p<0.001). However, we found that this shift in Lamin association is true throughout all genomic regions, in an overall dampening of the lamin association signal in HGPS patients (Supplementary Figure 2C), with small regions switching from associated to dissociated from the nuclear lamina. Comparing normal fibroblast cell lines HGADFN168 (belonging to a father of an HGPS patient) and HFFc6 (human foreskin fibroblast), we observe that both show a characteristic bimodal distribution in LAD intensity around markedly positive (LAD) and negative (non-LAD) values (log2 Lamin/input). Remarkably, in contrast, LAD values for the HGPS cell line HGADFN168 distribute normally, around a value of zero: the intensity of both types of association is reduced in HGPS cells, consistent with an overall misregulated interaction between chromosomes and the nuclear lamina (Supplementary Figure 2D).
Downregulated genes of interest related to osteogenesis and chondrogenic proliferation in the early childhood cohorts, as identified by transcriptomics, show alterations in compartment identity or lamina association, but interestingly these changes are not necessarily all the same. For example, the WNT5a gene is located exactly at the boundary between an A/B compartment in both HGPS samples, and in a well-defined A compartment in both parental cell lines. (Fig 9). Importantly, the HGPS in this comparison are from patients in the older childhood cohort, where the marked increase in WNT5a expression that occurs in the younger cohorts disappears. In a contrasting phenotype, the gene that encodes for parathyroid hormone-like hormone (PTHLH) is in a conserved A compartment in both parental and HGPS cell lines (Supplementary Figure 3).
Discussion
In this study we present a comprehensive analysis of the transcriptome of HPGS patient derived fibroblasts, stratified by age, compared to their age-matched controls. Since progeria patients present with accelerated aging, we also compared this data to that derived from middle-aged and old adults. Strikingly, HPGS expression profiles do not phenocopy gene ontology analysis from either set of adult controls, suggesting an altered aging paradigm.
Given the critical role that Lamin A plays in nuclear structure, formation of heterochromatin, and subsequent gene silencing (Lammerding et al., 2004; Leemans et al., 2019), it is not surprising that one of the main biological processes affected in our study, as per gene ontology analysis, is DNA maintenance and epigenetics. Upregulation of genes in the HDAC family is the largest discriminator between Progeria and age-matched normal controls in the 0–3-year-old cohort and 4–7-year-old age groups, and this upregulation persists in comparisons with the adult controls. Our data suggests an upregulation of genes belonging to ontology terms such as negative epigenetic regulation of genes, chromatin silencing, DNA repair, double strand DNA repair and DNA recombination across all age comparison with young patients. Overexpression of these genes are a potential overcompensation mechanism for defects found in DNA repair (Aguado et al., 2019; Gonzalo and Kreienkamp, 2015; Komari et al., 2020; Reviewed by Misteli and Scaffidi, 2005). In older HGPS samples, downregulation of genes related to histone methylation, histone modification, chromatin modifying enzymes, and DNA recombination was also observed.
Concomitant with changes in gene expression of epigenetic factors, Progeria fibroblasts show alterations in chromosome structure, as has been previously observed (McCord et al., 2013). The increased resolution and clarity of Hi-C data presented here enables us to describe these alterations more clearly. Spatial compartmentalization is weakened in Progeria patients with increasing passages consistent with microscopically observed loss of heterochromatin (Goldman et al., 2004). Unlike compartments, topologically associating domain (TAD) boundaries are preserved during in Progeria cells, even at higher passages. When we compare gene expression changes to chromosome structure changes, we observe that many genes consistently upregulated across Progeria patients are in the open, A compartment across all Hi-C samples, suggesting that these gene expression alterations are not associated with dramatic structural change. Downregulated genes are more likely to exhibit switching into the B compartment in Progeria patients.
A compromised transcriptional landscape, related to developmental milestones, points towards a compromised mesenchymal stem cell niche
We further describe up and down regulated biological function clusters that could play a detrimental role in bone, fat, joint and vascular homeostasis. It has been proposed that progerin accumulation in the nucleus results in a defect in the mesenchymal stem cell lineage which gives rise to osteoblasts, chondrocytes, adipocytes, pericytes and myocytes (Reviewed by Andrzejewska et al., 2019). Overall, our results are in concordance with previous reports of transcriptional mis regulation in mesenchymal lineages (Csoka et al., 2004), but the age group comparisons we present here further refine these observations to a temporal effect on lineage commitment.
iPSC models in which HGPS fibroblasts were reprogrammed to stem cells, and further differentiated into mesenchymal stem cells lineages were characterized by nuclear dysmorphia and increased DNA damage, but resulted in confounding results on differentiation, showing either limited differentiation potential, or no significant changes (Crasto and Di Pasquale, 2018; Xiong et al., 2013; Zhang et al., 2011). Interestingly, the most common finding in HGPS-derived iPSCs further differentiated into other lineages is premature senescence and the presence of progerin and misshapen nuclei (Reviewed by Lo Cicero and Nissan, 2015). In a more direct approach, overexpression of progerin A in umbilical cord derived MSCs and other MSC systems resulted in a reduced capacity of differentiation into chondrogenic, osteogenic and adipogenic capacity in vitro (Mateos et al., 2013) and deficient proliferation and migration (Pacheco et al., 2014). Spurred by these collective findings, our age-stratified comparisons seek to answer the question of whether this abnormal cell fate regulation occurs predominantly in a particular lineage.
Failed arrest of chondrocyte hypertrophy as a potential mechanism of MSC depletion in HGPS
Transcriptional upregulation of several members of the WNT5a signaling cascade and downregulation of expression of PTHLH, in early age comparisons described in this study, point to an essential defect in early endochondral ossification control and osteogenesis. Specifically, WNT signaling has been previously implicated as a connection between progeria-like syndromes and a defective deposition of extra cellular matrix, essential in bone development (Andrade et al., 2017; Green et al., 2015; Hernandez et al., 2010). Not unlike the phenotypes observed in HGPS patients, bone defects are also present in murine models for HGPS, with young mice (4-week-old) showing decreased trabecular thickness and number, bone volume and decreased mineral density (Hernandez et al., 2010). Not surprisingly, murine pre-osteoblasts forced to express mouse progerin failed to mineralize and show decreased levels of Runx2 expression and alkaline phosphatase upon osteoblastic induction (Tsukune et al., 2019). Further, Lamin A knockdown inhibits osteoblast proliferation and impairs osteoblast differentiation in an MSC in vitro differentiation model (Rauner et al., 2009).
Unlike other MSC fates, like adipogenesis, which occurs downstream of a positive energy intake all through life, bone formation occurs in a carefully orchestrated manner, at specific timepoints during development and early life. Secondary endochondral ossification that occurs in long bones like the humerus and femur, initiate in early childhood, activate intermittently, and finalize with the fusion of the growth plate during the teenage years (Dimeglio et al., 2005; Xie and Chagin, 2021; Zoetis et al., 2003). This process is tightly regulated by several signaling networks aimed at balancing longitudinal growth of the bone with maintenance of precursor cells, like MSCs and quiescent chondrocytes. The transcriptional repression of PTHLH and upregulation of WNT5a in this niche would result in early depletion of the latter, as cells continue down a pathway of chondrocyte proliferation and hypertrophy unimpeded (Bradley and Drissi, 2010; Olsen et al., 2000; Usami et al., 2016). The deficit in clavicular development in HPGS patients could provide some insight as to the spatiotemporal consequences of MSC/chondrocyte pool depletion. The clavicle is unique because while it is the first bone to start ossification during the embryonic period (Ogata and Uhthoff, 1990), it is the last to complete the process. In early events, around the fifth week of gestation, the two primary ossification centers, formed from mesenchymal tissues, fuse to form the middle of the clavicle. The complete ossification of the clavicular epiphysis occurs via endochondral ossification during the teenage years, with medial ossification beginning at the onset of puberty (Ferguson and Scott, 2015; Langley, 2016). In turn, bone deposition at the terminal ossification center, must occur in adolescents between 11 and 16 years of age, for the process to successfully proceed to epiphyseal plate ossification in early adulthood. (Schulz et al., 2008). Taken together, our observations point towards a “fail to arrest” phenotype that results in over-commitment towards the hypertrophic chondrocyte lineage in very young HGPS patients. We hypothesize that this results in a premature depletion of the chondro-osteoprogenitor pool and, at later stages, of the bone marrow-derived mesenchymal stem cells that feed into this niche with later in life result in the incomplete endochondral ossification of the claviculae of HGPS patients (Movie 1).
Deficient repair responses exacerbate the depletion of tissue resident MSC pools
A secondary function of tissue resident MSCs is to aid in repair during wounding and disease.
Our data shows that dermal fibroblasts derived from HPGS patients present with abnormal repair, including downregulated signatures of wound healing, hemostasis, and BMP signaling. These observations recapitulate previous reports of HGPS patient-derived skin precursors, which can differentiate intro fibroblasts and smooth muscle cells (smooth muscle alpha actin positive – myofibroblasts). In the precursor stage, these cells express low levels of progerin in vivo and in vitro, but differentiation into the lineages increases progerin expression and deposition in the nucleus (Wenzel et al., 2012). In concordance with our findings, a progeria murine model, deficient in the downstream processing of lamin A, shows abnormal skin wound repair, with prolonged time to wound closure, poor vasculogenic signaling and angiogenesis, which includes a limited mobilization of bone-marrow derived progenitor cells (Butala et al., 2012). Related to this phenotype, the LMNAΔ9 murine progeria model for HPGS show a significant decrease in postnatal fibroblast proliferation, with accelerated senescence levels in cell lines derived from kidney, lung, skin, and skeletal muscle (Hernandez et al., 2010). Further, a decreased epidermal population of adult stem cells was observed in another murine model of HGPS (Rosengardten et al., 2011), which are also deficient for skin wound repair.
Overall, growth, development, and maintaining homeostasis place a considerable burden on the mesenchymal stem cell pool and its many lineages. Under a paradigm of biology of priority during development the organism’s structural integrity in the shape of bones and joints, the homeostasis of blood microvessels – regulated by pericytes - should take precedence to fat deposition (Figure 10A).
We propose that these defects in repair in Progeria patients strain the mesenchymal stem cell pools from which repair reactive stroma originates, which have been characterized as both pools of CD44/CD99 positive cells in glandular tissues and as pluripotent cells located at the pericyte position (Crisan et al., 2008; Kim et al., 2014). Interestingly, it has been reported that normal fibroblast cell lines contain a subset of pluripotent MSCs, indistinguishable from those derived from the bone marrow (Denu et al., 2016) and it is established that normal MSCs lose proliferative and differentiation potential within a few passages when cultured in vitro. This phenomenon has been correlated with a defect in lamin A maturation that leads to cellular senescence (Bellotti et al., 2016), which is similar to the terminal differentiation into wound-repair-myofibroblasts resulting from the TGF-beta induction of prostate resident MSCs (Kim et al., 2014). HPGS fibroblasts experience an increased rate in apoptosis, with subsequent cell divisions, concomitant with mutant Lamin A accumulation (Bridger and Kill, 2004), and chronic DNA damage that lead to premature senescence (Wheaton et al., 2017),which could be indirect evidence of early depletion of the MSC pool in these cell lines. (Figure 10B).
Related to later differentiation events, the depletion of microvasculature resident MSCs occurring as a result of chronic, deficient wound repair will carry a severe impact to adipogenesis. Cells at the pericyte position have been characterized as the primary source for adipocytes in vivo (Traktuev et al., 2008; Zannettino et al., 2008). These cells, which are CD44, CD90 double positive, phenocopy the pluripotency of tissue resident mesenchymal stem cells, and are able to differentiate into repair-like myofibroblasts expressing smooth muscle alpha actin (Merfeld-Clauss et al., 2017). Coupled with a preadipocyte depletion phenotype, lipodystrophy in HGPS has also been experimentally induced by the introduction of progerin expression into a subset of pre adipogenic cells in mice, which led to fibrosis, senescence, and macrophage infiltration, with the ultimate result of white fat depletion (Revechon et al., 2017).
The fibroblast as a sentinel of general mesenchymal lineage health
The criticism can be made that it is a stretch to make hypotheses about different mesenchymal stem cell pools based only on gene expression from dermal fibroblasts. However, mounting evidence has shown that after birth, the bone marrow mesenchymal stem cell niche is in close synergy with pools of fibroblasts (LeBleu and Neilson, 2020; Soundararajan and Kannan, 2018), tissue resident mesenchymal stem cells (El Agha et al., 2017) and pericytes (Lamagna and Bergers, 2006). Since these cell pools are in flux, and can compensate for each other under duress (Di Carlo and Peduto, 2018; Direkze et al., 2004; Ozerdem et al., 2005; San Martin et al., 2014), comparing the global transcriptional status of fibroblasts between a patient cohort and its aged-matched controls can provide an insight into potential systemic deficits and compensatory mechanisms at play.
Ideas and speculation
Overall, the cohorts of biological processes that we observe to be misregulated across different age groups of Progeria patients lead us to speculate the impact of this misregulation across a series of events in development, which will warrant further investigation:
Biology of priority: the formation of bone “fires” at specific times, these events are “non-negotiable” and will take precedence over any other concomitant MSC fates. Deficits in endochondral ossification will draw from pericyte and tissue resident MSCs in an attempt to compensate.
The loss of bone fusion and regression of bone length observed in teenaged HGPS patients is an indirect result of deficits in arrest of chondrogenesis proliferation, that occurred in early childhood. These deficits resulted in depletion of MSC pools, both in the bone marrow and in tissues
Depletion of the pericyte niche is exacerbated by a chronic, incompetent, wound repair response. Ultimately, this depletion will result in loss of microvasculature integrity and subsequent vascular events observed in HGPS, such as vascular stiffness and atherosclerosis.
Elaborating on proposed MSC interventions (Infante and Rodriguez, 2021) and current gene editing efforts (Koblan et al., 2021), targeting the bone marrow niche in young patients could rescue the later HGPS phenotypes.
Materials and Methods
RNA-seq
Ten micrograms of purified RNA from several progeria cell lines were acquired from the Coriell Institute (Camden, NJ) as follows: AG06917 (HGPS. Male 3 y/o), AG10578 (HGPS. Male 17 y/o), AG11572 (HGPS. Female 2 y/o), AG10677 (HGPS. Male 4 y/o), AG08466 (HGPS. Female 8 y/o), AG03198 (HGPS. Female 10 y/o), AG07493 (HGPS. Female 2 y/o), AG01178 (HGPS. Male 20 y/o), AG11513 (HGPS. Female 8 y/o, GM01972 (HGPS. Female 14 y/o), GM01178 (HGPS. Male 20 y/o). After internal quality control upon receipt, RNA-seq library construction and sequencing was carried out by Genewiz (South Plainfield, NJ). Further, previously published RNA-seq datasets from HPGS cell lines and normal controls were included in the study (Supplementary Tables 1-3) (Fleischer et al., 2018; Ikegami et al., 2020; Köhler et al., 2020; Mateos et al., 2018).
RNA-seq data processing
The fastq reads were first processed with BBDuk tool (https://github.com/kbaseapps/BBTools), performing adapter trimming with parameters “ktrim=r k=23 mink=11 hdist=1”. Adapter trimmed reads were processed for quality trimming using the BBDuk tool to discard reads with quality score lower than 28 (parameters “qtrim=r trimq=28”). Following the adapter and quality trimming steps, the reads were aligned to the reference genome hg19 using STAR aligner (https://github.com/alexdobin/STAR) with both ‘--outFilterScoreMinOverLread’ and ‘ – outFilterMatchNminOverLread’ parameters set to 0.2. Finally, the mapped reads were sorted based on genomic coordinates and feature count was performed with HTSeq-Counts (https://github.com/simon-anders/htseq).
Batch effect removal
Since the RNA-seq files used in this study were generated in different laboratories, using different technologies, the raw gene counts produced by HTSeq-Counts suffer from batch effect. To mitigate that issue, the raw counts are batch effect adjusted using the ComBat-seq tool (https://github.com/zhangyuqing/ComBat-seq). For this purpose, files from the same laboratory are assigned the same batch number. In addition, the sex and status of the samples are provided as the biological covariates to the ComBat-seq tool to preserve that signal in the adjusted data.
Differential expression analysis
Differential gene expression analysis was performed between different groups of diseased and healthy samples using DESeq2 (https://bioconductor.org/packages/release/bioc/html/DESeq2.html) tool. HGPS samples from Young (0-7 y/o) or teenage (13+ y/o) patients were compared to age matched (AM), middle aged (M), and old (O) control samples (Supplementary Data). Further, to correlate gene expression profiles with known patient survival statistics, comparisons were further refined by stratifying early infancy progeria patients (0-3 years old) and older children (4-7 years old), comparing those populations to age matched, middle aged and old control samples, as before.
Gene ontology analysis
Genes defined as up/down regulated for each of the age group comparisons, were considered for gene ontology analysis based on a p adjusted value cutoff of 0.001. These gene lists, in turn, were then analyzed using Metascape (Zhou et al., 2019). Metascape output files were manually clustered into gene ontology themes, observed throughout the analysis. Gene lists derived from this manual clustering were then visualized via heatmaps to facilitate comparisons among age groups.
Gene expression heatmap plotting
To plot expression of specific genes across different samples as a heatmap, the batch adjusted counts for all the genes for each sample were log2 normalized with a pseudo count of 1. The expression of specific genes across specific samples was then extracted and Z-score normalization was performed for each of the genes. Finally, the Z-score normalized values were plotted using the Seaborn python package as a heatmap. (https://seaborn.pydata.org/)
DamID-seq data processing
All the DamID-seq data was processed as previously reported (Leemans et al., 2019), with modifications in the trimming step. Briefly, the bwa mem (https://github.com/lh3/bwa) tool was used to map gDNA reads starting with GATC to a combination of hg19 reference genome with a ribosomal model. For further processing, only the mapped reads having a mapping quality of at least 10 were considered as GATC fragments. Next, the reads were combined into the bins of 40 kb resolution depending on the middle of the GATC fragments and then scaled to 1M reads. For normalization, log2-ratio of the scaled target over the scaled Dam-only bins was calculated with a pseudo count of 1.
Cell culture
Human primary dermal fibroblast cell lines were obtained from The Progeria Research Foundation (PRF) Cell and Tissue Bank. The HGPS cell lines were HGADFN167 and HGADFN168. Normal dermal fibroblasts AG11513 and AG03257 were purchased from Coriell Institute (Camden, NJ). Cells were grown in DMEM (Gibco) supplemented with 15% FBS (Corning), 1% Pen-strep (Gibco), and 1% L-glutamine (Gibco). Cells were passaged at a density of 80%.
Chromosome conformation capture (Hi-C)
Hi-C experiments were performed according to standard protocols (Golloshi et al., 2018). The starting material was comprised of skin fibroblasts of parents of HGPS patients (Mother AG03257, Father HGADFN168) and two HGPS fibroblast cell lines (HGADFN167 and AG11513). Further, HGADFN168 and HGADFN167, belonging to a parent-child matched cells were analyzed both at an early and late passage: p12-27 and p12-19, respectively.
Briefly, ∼5 million cells were fixed with 1% formaldehyde, suspended in cell lysis buffer for permeabilization, and homogenized by douncing. Crosslinked chromatin was digested overnight with HindIII (HGADFN167 and HGADFN168) or DpnII (HGADFN167, HGADFN168, AG03257 and AG11513). Sticky ends were filled in with biotin-dATP (Invitrogen), and the blunt ends of interacting fragments were ligated together. DNA was purified by two phenol-chloroform extractions and ethanol precipitation. Biotin-dATP at unligated ends was removed, and the DNA was sheared to a target size of 200-400 bp by a Covaris sonicator (Covaris, M220). DNA between 100-400 bp was selected for using AMPure XP beads (Beckman Coulter). Biotinylated DNA was pulled down using streptavidin coated magnetic beads and prepared for multiplex sequencing on an Illumina platform using the NEBNext Ultra II kit (NEB, E7645S). All end preparation, adaptor ligation, and PCR amplification steps were carried out on bead bound DNA libraries. Sequencing was carried out on Illumina HiSeq 3000 or NovaSeq platforms with 75 bp or 150 bp paired end reads. All Hi-C data statistics are presented in Supplementary Table 4.
Analysis of Hi-C data
Sequencing reads were mapped to the reference human genome hg19, filtered, and iteratively corrected using previously published pipelines (Imakaev et al., 2012), available on github (https://github.com/dekkerlab/cMapping). Publicly available tools (https://github.com/dekkerlab/cworld-dekker) were used to produce Hi-C heatmaps at 2.5 Mb and 250 kb resolution and to perform principal component analysis (using the matrix2compartment script) to generate compartment tracks at 250 kb resolution, assigning A and B compartments to positive and negative PC1 values, respectively. Values from replicate experiments were averaged, by bin, to produce the final compartment track. TAD boundary strength analysis was carried out using the insulation score approach (Crane et al., 2015) (matrix2insulation) from this cworld package with 500 kb insulation square using 40 kb resolution contact maps.
ChIP-seq data and processing: Lamin Associated Domain Analysis
Raw LMNA ChIP-seq data was obtained from the NCBI Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE41764 (McCord et al., 2013). For comparison, DamID-seq for the fibroblast line HFFc6 (van Steensel Lab, Netherlands Cancer Institute) was obtained from the 4D Nucleome data portal (Dekker et al., 2017; Reiff et al., 2021), through the bio sample identifier 4DNBSR7TC87A https://data.4dnucleome.org/biosamples/4DNBSR7TC87A/.
The fastq reads were first processed for adapter and quality trimming with the help of BBDuk tool (https://github.com/kbaseapps/BBTools). The reads were then aligned to the hg19 reference genome using STAR aligner (https://github.com/alexdobin/STAR). Finally, both the target and input mapped reads were binned at 40 kb resolution and log2-ratio of the target over the input bins was calculated using the ‘bamCompare’ function of deepTools (https://deeptools.readthedocs.io/) with parameters “--operation log2 -bs 40000 --ignoreDuplicates --minMappingQuality 30 --scaleFactorsMethod SES --effectiveGenomeSize 2864785220”.
Data availability
All RNA-seq and Hi-C data contributed by this study is available on GEO at GSEXXXX.
Author Contributions
The study was conceived and designed by RSM, JTS, and RPM. Experimental data were collected by JTS and AH. Data analysis was performed by RSM, PD, JTS, and RPM. RSM and RPM wrote the manuscript with input from all authors. The project was supervised by RPM.
Competing Interests
The authors declare no competing financial interests.
Movie 1: A visual explanation of ideas presented in the Discussion, hypotheses derived from transcriptomics results.
Supplementary Figures and Tables
Supplementary Figure 1. Batch clustering correction.
A) Principal component analysis (PCA) of the datasets used in this study before batch correction. Samples segregate primarily by lab of origin (red circles). Solid color markers: Progeria, hollow markers: normal controls, triangles: female, circles: male.
B) PCA after batch correction. Samples primarily segregate by progeria/control phenotype. Dotted line denotes an arbitrary margin between cohorts. Solid color markers: Progeria, hollow markers: normal controls, triangles: female, circles: male. Purple circles denote patients that are mismatched from the rest of their respective groups.
Supplementary figure 2. Genes of interest vs LADS
A) The genomic regions for downregulated genes identified in our age comparisons were compiled with the LAD identity found in a father-HGPS child fibroblasts (HGADFN168 and HGADFN167, respectively). Comparisons are presented as follows, from left to right: Young HGPS (0-7 years old) compared to middle-aged controls, young HGPS patients compared to old controls, teenaged HGPS compared to middle-aged controls and teenaged HGPS compared to old controls. Differences between al groups are significant (Kruskal-Wallis p<0.001) Violin plots for the global distribution of values; median denoted by thick dashed line, 25th and 75th percentiles highlighted as thin dashed lines
B) The genomic regions for up regulated genes identified in our age comparisons were compiled with the LAD identity found in a father-HGPS child fibroblasts (HGADFN168 and HGADFN167, respectively). Comparisons are presented as follows, from left to right: Young HGPS (0-7 years old) compared to middle-aged controls, young HGPS patients compared to old controls, teenaged HGPS compared to middle-aged controls and teenaged HGPS compared to old controls. Differences between al groups are significant (Kruskal-Wallis p<0.001) Violin plots for the global distribution of values; median denoted by thick dashed line, 25th and 75th percentiles highlighted as thin dashed lines
C) LAD distribution along chromosome 1 and 21. LAD identity remains consistent among samples, with small regional changes in LAD definition in specific areas (highlighted-rectangle). The greatest difference among samples is the strength in LAD definition, as characterized by smaller positive and negative values
D) The genome-wide distribution of LADs, as characterized by DamID-seq shows that normal fibroblast lines HGADFN168 (belonging to a father of an HGPS patient) and HFFc6 (human foreskin fibroblast), show a characteristic bimodal distribution in LAD intensity around positive (LAD) and negative (non-LAD) values. In contrast, LAD values for the HGPS cell line HGADFN168 distribute normally, around zero.
Supplementary Figure 3
40-kb resolution heatmaps for the parental and HGPS cell lines, around the PTHLH gene (highlight: blue), aligned to their associated compartment and LAD tracks. The gene localizes to the A compartment in both HGPS and control cell lines. A called LAD domain is present in HGPS cells.
Supplementary Table 1. RNA-seq datasets from progeria patients used in this study
Supplementary Table 2. RNA-seq datasets from adult controls used in this study
Supplementary Table 3. RNA-seq datasets from children controls used in this study
Supplementary Table 4. Hi-C Data Statistics
Acknowledgements
The authors would like to thank Enrique Pacheco San Martin for his assistance with graphics and media design. The concept of the biology of priority during development and repair was adapted from discussion with, and coursework designed by, David Rowley Ph.D. (Baylor College of Medicine). This work was supported by NIH NIGMS Grant R35GM133557 to R.P.M. R. San Martin was supported by a postdoctoral fellowship from the American Cancer Society (134060-PF-19-183-01-CSM).