Humans exhibit marked musculoskeletal changes compared to their African ape relatives, such as chimpanzees, bonobos, and gorillas. These changes reflect adaptive shifts during evolution in spine, pelvis, knee, and foot morphology toward bipedalism, shoulder, elbow, and hand morphology for propulsive throwing and object manipulation, and brain size expansion and craniofacial morphology for enhanced cognition related to complex culture and language. Most of these shifts are apparent by birth or arise during early postsnatal life, indicating that developmental genetic circuitry was targeted by selection to shape phenotypes. Studies have suggested that such adaptive traits have complex polygenic underpinnings, with causal variants arising over deep time 1•, 2, 3, 4.

Both the above aspects have made it difficult to connect genotype to phenotype and reveal the underlying mechanisms making humans unique amongst primates 1•, 2, 5 or that contribute to population variation in modern human biology worldwide. Importantly, much of the adaptive condition and its genetic architecture may also hold the key to understanding complex inherited human disease risk in modern contexts (e.g. see Refs. 2, 6, 7). In the past several years, new methodological advances have emerged both in the wet lab and computationally that are setting the stage for scientists to make massive leaps in addressing these gaps. This review focuses on recent developmental, genetic, and comparative primate studies that, through their methodological advances, have sought to identify the complex molecular underpinnings of the derived musculoskeletal trait and disease biology of humans compared to extant and extinct primate relatives.

Human musculoskeletal tissues have been historically understudied at the molecular level because their dense, mineralized nature makes them difficult to access and analyze without damaging cells 8, 9. Consequently, they were excluded from major genomics consortia like encyclopedia of DNA elements [10], Roadmap Epigenomics [11], and dGTEx [12], despite musculoskeletal disorders affecting nearly half the global population [13]. Most postcranial bones form through endochondral ossification, where mesenchymal cells differentiate into chondrocytes that prefigure adult bone structures early in gestation (e.g. 14, 15). These cartilage models then elongate through growth plates and are gradually replaced by bone through processes involving vascular invasion and the differentiation or trans-differentiation of chondrocytes and perichondral cells into osteoblasts 16, 17. More recently, via studies in distinct vertebrate systems, some appendicular skeletal elements may experience another form of trans-differentiation in which perichondral cells (cells lining each chondrocyte model) and fibroblast cells directly become osteoblasts 18, 19, 20. In contrast, certain cranial bones form directly through intramembranous ossification, underscoring the diverse developmental mechanisms underlying skeletal patterning and evolutionary variation 21, 22, 23. These observations across different bone developmental processes now reveal that cartilage and bone cell differentiation processes are plastic and evolvable over time. Moreover, the blueprint that makes one bone develop and take form cannot necessarily be applied directly to another bone, in part due to the complexity of development at each anatomical site, and evolution does not work universally but specifically on anatomy to dictate function.

Both endochondral and intramembranous ossification (Figure 1) are governed by exquisitely choreographed gene-regulatory programs that also underlie growth, homeostasis, and repair of skeletal tissues. Historically, these programs were identified via rare coding mutations in chondrocyte and osteocyte genes that result in syndromic effects, with many skeletal elements being disrupted due to the pleiotropic effects of each altered gene on genetic circuitry (see Ref. [24]). With advances in next-generation sequencing, including its use in functional genomics approaches, the importance of the noncoding genome has taken precedent, with examples of noncoding mutations having massive impacts on skeletogenesis (e.g. SOX9 noncoding mutations underlying campomelic dysplasia) 25, 26. More recently, genome-wide association studies (GWAS) demonstrate that the most common noncoding mutations have limited impacts on skeletal trait variation and disease [27].

Now, advanced technologies have begun to lift the veil on how the noncoding genome impacts gene regulation during unique skeletal element development. High-throughput single-cell (sc) and single-nucleus (sn) RNA-seq, sc-multiomics, high definition (HD) spatial transcriptomics, CRISPR-based perturbation screens, functional screening of identified putative enhancers, and cell culture work systematically link regulatory sequences and genes to phenotypes (Figure 1). These advances are transforming skeletal biology from a sparsely sampled landscape into a dynamic, cell-resolved atlas. With these technological breakthroughs, researchers can now molecularly, cellularly, and histologically document previously inaccessible stages of human skeletogenesis in detail.

Single cell (sc)RNA and scATAC sequencing molecularly segregate tissues into their constituent cells, revealing each cell’s transcriptome and open chromatin landscape, respectively [28]. When these profiles are conducted simultaneously from the same nuclei (i.e., multiomics) and then mapped onto their native tissue coordinates using spatial transcriptomics, lineage trajectories and regulatory events can be tracked directly in situ, rather than inferred from bulk averages or disaggregated cell suspensions (Figure 1). Compared to bulk RNA-seq, these advanced technologies reveal skeletal progenitors and novel cell types, resolve transient chondro-osteogenic intermediates, and pinpoint enhancer-promoter interactions that drive zone-specific as well as skeletal element-specific gene regulation during chondrification and ossification.

By sequencing over 120,000 cells from human hind limbs (Carnegie weeks 5–9) and integrating spatial transcriptomics, Zhang et al. [29] identified 67 transcriptional clusters, mapped early limb patterning, and revealed distinct waves of myogenesis and progenitor diversification, providing the first human-specific timeline for stylopod, zeugopod, and autopod development. Their spatial atlas linked genes such as NOG, PTH1R, DLX5, and TWIST1 to congenital limb malformations. Building on this, To et al. [30] combined scRNA-seq and scATAC-seq across joint regions (E36–E77) to study mesenchyme-to-chondrocyte-to-osteoblast lineages, while doing so identified a novel Schwann cell–derived chondrocyte population. The authors also uncovered regulatory programs shaping synovial and suture joints. They linked regulatory element activity to traits like knee and hip osteoarthritis. While these studies are transformational, they are still coarse-grained and not deep dives into the developmental trajectory of a single tissue, which can have its own unique developmental context and evolutionary path.

When the above studies are considered within an evolutionary framework, as exemplified by recent work using bulk ATAC-seq and RNA-seq on microdissected chondrocytes from numerous postcranial elements by Richard et al. [15], and those on the human and mouse pelvis and knee 2, 3, such studies can transform schematic views of human skeletogenesis into data-rich maps that expose new cell types, progenitor origins, developmental checkpoints, transcription factor (TF) regulatory networks, and disease-relevant regulatory elements, showcasing the power of single-cell and spatial multiomics to close long-standing gaps in skeletal biology with direct ties to evolution. They can also identify signals of adaptive evolution in noncoding regulators of specific skeletal elements, as highlighted in Richard et al. [2], and their focus on how adaptive noncoding regulatory element evolution in human knees shaped modern inherited knee osteoarthritis risk.

Comparative primate sequence analyses complement single-cell and spatial studies by revealing regulatory changes underlying human-specific skeletal traits. Human accelerated regions (HARs) 31, 32, 33, 34, 35 are conserved sequences that evolved rapidly in humans and often act as developmental enhancers; lineage-specific analogs like chimpanzee accelerated regions [36] have also been identified. HARs are enriched in regulatory elements active during limb and pelvic development (e.g. [2,3]) and also shape neurodevelopment by modulating gene expression and 3D chromatin architecture in human vs. chimpanzee cells 36, 37, 38, 39. Additional classes include human-ancestor quickly evolved regions (HAQERs), often poised as enhancers 40, 41, 42•• and human-specific deletions in otherwise conserved elements (hCONDELs), thousands of which alter enhancer activity [43] [44].

Although neurodevelopmental cell types were highlighted in the above studies, a few of these genomic regions, which have been targeted by evolution, also reside near key chondrogenic genes (e.g. SOX9, SOX5, PRG4), making them candidates for skeletal innovation. And approaches like massively parallel reporter assays (MPRAs) can flag functional noncoding variants and enhancers, but because they read out transcriptional activity without definitive targets and often use nonphysiologic cell lines, pairing them with 3D genome contact maps (e.g. Hi-C) and skeleton-wide epigenomic data will more accurately link enhancers to genes and reveal evolved regulators of skeletal growth and human-specific morphology.

A few caveats to studies on HARs or other sequence changes persist. It remains unresolved as to how cell-type-specific versus pleiotropic HAR-containing regulatory sequences act. Moreover, it remains unclear how often HARs (and other sequence changes) reside in skeletal element-specific versus more pan-acting regulatory elements. Another issue concerns how strong or weak HAR-containing regulatory sequences are in driving gene expression, versus performing some other functionalities in biology. It will be equally important to understand how significant or potent HARs versus other classes of sequence changes are in driving regulatory evolutionary events, vis-à-vis gene up- or down-regulation. One such class of changes is single-base pair substitutions between humans and other primates, of which there are millions across the genome, many in highly conserved sequences. These remain untested for functional outcomes even in MPRA.

Concerning single-base-pair substitutions, a proportion of them are now highly polymorphic within humans, and especially between populations. While these have been studied from speciation and population history perspectives (see below), they also have been the subject of GWAS, whereby common single-nucleotide polymorphisms (SNPs) are typed to detect associations/correlations with variation in traits and disease risks. GWAS data, when integrated with multiomic datasets, is rapidly clarifying how genetic variation shapes human skeletal development, disease mechanisms, and even how it reflects archaic interbreeding events effects on phenotypic diversity today. For example, a height meta-GWAS, analyzing 5.4 million individuals, mapped >12,000 independent SNPs, collectively explaining about 40% of height variance in Europeans [45]. This demonstrates how thousands of small-effect variants can saturate polygenic traits and set a framework for pathway-level interpretation of skeletal growth. These studies also illustrate how thousands of variants can collectively account for a highly heritable trait such as height [15]. To et al.’s [30] multiomic atlases of human embryonic bone and joint development also highlight osteoarthritis GWAS signals in specific chondrogenic and osteogenic cell states; this study links early osteo- and chondrogenic programs to later musculoskeletal disease. Extensive work has likewise focused on craniofacial development and GWAS datasets. A notable study by Goovaerts et al. [46] using advanced phenotyping of complex 3D shapes, has enabled more precise analyses than traditional distance-based measures, and, in a joint multi-ancestry/admixed GWAS of 3D cranial vault shape, identified 30 loci whose activity is enriched in cranial neural crest cells and overlaps risk regions for sagittal craniosynostosis, directly linking normal cranial variation to congenital disease pathways. Another study [47] focusing on 4–8 weeks post-conception integrated bulk and single-cell RNA-seq with epigenomics to identify key genes and regulatory networks critical for normal craniofacial development; many are implicated in disorders such as orofacial clefting, offering mechanistic entry points for future functional studies. Finally, GWAS has also been used to probe long-standing evolutionary questions such as the obstetrical dilemma; a recent study by Xu et al. [48] provided empirical evidence supporting and refining this hypothesis by demonstrating the complex genetic architecture and the interplay of selective pressures from childbirth, locomotion, and the human pelvis. They used a deep-learning–assisted GWAS of 31,115 pelvic DXA scans and identified 180 loci affecting seven highly heritable pelvic traits, which revealed sex-specific genetic architecture and trade-offs between childbirth, locomotion, back pain, pelvic floor disorders, and hip osteoarthritis. Collectively, these studies show how integrating GWAS with developmental and evolutionary frameworks is illuminating the genetic architecture of the skeleton and its continuum from normal variation to disease.

Ancient DNA (aDNA), the endogenous genetic material preserved in archeological or historical remains of humans, most often recovered from skeletal tissues, has emerged as a major area of study not just for population history reconstruction but to understand the evolution of human phenotypes. Over the past decade, advances in next-generation sequencing have driven a rapid expansion of aDNA studies, now with tens of thousands of ancient genomes available. However, low coverage and contamination remain key caveats, requiring stringent clean-room protocols and authentication of characteristic damage patterns. Despite these challenges, recent work yielded clear insights into skeletal development and its evolution (e.g. 49, 50, 51). For example, time-series analyses of European aDNA have highlighted selection on genes involved in metabolism, immunity, and pigmentation, revealing a complex interplay of demographic change and adaptation that shaped the genetic and phenotypic landscape of modern Europeans [49]. These studies also clarify how major Holocene cultural shifts (e.g. agriculture) produced specific genetic adaptations with consequences for present-day health and traits. Recent dental GWAS further illustrates this synergy. Mishol et al. [50] introduced a gene-regulatory phenotyping strategy that leverages reconstructed DNA methylation maps to distinguish Denisovan, Neanderthal, and modern human lineages offering a way to flag candidate Denisovan fossils despite scarce diagnostic morphology. The strength of their approach was its link between lineage-specific regulatory shifts and predicted tissue-level phenotypes, complementing traditional skeletal comparisons. Li et al. [51] combined population genetics, multiomics, and developmental assays to identify 18 loci for tooth-crown dimensions, implicating PITX2 and HS3ST3A1 activity in enamel-knot/dental epithelium and demonstrating knockout-driven effects on dental morphology. Notably, the HS3ST3A1 association lies on a Neanderthal-introgressed haplotype, directly linking archaic ancestry to modern variation in tooth size. Taken together, these studies highlight a single takeaway: integration is essential whenever possible, across GWAS, multiomics, aDNA, and beyond.

Advancements in comparative cell-based methodologies have already significantly enhanced our understanding of human skeletal development within an evolutionary framework (Figure 1). Housman et al. [52] introduced a comparative primate skeletal cell culture model utilizing induced pluripotent stem cells (iPSCs) derived from humans and chimpanzees. These iPSCs were differentiated into mesenchymal stem cells and subsequently into osteogenic cells, allowing for a controlled comparison of gene expression during skeletal differentiation. They discovered hundreds of differentially expressed genes between the two species by conducting scRNA-seq across different timepoints, highlighting the regulatory mechanisms that may underlie species-specific skeletal traits. They discovered that human cells advance to late osteogenic states faster than chimpanzee cells (supported by scRNA-seq and lower calcium deposition in chimp cultures), and the differentially expressed genes are enriched for skeletal development and extracellular matrix programs. This model overcomes the previous limitation of having few to no osteogenic cell lines for comparisons, albeit it is still broadly osteogenic, and thus it may or may not capture drivers of localized anatomical differences in the skeleton.

We have also seen the emergence of human and chimpanzee allotetraploid (and autotetraploid) cell models whereby iPSCs of each species have been fused to generate stem cells with both species genomes in the same nucleus 53, 54, 55, 56 (Figure 1). This provides the ability to separate effects in cis- versus trans-regulation in the manner in which cells utilize genomes to orchestrate developmental fates. Using this tetraploid human-chimp cell model, Gokhman et al. [56] focused on cis-regulatory divergence and found evidence that contributed to the distinct facial and vocal tract morphology in humans. In parallel, Song et al. [53] demonstrated that cis-regulatory changes exert larger effect sizes on gene expression differences compared to trans-regulation. They also successfully employed this system to generate species-specific CRISPR-cas9 deletion lines using both allotetraploid and autotetraploid cell lines, providing a powerful platform for functional genetics work. These advances in interspecies cell fusion and creation of novel pluripotent lines have prompted an important reconsideration of their terminology, especially along ethical lines [57]. However, these cell models come with caveats of their own: the artificial doubling of chromosomes introduces potential confounds, and the combination of genomes from different species could disrupt normal regulatory dynamics. Even so, these models offer unprecedented opportunities for in vitro experiments that were previously unfeasible.

Building on these pluripotent platforms, investigators are now using stem-cell lines to construct skeletal organoids and even more complex ‘assembloids’ [58]. A recent study spatially combined two region-specific hPSC-derived cartilage models (interzone/articular chondrocyte progenitors and growth-plate chondrocytes) into a single 3D structure. When cultured in vitro and later transplanted, the assembloid recapitulated endochondral ossification and even exhibited longitudinal growth, mimicking the behavior of embryonic limb elements [58]. This modular design offers a powerful tool for probing into human skeletal patterning and the effects of different species’ genomes on this process (Figure 1). Integrating large-scale single-cell datasets with cell-culture experiments can be prohibitively expensive and time-consuming. The new CellBouncer toolkit [59] helps mitigate these costs by allowing researchers to pool cells from many individuals. This streamlined demultiplexing strategy will be invaluable for future high-throughput human/primate evo-devo studies.

While not in vitro focused, the recent implementation of the developmental Genotype-Tissue Expression (dGTEx) project [12] will complement the above cell-based systems. dGTEx serves to establish a comprehensive resource encompassing gene expression, regulation, and genetic data across developmental stages in humans and nonhuman primates (NHPs) (i.e. macaques and marmosets). By integrating whole-genome sequencing with transcriptomic profiles across 120 human donors (across different developmental stages, 74 tissue types) and ∼200 NHPs (stage-matched 126 macaques and 72 marmosets), dGTEx will provide an unprecedented platform for the scientific community, by providing readily accessible data, to compare developmental gene regulation across species. However, it remains unclear whether hard-to-access skeletal tissues will be included; if so, then findings from dGTEx, when coupled with human in vivo studies of skeletal development 60, 61 and used in conjunction with the cell-based assays described above, will provide massive power to unravel the complexities of skeletal development, disease, and evolution.

Comparative morphological data are essential for fully interpreting genetic findings, particularly within evolutionary developmental biology. High-resolution micro-computed tomography (microCT) scanning provides critical non-destructive visualization of bones, cartilage, and soft tissues, supporting robust comparative phenotyping studies (Figure 1). The NSF-funded openVertebrate (oVert) Thematic Collections Network [62] led a major effort to scan and share thousands of museum specimens via MorphoSource [63], an extensive imaging repository, which now hosts >140,000 CT scans from international researchers, including over 600 CT-scanned specimens spanning 15 primate taxa [64]. While we currently possess comprehensive databases for extant species, developing a similarly expansive, high-quality microCT database for fossil or subfossil specimens is equally essential. Such a resource would significantly enhance our capacity to conduct more rigorous and interdisciplinary deep time studies examining human and primate skeletal evolution by enabling researchers worldwide to correlate recent genetic discoveries directly with observable phenotypic changes (Figure 1). Furthermore, detailed imaging of fossil specimens, including precise visualization of soft tissue proxies (e.g. muscle origin/insertion sites), which can be carefully validated in extant species 65, 66, would significantly advance our understanding of functional morphological adaptations through evolutionary time. Recent studies leveraging microCT data of fossil hominins, such as in Australopithecus and early Homo species, have provided crucial insights into skeletal evolution by revealing transitional morphologies in pelvic and cranial structures 67, 68, 69. By integrating morphological data from extinct and extant primates, researchers can more accurately reconstruct evolutionary trajectories, identify ancestral states, and elucidate adaptive transitions leading to modern human skeletal innovations.

Comprehensively documenting primate embryonic material is equally important; it provides a developmental context for pattern formation rather than analyzing adult morphology alone. Museums worldwide house extensive embryonic specimens, often undocumented and stored in jars or as histological sections, that represent a significant yet largely untapped resource (Figure 1). Ethical constraints significantly limit direct access to living primate embryonic material, especially for hominoids, making historical museum collections especially valuable. Non-destructive imaging methods such as microCT scanning combined with contrast staining solutions like buffered Lugol’s iodine (diceCT) [70] and phosphotungstic acid [71], when permitted, offer ideal approaches to examining these rare specimens without compromising their integrity. Collections like the R. Glenn Northcutt Collection [72], which include thousands of histological sections pertaining to vertebrate neurocranial anatomy, highlight the importance of creating comprehensive archives dedicated to postcranial embryonic anatomy. Utilizing diceCT scanning facilitates high-resolution visualization of soft tissues such as muscles, blood vessels, cartilage, and bone, greatly enriching our comparative morphological analyses. Furthermore, cross-validating microCT imaging with histological studies will serve to link specific gross morphology to underlying tissue structure, setting the stage to understand the tissues/cell types targeted by evolution.

Recent work by Senevirathne et al. [73] demonstrates how a multifaceted approach — combining museum specimens, CT scanning, histology, and single-cell methods — can illuminate the development of a structure central to discussions of bipedalism. The study shows that the human ilium grows in two distinctive steps relative to NHPs: first, the cartilaginous iliac growth plate exhibits an ∼90° reorientation compared with NHP and mouse iliac blades and with long bones in humans, NHPs, and mice; second, ossification initiates at the posterior-most region of the ilium (vs. midshaft in NHPs), internal mineralization is delayed, and mineralization remains largely peripheral (perichondral) until ∼24 gestational weeks. Together, these findings indicated a unique growth-plate orientation and heterotopic (site) and heterochronic (timing) shifts in human ossification patterns. The authors further showed that these human-specific features arise from noncoding regulatory evolution, with implicated elements suggesting potential disease-risk trade-offs.

The advancement of sequencing and imaging technologies over time has facilitated our profound understanding of human evolution. The datasets bridge previously isolated domains of regulatory evolution, gene expression, and phenotypic morphology, effectively aiding in closing the gap between genotype-phenotype relationships (Figure 1). Recently introduced spatial transcriptomic technologies, such as Visium Spatial HD, further enhance these approaches by providing single-cell resolution, a substantial improvement over the 5–10 cell resolution of earlier Visium spot-based methods 74, 75. Integrating regulatory elements containing evolved sequences such as HARs, HAQERs, and hCONDELs with advanced sc-multiomics, MPRAs, cell culture, and imaging techniques represents a promising frontier for evolutionary developmental biology research. Leveraging these cutting-edge methodologies alongside detailed phenotypic analyses across developmental stages and phylogenetic comparisons among primates will enable a robust connection between genetic innovations and anatomical phenotypes (Figure 1). This suggests the need for an initiative dedicated to systematically scanning and cataloging embryonic primate material across global museum collections, as this would significantly advance our understanding of human evolution, and would allow us to map genetic changes onto phenotypic variations and thus morphological adaptations unique to humans or other primates.