The Role of Human-Specific lncRNA in Hyaline Cartilage Development

31 Limb bud like mesenchymal cells・Human-specific lncRNA・Hyaline cartilage development 32

33 Humans possess a markedly different body structure compared to other organisms. The human 34 brain is notably larger than that of other animals, with a particularly well -developed frontal 35 lobe(Hofman, 2014). Humans possess numerous facial muscles, enabling rich expression of 36 emotions(Burrows, 2008). The throat and vocal cords can produce complex sounds, facilitating 37 language-based communication(Nishimura et al., 2022). The hands exhibit opposable thumbs, 38 permitting intricate manipulation (Richmond et al., 2016) . Humans have less body hair and 39 more developed sweat glands (Kamberov et al., 2018) . Additionally, humans are the only 40 species that practises upright bipedalism. 41 In evolutionary biology, several hypotheses have been proposed to explain the attainment of 42 upright bipedalism. These include the Savannah Hypothesis, the Carrying Hypothesis, and the 43 Thermoregulatory Hypothesis (Harcourt-Smith, 2010) . The Savannah Hypothesis attributes 44 bipedalism to environmental changes. Approximately six million years ago, the African forests 45 contracted, giving rise to expansive savannahs. Consequently, human ancestors needed to 46 traverse open grasslands, and it is believed that upright bipedalism allowed them to efficiently 47 survey large areas. The Carrying Hypothesis suggests that the freeing of the hands enabled the 48 transport of food, tools, and offspring, which was advantageous for survival and reproduction. 49 The use of tools is considered a factor in human evolution, contributing to the development of 50 intelligence. The Thermoregulatory Hypothesis posits that bipedalism was an adaptation for 51 survival in hot environments. By standing upright, the body reduced the surface area exposed 52 to direct solar radiation and heat reflected from the ground. Moreover, the increased surface 53 area exposed to wind facilitated effective heat dissipation. 54 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 3 Despite the various hypotheses proposed in evolutionary biology, the molecular biological 55 mechanisms underlying human bipedalism remain elusive. The evolution of bipedalism in 56 humans is closely related to changes in the structure of the lower limbs. It is well known that 57 the lower limbs of humans are significantly longer compared to those of the Pan genus, and 58 there are notable differences in the angle of the pelvis (O’Neill et al., 2015) . Therefore, 59 elucidating the human -specific mechanisms during the developmental process when the 60 skeleton is formed could lead to the formulation of hypotheses regarding the molecular 61 biological mechanisms of evolution. The skeleton is formed through endochondral ossification, 62 where cartilage is replaced by bone, and secondary ossification occurs at the epiphyses (Long 63 and Ornitz, 2013). Hence, it is essential to focus on human-specific cartilage formation during 64 development. Previously, a method was established for differentiating human iPS cells into 65 expandable limb bud like mesenchymal cells (ExpLBM) via lateral plate mesoderm, and 66 subsequently into hyaline cartilage-like tissue (HCT)(Yamada et al., 2021). Since this method 67 mimics the developmental process, it was considered that comparing RNA expression between 68 ExpLBM and HCT could reveal RNAs involved in human-specific cartilage formation. 69 However, since mRNA exhibits a high degree of sequence conservation across species, it might 70 be challenging to elucidate the molecular biological mechanisms using mRNA alone. On the 71 other hand, long non -coding RNAs (lncRNAs), which are RNA molecules longer than 200 72 nucleotides that do not encode proteins, show less sequence conservation compared to 73 mRNA(Makałowski et al., 1996; Noviello et al., 2018) . Moreover, lncRNAs are known to 74 exhibit more tissue -specific expression patterns than mRNA, suggesting significant roles in 75 cell type-specific processes(Statello et al., 2021) . Unlike the approximately 20,000 types of 76 mRNA, the number of lncRNAs reaches hundreds of thousands, and due to alternative splicing, 77 the number of isoforms increases dramatically (“RNAcentral: a comprehensive database of 78 non-coding RNA sequences,” 2017) . This complexity makes functional 79 evaluation challenging, and consequently, lncRNA research is not as advanced. lncRNAs play 80 crucial roles in various cellular processes, such as the cell cycle, differentiation, and 81 metabolism. Their mechanisms involve interactions with DNA, RNA, and proteins, leading to 82 the regulation of epigenetic modifications, transcription, translation, post -translational 83 modifications, and the stability of proteins and RNAs (Herman et al., 2022) . Particularly in 84 transcriptional regulation, lncRNAs are known to cis -acting by regulating the expression of 85 genes located near their own genetic loci, or by binding to promoter and enhancer regions of 86 target genes to control transcription (Gil and Ulitsky, 2020) . This regulation occurs by either 87 inhibiting or promoting the binding of transcription factors and the transcription initiation 88 complex. 89 LncRNAs are known to interact with various biomolecules; consequently, numerous tools have 90 been developed to predict their functions. For predicting interactions with DNA, tools such as 91 Triplex Domain Finder (TDF) and PATO have been developed (Amatria-Barral et al., 2023; 92 Kuo et al., 2019) . TDF utilises an efficient bit-parallel algorithm and a concept akin to motif 93 overrepresentation analysis to rapidly predict regions where lncRNAs form triple helices with 94 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 4 DNA. For predicting interactions with proteins and RNA, tools such as linc2function and 95 EDLMFC have been developed (Ramakrishnaiah et al., 2023; Wang et al., 2021) . 96 Linc2function is a tool that extracts features through machine learning from nucleotide 97 sequences and predicts functional characteristics via interactome and secondary structure 98 analysis. It can be executed locally or through a web server. 99 By utilising the characteristics of lncRNAs and these tools, it is hypothesised that the 100 relationship between human -specific lncRNAs and the evolution of bipedalism, which is a 101 significant difference between humans and other primates such as Pan, can be elucidated. 102 Therefore, in this study, an attempt was made to evaluate the functions of cartilage -related 103 lncRNAs from an evolutionary perspective by predicting the roles of human-specific lncRNAs 104 that are significantly up-regulated in chondrocytes. 105

106 Cell culture and Differentiation 107 The methods employed were based on those outlined in our previous publication(Yamada et 108 al., 2021). Briefly, human induced pluripotent stem cells (414C2 PRRX1-tdTomato; 3 × 104 109 cells) were suspended in 1 ml of StemFit (Ajinomoto) containing 10 μM Y27632 (Wako) and 110 added to a 3.5 cm culture dish containing 4 μl iMatrix511. The culture medium was replaced 111 the next day with fresh StemFit without Y27632. After culturing for 2 d, the cells were 112 washed with PBS and differentiation was induced by changing the culture medium at each 113 time point. Accutase (Thermo Fisher) was used to dissociate LBM-like cells, and 2 × 105 114 cells were suspended in ExpLBM medium containing 16 μl iMatrix511 and then added to a 6 115 cm culture dish. The culture media were replaced with fresh ExpLBM medium every 2 d. To 116 induce chondrogenic induction using pellet culture conditions (3DCI), 1 × 105 ExpLBM cells 117 were suspended in 200 μl chondrogenic culture step 1 medium and added to 96-well ultralow 118 U-bottom plates (Corning). The cells were pelleted by centrifugation at 2,000 r.p.m. for 5 119 min. At the end of steps 1 and 2, the culture media were changed to chondrogenic culture step 120 2 and step 3 media, respectively. 121 122 Bulk RNA-Seq 123 Total RNA was extracted using an RNeasy kit (Qiagen), and sequencing libraries were 124 prepared using a KAPA RNA HyperPrep Kit with RiboErase (HMR) (Kapa Biosystems, USA) 125 and a SeqCap Adapter Kit (Set A or Set B, Roche, USA) according to the manufacturer’s 126 instructions. Sequencing libraries were transferred to AZENTA (Suzhou, China) and were 127 loaded onto a HiSeq 2500 system (Illumina, USA) for sequencing. All sequence reads were 128 extracted in FASTQ format using the CASAVA 1.8.4 pipeline. Fastp (version 0.23.4) was used 129 to remove adapters and filter raw reads of < 60 bases in addition to leading and trailing bases 130 with a read quality of less than 30. For the pseudoalignment of the filtered reads to the 131 GRCh38.p14, Kallisto (version 0.48.0) was used. GeTMM normalization was performed using 132 edgeR (version 4.0.3) to account for sample variation. Differentially expressed genes were 133 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 5 identified through NOISeq (version 2.46.0) analysis. The raw RNA-seq data were deposited in 134 the NCBI GEO database under accession number GSE165620, . 135 Databases 136 RNACentral(“RNAcentral: a comprehensive database of non -coding 137 RNA sequences,” 2017) was utilised for the identification of lncRNAs. For the 138 identification of human-specific lncRNAs, LncBook2.0(Li et al., 2023) was employed. 139 Gene Ontology Analysis 140 Gene Ontology (GO) analysis was performed using the clusterProfiler package (v4.10.1) in 141 the R (v4.3.3) environment against the org.Hs.eg.db database (v3.18.0). The parameters used 142 for the analysis were ont="ALL", pAdjustMethod="BH", pvalueCutoff=0.05, and 143 qvalueCutoff=0.0. Visualisation was conducted using the ggplot2 package (v3.5.0). 144 lncRNA Function Prediction 145 To predict the proteins binding to lncRNAs, the linc2function web server (Ramakrishnaiah et 146 al., 2023) was utilised with parameters set to ANN, HumanSpecific, and Full Model. For 147 peptide function prediction, DeepGOWeb (v1.0.18)(Kulmanov et al., 2021) was used, with the 148 prediction threshold set to the default value of 0.3. Triplex prediction was conducted using the 149 TDF(Kuo et al., 2019) included in the Regulatory Genomics Toolbox (v1.0.2), executed in 150 promotertest mode against the hg38 genome assembly. All parameters were set to their default 151 values. 152

153 lncRNA Profiles During Chondrogenic Differentiation 154 Bulk RNA -Seq was performed on ExpLBM and HCT (Fig. 1A). From the obtained RNA 155 expression profiles, non-coding RNAs were extracted, and MA plots were generated (Fig. 1B). 156 There were 1,346 HCT -enriched lncRNAs (log2FC ≥ 1, prob > 0.8) and 1,082 ExpLBM -157 enriched lncRNAs (log2FC ≤ -1, prob > 0.8). Known cartilage -related lncRNAs were 158 significantly highly expressed in HCT (Table 1). There were no significant differences in the 159 expression of other lncRNAs reported in the reference literature. Although the referenced paper 160 also reports on microRNAs, small RNA-Seq was not conducted; hence, the quantification was 161 based on the expression levels of pri-miRNA and miRNA host gene. To extract human-specific 162 lncRNAs from the HCT -enriched lncRNAs, the LncBook2.0 database, which includes 163 conservation information, was used (Fig. 1C). As a result, 740 lncRNAs were identified (Fig. 164 1D, Supp. Fig. 1). Interestingly, the greatest number of conserved lncRNAs was found among 165 the Eutheria, while those conserved for a longer period, excluding those conserved among the 166 Euteleostomi, were present in similar quantities. Additionally, the lncRNAs conserved for a 167 longer period were predominantly known to encode peptides , as identified by Ribo -Seq, 168 whereas the more recently evolved lncRNAs post -Hominoidea scarcely encoded peptides. 169 Furthermore, despite the relatively lower quantity of lncRNAs conserved between Primates 170 and Hominini, a higher number of Homo-specific lncRNAs were observed in comparison. 171 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 6 172 173 Functional Prediction of Human-Specific lncRNAs 174 Among the differentially expressed lncRNAs (DELs) between ExpLBM and HCT, 36 human-175 specific lncRNAs (DEL_HCT_hs) were identified as having significantly increased expression 176 levels in HCT. Among DEL_HCT_hs, HSALNG0113101 and HSALNG0142302 are known 177 to be translated into peptides, capable of translating into 5 and 10 types of peptides, respectively 178 (Suppl. Fig. 2A, 2B). Functional predictions for these 15 peptides were conducted using 179 DeepGOWeb (Suppl. Fig. 2C). Only two peptides derived from HSALNG0142302 could have 180 their functions predicted. Both were associated with the same Gene Ontology Molecular 181 Function (GOMF) and Biological Process (GOBP), with GOMF related to binding functions 182 and GOBP related to responses to stress and stimuli. 183 Additionally, the prediction of proteins binding to DEL_HCT_hs was conducted using 184 linc2function for all proteins. As a result, it was predicted that 28 proteins would bind to 14 185 DEL_HCT_hs (Fig. 2A). EIF4B, FUS, MBNL1, and SFR1 were predicted to have binding 186 potential with all the identified lncRNAs, whereas some proteins, such as A2BP1, were 187 predicted to bind only to specific lncRNAs (Fig. 2B). Gene Ontology analysis of the 28 proteins 188 predicted to bind to the 14 DEL_HCT_hs indicated that they are involved in RNA splicing and 189 mRNA metabolic processes (Fig. 2C). Among the top 10 terms by GeneRatio, six functions 190 involved in interactions with mRNA were identified: RNA splicing, regulation of mRNA 191 metabolic process, RNA transport, establishment of RNA localization, mRNA stabilization, 192 and positive regulation of translation. 193 Table 1 known cartilage-associated lncRNAs lncRNA ref ROCR (Zhu et al., 2019) TUG1 GAS5 SNHG5 (Razmara et al., 2019) MEG3 miR-140 miR-483 miR-218 MIR503HG .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 7 Prediction of Gene Expression Regulation by Human-Specific lncRNA 194 To investigate whether DEL_HCT_hs could regulate the expression of the 2,841 mRNAs 195 (DEG_HCT), which were significantly upregulated in HCT compared to ExpLBM, the 196 interactions between the promoter regions of DEG_HCT and DEL_HCT_hs were predicted 197 using TDF (Fig. 3A). The results indicated that 9 DEL_HCT_hs could form triplex structures 198 with the promoter regions of DEG_HCT (Table 2). Among these, HSALNG0055279 was 199 found near SMOC2, and HSALNG0123261 was located near the transcription factor NFIC. 200 Using the NFIC GeneSet from Harmonizome 3.0, target genes of NFIC were identified, 201 including the chondrocyte marker genes COL2A1 and SOX9, as well as the joint marker genes 202 BARX1, COL14A1, EPS8, GEM, and NFIA. Analysis of the number of target genes for each 203 lncRNA revealed that lnc-GOLGA6L4-2_4-HSALNG0107783_2 had the most targets, while 204 HSALNG0123261 had the fewest (Fig. 3B). It should be noted that lncRNA names connected 205 by a hyphen, such as lnc -GOLGA6L4-2_4-HSALNG0107783_2, represent different names 206 from various databases, with the underscore being used to distinguish variants. Additionally, 207 HSALNG0055648 had the most total triple helix-forming regions, and HSALNG0061970 had 208 the fewest. Examination of genes with many triple helix -forming regions with DEL_HCT_hs 209 revealed a high number of such regions between HSALNG0151253 and COL6A1 (Suppl. Fig. 210 3A). Focusing on chondrocyte marker genes and joint marker genes, it was predicted that 211 ACAN, COL2A1, COL9A1, COL11A2, COMP, and SOX5 among the chondrocyte marker 212 genes, and ABI3BP, BARX1, CHI3L1, EPS8, GEM, NFIA, THBS4, and TMEM30B among 213 the joint marker genes, would form triplexes. While some genes, such as COL2A1, BARX1, 214 EPS8, and THBS4, had very few triplex formation regions compared to other genes, there were 215 also genes like CHI3L1 and TMEM30B that were predicted to form triplexes with all nine 216 types (Fig. 3C, 3D). The DEL_HCT_hs with the highest number of triplex formation regions 217 with chondrocyte marker genes was lnc -GOLGA6L4-2, whereas the DEL_HCT_hs with the 218 highest number of triplex formation regions with joint marker genes were HSALNG0055648 219 and HSALNG0108834. Among the human-specific proteins within DEG_HCT, it was shown 220 that GSTT2B, ZFP36L1, and ANGPTL5 could form triple helices with DEL_HCT_hs (Suppl. 221 Fig. 3 B). Among them, ANGPTL5 exhibited the highest number of total triplex formation 222 regions with HSALNG0055648 and HSALNG0108834. Information gathered from 223 GeneCards indicated that GSTT2B catalyses the conjugation of reduced glutathione to various 224 electrophilic and hydrophobic compounds, ZFP36L1 functions as a transcription factor, and 225 ANGPTL5 is predicted to act in the extracellular matrix and extracellular space containing 226 collagen. Despite attempts to identify the target genes of ZFP36L1 using JASPAR, ZFP36L1 227 was not registered in the JASPAR database. 228 Given the potential of DEL_HCT_hs to form triple helices in the promoter regions of genes 229 associated with cartilage, joints, and ECM, GO analysis was performed on DEG_HCT, the 230 1,229 genes predicted to form triple helices with DEL_HCT_hs in their promoter regions, and 231 the remaining 1,612 genes not predicted to form triple helices with any DEL_HCT_hs (Fig. 232 3F). In DEG_HCT, ontologies related to ECM and bone/cartilage were enriched (Fig. 3G). To 233 investigate the characteristics of the genes targeted by DEL_HCT_hs, a comparison was made 234 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 8 between the genes predicted to form triple helices with DEL_HCT_hs in their promoter regions 235 and the genes predicted not to form triple helices with any DEL_HCT_hs (Fig. 3F, Suppl. Table 236 1). This comparison was conducted under the conditions of an adjusted p -value < 0.01, Fold 237 Enrichment ≥ 0.001, and |log2(Fold Change of Fold Enrichment)| ≥ 1. Consequently, it was 238 revealed that ECM -related genes and cell growth -related genes were included among the 239 DEL_HCT_hs target genes, whereas vesicle -related genes and genes associated with 240 topologically incorrect proteins were included among the genes not targeted by any 241 DEL_HCT_hs(Fig. 3H, Suppl. Table 2, 3) . Among the top five GOBP terms for 242 FoldEnrichment of DEL_HCT_hs target genes, the terms related to ECM contained genes that 243 were not targeted by DEL_HCT_hs. However, the log2(Fold Change of FoldEnrichment) was 244 1.36, indicating an enrichment towards DEL_HCT_hs target genes. The remaining terms, 245 which were significantly associated with cell growth and positive regulation of cell adhesion, 246 did not include any genes among the non-targets of DEL_HCT_hs. Conversely, for the top five 247 GOBP terms with the highest FoldEnrichment among non-target genes of DEL_HCT_hs, none 248 of the significantly associated genes were found within the DEL_HCT_hs target genes. 249

250 Understanding the function of lncRNAs is crucial for a deeper comprehension of biological 251 phenomena. LncRNAs have garnered significant attention in recent years due to their 252 involvement in nearly all aspects of the central dogma, from the epigenome to protein 253 stability(Herman et al., 2022). This involvement suggests their contribution to the evolution of 254 life. However, previous research has primarily focused on their potential as biomarkers for 255 diseases and drug targets (Nemeth et al., 2024) , leaving their role in evolution largely 256 unexplored. One notable difference between humans and closely related species, such as the 257 Pan genus, is the presence of bipedalism. The relationship between lncRNAs and bipedalism 258 remains unclear. Bipedalism is a result of structural changes in the lower limbs, with the 259 skeleton being formed from cartilage during development. Given the skeletal differences in the 260 lower limbs between humans and Pan, it is hypothesised that the cellular arrangement of 261 cartilage in humans significantly diverges from that of other organisms during developmental 262 stages. Although mature chondrocytes may not necessarily retain memories of their 263 developmental phase, focusing on the distinctive RNA expression profiles of chondrocytes 264 during development may elucidate genes involved in bipedal locomotion. Previously, we 265 developed a method to differentiate iPS cells into chondrocytes by mimicking developmental 266 stages(Yamada et al., 2021) . However, this method is inherently artificial, making it 267 challenging to discern whether the observed differences from mature chondrocytes are specific 268 to the developmental stage or artifacts of the artificial process. Furthermore, the mechanisms 269 of limb development are widely conserved, and the specific contributions of various genes at 270 different stages are well -documented(Ornitz and Marie, 2015) . In contrast, lncRNAs exhibit 271 characteristics such as lower conservation compared to mRNAs, involvement in nearly all 272 aspects of the central dogma, and in some cases, peptide coding. Therefore, by focusing on the 273 functions of human-specific lncRNAs, it is possible to elucidate contributions to chondrocyte 274 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 9 differentiation and, consequently, to bipedal locomotion from a perspective distinct from the 275 conserved mechanisms of limb development. This study aims to investigate the expression of 276 lncRNAs in cartilage during human development by comparing ExpLBM, as reported 277 previously, with HCT differentiated from it, to infer the function of human-specific lncRNAs. 278 Initially, the functions of peptides encoded by DEL_HCT_hs were predicted, revealing that 279 only 2 out of 15 peptides could be estimated, both related to responses to stimuli and stress 280 (Suppl. Fig. 2). Furthermore, predictions of proteins that could bind to DEL_HCT_hs, followed 281 by GO analysis of these proteins, indicated that they possess RNA splicing functions (Fig. 2C). 282 Additionally, they were found to possess functions related to interactions with mRNAs. This 283 study did not employ RNA -RNA interaction analyses such as hiCLIP or CLASH -Seq; thus, 284 the specific genes targeted by human -specific lncRNAs for RNA splicing remain unknown. 285 However, the highly ranked GO biological process terms such as RNA splicing, regulation of 286 mRNA metabolic process, RNA transport, establishment of RNA localization, mRNA 287 stabilization, and positive regulation of translation contribute to the structure and expression 288 levels of proteins. Particularly, if human-specific splicing is regulated by DEL_HCT_hs, slight 289 variations in protein function may be involved in bipedal locomotion. Investigating RNA 290 splicing regulated by DEL_HCT_hs remains a task for future research. 291 Next, to investigate the potential regulatory role of DEL_HCT_hs on gene expression, the 292 ability to form triple helices with DEG_HCT promoter regions was predicted (Fig. 3). The 293

demonstrated that promoters of genes related to cartilage formed triple helices (Fig. 3C, 294 D), and that target genes of DEL_HCT_hs were enriched in ECM-related genes (Fig. 3H). The 295 potential for triplex formation with human -specific genes among DEG_HCT was also 296 investigated. It was demonstrated that triplexes could form with GSTT2B, ZFP36L1, and 297 ANGPTL5, with a particular propensity for formation in the promoter region of ANGPTL5, 298 which is predicted to be active in the extracellular matrix and extracellular space, including 299 collagen. It is known that lncRNAs bind to nearby gene promoter regions to regulate 300 expression(Dhaka et al., 2024), and among the lncRNAs predicted to form triple helices with 301 DEG promoter regions, HSALNG0055279 and HSALNG0123261 were near SMOC2 and 302 NFIC, respectively (Table 2). In mice, knockdown of SMOC1 along with SMOC2 results in 303 impaired bone formation in the cranium, limbs, and mandible (Takahata et al., 2021). NFIC, a 304 transcription factor, targets chondrocyte and joint marker genes. The mechanical stimuli 305 transmitted by the ECM are relayed into the cell by various stimulus and stress response 306 proteins, and it is intriguing that peptides derived from DEL_HCT_hs were predicted to 307 respond to such stimuli and stress (Suppl. Fig. 2C). These findings suggest that human-specific 308 lncRNAs may influence the composition of the ECM. 309 Another significant difference in skeletal structure between humans and other animals is the 310 angle of the femur relative to the body axis. In chimpanzees and gorillas, close relatives of 311 humans, this angle is 75°, whereas in humans it is 150° (Kozma et al., 2018) , increasing the 312 load on the joints. It has been revealed that Japanese macaques, known for their performance 313 in "saru-mawashi" and trained to walk bipedally, adapt to the stresses of bipedal locomotion 314 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 10 by increasing the strength of the trabecular bone in the femoral head and enhancing ligament 315 attachments compared to their wild counterparts (Cazenave et al., 2024) . Given that human -316 specific lncRNAs regulate the expression of ECM -related genes, it is suggested that these 317 lncRNAs may have evolved to control ECM composition in response to the increased joint 318 load due to bipedalism. This study focused only on the functionally estimated DEL_HCT_hs, 319 but cellular and animal experiments on other lncRNAs might reveal those that have adapted to 320 and contribute to bipedalism. Though this study primarily involved functional predictions using 321 tools, empirical verification using cells and animals is necessary. Nevertheless, this research 322 suggests that investigating lncRNAs from an evolutionary perspective could elucidate their role 323 as drivers of cellular processes associated with evolution and adaptation. 324

325 This study suggests that human -specific lncRNAs expressed during chondrogenesis may be 326 associated with the ECM, potentially as an adaptation resulting from bipedalism. These 327 findings have implications not only for evolutionary biology but also for therapeutic 328 applications. Specifically, regulating human -specific lncRNAs could improve the quality of 329 the ECM, potentially leading to the development of regenerative cartilage tissues that more 330 closely resemble normal human joint cartilage. Furthermore, elucidating the mechanisms of 331 evolution-related lncRNAs could contribute to understanding the pathogenesis of conditions 332 such as osteoarthritis and neural tube defects, which are more prevalent in humans compared 333 to other species. 334 Supplementary Materials 335 Figure S1: Phylogenetic tree of biological classification 336 Figure S2: DEL_HCT_hs-Protein Interactions 337 Figure S3: Triplex Predictions of DEL_HCT_hs 338 Table S1: Fold enrichment for the GO analysis conducted on DEG_HCT, predicted to be 339 targeted by DEL_HCT_hs in the promoter region, and other DEG_HCT. 340 Table S2: GO analysis results of DEG_HCT, predicted to be targeted by DEL_HCT_hs in the 341 promoter region 342 Table S3: GO analysis results of DEG_HCT, predicted not to be targeted by DEL_HCT_hs 343 in the promoter region 344 Acknowledgments 345 Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science 346 (23K14384 to T. Osone; 23K08677 to T. Takao; 23K21368 to T. Takarada) and JST FOREST 347 Program (JPMJFR225H to T. Takarada). These funders had no role in the study design, data 348 collection and analysis, decision to publish, or preparation of the manuscript. 349 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 11 The computation was carried out using the General Projects on supercomputer "Flow" at 350 Information Technology Center, Nagoya University. 351 Conflicts of Interest 352 The authors declare no conflict of interest. 353

354 Amatria-Barral, I., González-Domínguez, J., Touriño, J., 2023. PATO: 355 genome-wide prediction of lncRNA–DNA triple helices. Bioinformatics 356 39, btad134. https://doi.org/10.1093/bioinformatics/btad134 357 Burrows, A.M., 2008. The facial expression musculature in primates and its 358 evolutionary significance. BioEssays 30, 212–225. 359 https://doi.org/https://doi.org/10.1002/bies.20719 360 Cazenave, M., Nakatsukasa, M., Mazurier, A., M. Skinner, M., 2024. 361 Identification of functionally related adaptations in the trabecular 362 network of the proximal femur and tibia of a bipedally trained Japanese 363 macaque. Anthropological Science 132, 13–26. 364 https://doi.org/10.1537/ase.2307142 365 Dhaka, B., Zimmerli, M., Hanhart, D., Moser, M.B., Guillen-Ramirez, H., 366 Mishra, S., Esposito, R., Polidori, T., Widmer, M., García-Pérez, R., 367 Julio, M.K., Pervouchine, D., Melé, M., Chouvardas, P., Johnson, R., 368 2024. Functional identification of cis-regulatory long noncoding RNAs 369 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 12 at controlled false discovery rates. Nucleic Acids Res 52, 2821–2835. 370 https://doi.org/10.1093/nar/gkae075 371 Gil, N., Ulitsky, I., 2020. Regulation of gene expression by cis-acting 372 long non-coding RNAs. Nat Rev Genet 21, 102–117. 373 https://doi.org/10.1038/s41576-019-0184-5 374 Harcourt-Smith, W.H.E., 2010. The First Hominins and the Origins of 375 Bipedalism. Evolution: Education and Outreach 3, 333–340. 376 https://doi.org/10.1007/s12052-010-0257-6 377 Herman, A.B., Tsitsipatis, D., Gorospe, M., 2022. Integrated lncRNA 378 function upon genomic and epigenomic regulation. Mol Cell 82, 2252–379 2266. https://doi.org/https://doi.org/10.1016/j.molcel.2022.05.027 380 Hofman, M., 2014. Evolution of the human brain: when bigger is better. 381 Front Neuroanat 8. https://doi.org/10.3389/fnana.2014.00015 382 Kamberov, Y.G., Guhan, S.M., DeMarchis, A., Jiang, J., Wright, S.S., 383 Morgan, B.A., Sabeti, P.C., Tabin, C.J., Lieberman, D.E., 2018. 384 Comparative evidence for the independent evolution of hair and sweat 385 gland traits in primates. J Hum Evol 125, 99–105. 386 https://doi.org/https://doi.org/10.1016/j.jhevol.2018.10.008 387 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 13 Kozma, E.E., Webb, N.M., Harcourt-Smith, W.E.H., Raichlen, D.A., D’Août, 388 K., Brown, M.H., Finestone, E.M., Ross, S.R., Aerts, P., Pontzer, H., 389 2018. Hip extensor mechanics and the evolution of walking and climbing 390 capabilities in humans, apes, and fossil hominins. Proceedings of the 391 National Academy of Sciences 115, 4134–4139. 392 https://doi.org/10.1073/pnas.1715120115 393 Kulmanov, M., Zhapa-Camacho, F., Hoehndorf, R., 2021. DeepGOWeb: fast and 394 accurate protein function prediction on the (Semantic) Web. Nucleic 395 Acids Res 49, W140–W146. https://doi.org/10.1093/nar/gkab373 396 Kuo, C.-C., Hänzelmann, S., Sentürk Cetin, N., Frank, S., Zajzon, B., 397 Derks, J.-P., Akhade, V.S., Ahuja, G., Kanduri, C., Grummt, I., Kurian, 398 L., Costa, I.G., 2019. Detection of RNA–DNA binding sites in long 399 noncoding RNAs. Nucleic Acids Res 47, e32–e32. 400 https://doi.org/10.1093/nar/gkz037 401 Li, Z., Liu, L., Feng, C., Qin, Y., Xiao, J., Zhang, Z., Ma, L., 2023. 402 LncBook 2.0: integrating human long non-coding RNAs with multi-omics 403 annotations. Nucleic Acids Res 51, D186–D191. 404 https://doi.org/10.1093/nar/gkac999 405 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 14 Long, F., Ornitz, D.M., 2013. Development of the Endochondral Skeleton. 406 Cold Spring Harb Perspect Biol 5, a008334. 407 https://doi.org/10.1101/cshperspect.a008334 408 Makałowski, W., Zhang, J., Boguski, M.S., 1996. Comparative analysis of 409 1196 orthologous mouse and human full-length mRNA and protein 410 sequences. Genome Res 6, 846–857. https://doi.org/10.1101/gr.6.9.846 411 Nemeth, K., Bayraktar, R., Ferracin, M., Calin, G.A., 2024. Non-coding RNAs 412 in disease: from mechanisms to therapeutics. Nat Rev Genet 25, 211–413 232. https://doi.org/10.1038/s41576-023-00662-1 414 Nishimura, T., Tokuda, I.T., Miyachi, S., Dunn, J.C., Herbst, C.T., 415 Ishimura, K., Kaneko, A., Kinoshita, Y., Koda, H., Saers, J.P.P., Imai, 416 H., Matsuda, T., Larsen, O.N., Jürgens, U., Hirabayashi, H., Kojima, 417 S., Fitch, W.T., 2022. Evolutionary loss of complexity in human vocal 418 anatomy as an adaptation for speech. Science (1979) 377, 760–763. 419 https://doi.org/10.1126/science.abm1574 420 Noviello, T.M.R., Di Liddo, A., Ventola, G.M., Spagnuolo, A., D’Aniello, 421 S., Ceccarelli, M., Cerulo, L., 2018. Detection of long non–coding RNA 422 homology, a comparative study on alignment and alignment–free metrics. 423 BMC Bioinformatics 19, 407. https://doi.org/10.1186/s12859-018-2441-6 424 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 15 O’Neill, M.C., Lee, L.-F., Demes, B., Thompson, N.E., Larson, S.G., Stern, 425 J.T., Umberger, B.R., 2015. Three-dimensional kinematics of the pelvis 426 and hind limbs in chimpanzee (Pan troglodytes) and human bipedal 427 walking. J Hum Evol 86, 32–42. 428 https://doi.org/https://doi.org/10.1016/j.jhevol.2015.05.012 429 Ornitz, D.M., Marie, P.J., 2015. Fibroblast growth factor signaling in 430 skeletal development and disease. Genes Dev 29, 1463–1486. 431 https://doi.org/10.1101/gad.266551.115 432 Ramakrishnaiah, Y., Morris, A.P., Dhaliwal, J., Philip, M., Kuhlmann, L., 433 Tyagi, S., 2023. Linc2function: A Comprehensive Pipeline and Webserver 434 for Long Non-Coding RNA (lncRNA) Identification and Functional 435 Predictions Using Deep Learning Approaches. Epigenomes 7. 436 https://doi.org/10.3390/epigenomes7030022 437 Razmara, E., Bitaraf, A., Yousefi, H., Nguyen, T.H., Garshasbi, M., Cho, 438 W.C., Babashah, S., 2019. Non-Coding RNAs in Cartilage Development: An 439 Updated Review. Int J Mol Sci 20. https://doi.org/10.3390/ijms20184475 440 Richmond, B.G., Roach, N.T., Ostrofsky, K.R., 2016. Evolution of the Early 441 Hominin Hand, in: Kivell, T.L., Lemelin, P., Richmond, B.G., Schmitt, 442 D. (Eds.), The Evolution of the Primate Hand: Anatomical, 443 Developmental, Functional, and Paleontological Evidence. Springer New 444 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 16 York, New York, NY, pp. 515–543. https://doi.org/10.1007/978-1-4939-445 3646-5_18 446 RNAcentral: a comprehensive database of non-coding RNA sequences, 447 2017. . Nucleic Acids Res 45, D128–D134. 448 https://doi.org/10.1093/nar/gkw1008 449 Statello, L., Guo, C.-J., Chen, L.-L., Huarte, M., 2021. Gene regulation by 450 long non-coding RNAs and its biological functions. Nat Rev Mol Cell 451 Biol 22, 96–118. https://doi.org/10.1038/s41580-020-00315-9 452 Takahata, Y., Hagino, H., Kimura, A., Urushizaki, M., Kobayashi, S., 453 Wakamori, K., Fujiwara, C., Nakamura, E., Yu, K., Kiyonari, H., Bando, 454 K., Murakami, T., Komori, T., Hata, K., Nishimura, R., 2021. Smoc1 and 455 Smoc2 regulate bone formation as downstream molecules of Runx2. Commun 456 Biol 4, 1199. https://doi.org/10.1038/s42003-021-02717-7 457 Wang, J., Zhao, Y., Gong, W., Liu, Y., Wang, M., Huang, X., Tan, J., 2021. 458 EDLMFC: an ensemble deep learning framework with multi-scale features 459 combination for ncRNA–protein interaction prediction. BMC 460 Bioinformatics 22, 133. https://doi.org/10.1186/s12859-021-04069-9 461 Yamada, D., Nakamura, M., Takao, T., Takihira, S., Yoshida, A., Kawai, S., 462 Miura, A., Ming, L., Yoshitomi, H., Gozu, M., Okamoto, K., Hojo, H., 463 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 17 Kusaka, N., Iwai, R., Nakata, E., Ozaki, T., Toguchida, J., Takarada, 464 T., 2021. Induction and expansion of human PRRX1+ limb-bud-like 465 mesenchymal cells from pluripotent stem cells. Nat Biomed Eng 5, 926–466 940. https://doi.org/10.1038/s41551-021-00778-x 467 Zhu, J., Yu, W., Wang, Y., Xia, K., Huang, Y., Xu, A., Chen, Q., Liu, B., 468 Tao, H., Li, F., Liang, C., 2019. lncRNAs: function and mechanism in 469 cartilage development, degeneration, and regeneration. Stem Cell Res 470 Ther 10, 344. https://doi.org/10.1186/s13287-019-1458-8 471 472 473 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 18 Figure Legend 474 Figure 1. lncRNA Expression Profiles 475 (A) Schematic overview of the methodology used in this study. 476 (B) MA plots comparing ExpLBM and HCT respective lncRNA expression levels. red: HCT 477 enrichment ncRNA, blue: ExpLBM enrichment ncRNA. 478 (C) Methodology for obtaining the conservation features. 479 (D) Bar plot of Conservation features of DEL_HCT. The number of human lncRNAs per 480 phylogenetic category is plotted against 740 HCT -enriched lncRNAs, classified by 481 conservation age on the horizontal axis and by phylogenetic category on the vertical axis. 482 483 Figure 2. DEL_HCT_hs-Protein Interactions 484 (A) Schematic overview of the prediction methodology. 485 (B) Correspondence map of predicted binding proteins with lncRNAs. Red indicates predicted 486 binding pairs; white indicates non-predicted binding pairs. 487 (C) Bubble plot of GO analysis results for 28 proteins predicted to bind lncRNAs 488 489 Figure 3. Triplex Predictions of DEL_HCT_hs 490 (A) Schematic overview of the prediction methodology. 491 (B) Predicted lncRNAs forming triplex structures with DEG_HCT promoter regions, their 492 target gene counts, and total formation sites. 493 (C) Heatmap of formation sites in cartilage marker genes, plotted using the raw values of 494 region numbers instead of Z-scores. 495 (D) Heatmap of formation sites in chondrocyte marker genes, plotted using the raw values of 496 region numbers instead of Z-scores. 497 (E) Heatmap of formation sites in human-specific coding genes, plotted using the raw values 498 of region numbers instead of Z-scores. 499 (F) Subjects of Gene Ontology analysis. 500 (G) Bubble plots representing the results of GO analysis in DEG_HCT. 501 (H) The heatmap depicting Fold Enrichment for the GO analysis conducted on DEG_HCT 502 predicted to be targeted by DEL_HCT_hs at the promoter region and other DEG_HCT. The 503 top five and bottom five values of the log2(Fold Change of Fold Enrichment) were extracted, 504 respectively. 505 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint 19 506 Supplementary Figure 1. Phylogenetic tree of biological classification 507 The range indicated by the classification used in Fig. 1E is shown in a box. 508 509 Supplementary Figure 2. Predicted coding peptide functions of DEL_HCT_hs 510 (A) Schematic overview of the prediction methodology. 511 (B) Bar graph showing lncRNAs reported to code for peptides and the number of peptides 512 coded by these lncRNAs. 513 (C) Bar graph of predicted peptide functions: the upper panel for Gene Ontology Molecular 514 Function, and the lower panel for Gene Ontology Biological Process. 515 516 Supplementary Figure 3. Triplex Predictions of DEL_HCT_hs 517 (A) Heatmap of formation sites for 9 DEL_HCT_hs predicted to form triplexes with 518 DEG_HCT promoter regions. The top 50 formation sites for each lncRNA were extracted 519 and combined into one heatmap. 520 (B) Heatmap of formation sites for human -specific proteins, plotted using the raw values of 521 region numbers instead of Z-scores. 522 523 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint A C Fig. 1 B d .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint A B C Fig. 2 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint A B C E D F Fig. 3 DEG_HCT G H .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint Supl. Fig. 1 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint A B C Supl.Fig. 2 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint Supl.Fig. 3 A B .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint Table 2 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 18, 2026. ; https://doi.org/10.64898/2026.02.17.706478doi: bioRxiv preprint