Could Metabolism-Related Long Non-Coding RNAs Be More Conserved than Their Brain-Related Counterparts?

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The topic is potentially interesting, but in its current form the manuscript provides little new insight. It reads largely as a descriptive/literature overview that leans heavily on existing reviews, and many of the central claims are stated without sufficient supporting evidence or clear rationale. Below I summarize the major concerns and suggest ways to strengthen the paper.

 

Major Points

 

1. Unclear rationale for the “metabolism vs. brain” framing (p.5).

The manuscript contrasts “metabolism” (presented as ancient/conserved) with “higher-order, species-specific traits” such as cognition, using the brain as the exemplar. The choice of brain as the comparison group is not justified. Please make explicit why the brain is the most appropriate comparator, and clarify the biological rationale linking “complexity” to expected patterns of evolutionary conservation. This will prevent the framing from appearing as an unsupported assumption.

 

2. Attribution of evolutionary conclusions to GWAS (p.5).

Several statements about lncRNA conservation (limited primary-sequence conservation; conservation detectable at promoter/splice/synteny/structure levels; rapid lineage-specific turnover) are attributed to “GWAS.” Those points are more typically derived from comparative genomics/evolutionary transcriptomics rather than GWAS per se. Either (a) clarify the specific GWAS analyses that support these statements (e.g., fine-mapped variants overlapping conserved promoters, heritability enrichment analyses, stratified LD score regression results), or (b) reword the text to attribute these observations to the appropriate genomic/evolutionary literature and cite the relevant studies.

 

3. Logical gap in concluding which features predict constraint (p.6).

The claim that mechanism (scaffold/decoy/guide), requirement for RNA structure, interaction with conserved proteins, and genomic context are more reliable predictors of evolutionary constraint than broad biological process (metabolism vs neurology) is asserted without direct evidence. Please either provide data or citations that directly compare these predictors, or explicitly acknowledge that this is a hypothesis and describe how it could be tested.

 

4. Bias in dataset assembly and group comparisons (p.6).

The proposed comparison of “metabolism-related” versus “neurobiology-related” lncRNAs depends on assigning lncRNAs to function via GO/pathway annotations. But lncRNA annotation is sparse and strongly biased by research focus (brain is commonly overrepresented). This can create annotation/ascertainment bias. Please describe how you will (or did) control for annotation depth and publication bias when assembling the two groups (for example: using uniform, evidence-based inclusion criteria, matching on annotation density or expression breadth, or performing sensitivity analyses).

 

Minor points

Reduce reliance on secondary reviews; include primary comparative genomics and functional studies where possible.

 

When making comparative claims, provide explicit examples or quantitative evidence rather than broad statements.

 

Clarify methods for any proposed dataset assembly, including inclusion/exclusion criteria and how functional attribution will be handled.

Author Response

Reviewer 1

The topic is potentially interesting, but in its current form the manuscript provides little new insight. It reads largely as a descriptive/literature overview that leans heavily on existing reviews, and many of the central claims are stated without sufficient supporting evidence or clear rationale. Below I summarize the major concerns and suggest ways to strengthen the paper.

Answer : We thank the reviewer for his justified and clever remarks, and have tried to our best to implement his insightful suggestions. We believe that the revised manuscript is stronger, thanks in great part to him, and thank him deeply.

 

Major Points

1. Unclear rationale for the “metabolism vs. brain” framing (p.5).

The manuscript contrasts “metabolism” (presented as ancient/conserved) with “higher-order, species-specific traits” such as cognition, using the brain as the exemplar. The choice of brain as the comparison group is not justified. Please make explicit why the brain is the most appropriate comparator, and clarify the biological rationale linking “complexity” to expected patterns of evolutionary conservation. This will prevent the framing from appearing as an unsupported assumption.

 

Answer : We thank the reviewer for highlighting the need to clarify the rationale behind using the brain as the comparison group in our framing. We aimed  to contrast two biological systems that differ in well-established evolutionary and functional properties.

Core metabolic pathways are among the most ancient and broadly conserved components of cellular biology, shared across nearly all forms of life due to their essential role in energy production and basic homeostasis. As a result, metabolic genes and processes are often expected to show strong evolutionary constraint across species.

In contrast, the brain was chosen as an paradigm of a tissue system characterized by high cellular diversity, complex regulatory architecture, and lineage-specific adaptations. Neural and cognitive traits are known to evolve rapidly in certain clades, driven by ecological and behavioral pressures, and are frequently associated with expanded gene regulatory innovation rather than strict conservation of molecular components.

We have revised the manuscript to explicitly state why the brain provides an informative comparator to metabolism in the context of evolutionary conservation, and to clarify that our use of “complexity” refers to differences in regulatory and phenotypic specialization, which can influence patterns of evolutionary constraint.

Manuscript revision:

We have added text in the Introduction (lines 82-92) explaining the biological basis for contrasting core metabolic processes with brain-related traits and clarifying the link between functional specialization, regulatory complexity, and expected conservation patterns. « In this work, we aimed to contrast two biological systems that differ in well-established evolutionary and functional properties. Core metabolic pathways are among the most ancient and broadly conserved components of cellular biology, shared across nearly all forms of life due to their essential role in energy production and basic homeostasis. As a result, metabolic genes and processes are often expected to show strong evolutionary constraint across species. In contrast, the brain was chosen as an paradigm of a tissue system characterized by high cellular diversity, complex regulatory architecture, and lineage-specific adaptations(7). Neural and cognitive traits are known to evolve rapidly in certain clades, driven by ecological and behavioral pressures, and are frequently associated with expanded gene regulatory innovation rather than strict conservation of molecular components.

Lines114-132 We added a nex paragraph to more properly descibe roles on lncRNAS in the brain.

« 3. lncRNAs in cerebral activity

We use the brain as an illustrative comparator because, unlike core metabolic pathways that are deeply conserved across taxa, neural systems exhibit high regulatory complexity and have undergone substantial lineage-specific diversification associated with cognition and behavior. This contrast provides a useful framework for examining how evolutionary constraint differs between ancient housekeeping functions and more specialized, rapidly evolving traits. Indeed, the neural tissues require finely tuned gene expression programs to coordinate the development and function of numerous specialized cell types. Certain evolutionary lineages exhibit accelerated changes in neural and cognitive features, driven by environmental and behavioral factors, and this flexibility is often associated with novel regulatory developments instead of strong molecular conservation(7). In this context, lncRNAs have emerged as key contributors to regulatory diversification, as they can modulate gene expression through interactions with chromatin, transcription factors, and RNA-binding proteins. Many lncRNAs display strong tissue- and cell-type specificity in the brain, supporting their role in shaping lineage-specific regulatory networks. Moreover, the rapid evolutionary turnover of lncRNAs suggests that they may provide a substrate for adaptive changes in neural gene regulation, ultimately contributing to the emergence of novel cognitive and behavioral traits. One could thus hypothesize that metabolic-Related lncRNAs could Be more conserved than their brain-related counterparts.

»

 

1. Attribution of evolutionary conclusions to GWAS (p.5).

Several statements about lncRNA conservation (limited primary-sequence conservation; conservation detectable at promoter/splice/synteny/structure levels; rapid lineage-specific turnover) are attributed to “GWAS.” Those points are more typically derived from comparative genomics/evolutionary transcriptomics rather than GWAS per se. Either (a) clarify the specific GWAS analyses that support these statements (e.g., fine-mapped variants overlapping conserved promoters, heritability enrichment analyses, stratified LD score regression results), or (b) reword the text to attribute these observations to the appropriate genomic/evolutionary literature and cite the relevant studies.

 Response:
We thank the reviewer for raising this important point. We agree that statements regarding limited primary-sequence conservation of lncRNAs, conservation detectable at promoter/splice/synteny/structural levels, and rapid lineage-specific turnover are primarily supported by comparative genomics and evolutionary transcriptomics studies rather than by GWAS analyses per se.

We have therefore revised the manuscript to attribute these observations to the appropriate evolutionary and comparative genomics literature and have adjusted the wording accordingly.

Manuscript revision:
The relevant section (lines 135-139) has been rephrased to remove the implication that GWAS provides the primary evidence for lncRNA conservation patterns, and additional citations from comparative genomics studies have been included. « “Despite the intuitive appeal of the preceding hypothesis, evidence from comparative genomics and evolutionary transcriptomics supports a more nuanced interpretation. While GWAS frequently implicate noncoding loci in disease susceptibility (15), evolutionary conservation patterns of lncRNAs have been primarily characterized through comparative genomics approaches (16).” 

Additional citations, 15 . Mirza et al. PloS One. 2014;9(8):e105723.  16.   Ulitsky I. et al . Nat Rev Genet. oct 2016;17(10):601‑14.

 

1. Logical gap in concluding which features predict constraint (p.6).

The claim that mechanism (scaffold/decoy/guide), requirement for RNA structure, interaction with conserved proteins, and genomic context are more reliable predictors of evolutionary constraint than broad biological process (metabolism vs neurology) is asserted without direct evidence. Please either provide data or citations that directly compare these predictors, or explicitly acknowledge that this is a hypothesis and describe how it could be tested.

 

Response:
We thank the reviewer for pointing out that this statement was too strongly phrased. We agree that, while lncRNA evolutionary constraint is often better explained by factors such as promoter conservation, structural requirements, genomic context, and interactions with conserved protein partners, direct quantitative comparisons across predictors remain limited.

We have therefore revised the text to clarify that this represents a working hypothesis rather than a definitive conclusion. We also added supporting citations from comparative genomics and lncRNA functional literature and explicitly note that our proposed framework (Section 4) provides a way to empirically test which molecular features best predict evolutionary constraint across functional categories.

Manuscript revision:
The relevant passage (lines 152-159) has been reworded to avoid overstatement and to frame these predictors as hypotheses supported by emerging evidence, with additional references and a clearer link to future testing. « “Taken together, these observations suggest that evolutionary constraint in lncRNAs may depend less on the broad biological domain in which a transcript operates (e.g., metabolism versus neurology) and more on specific molecular features such as mechanism of action (scaffold, decoy, guide), requirements for conserved RNA structure, genomic context, and interactions with conserved protein partners (Necsulea et ). However, we note that this remains an emerging hypothesis rather than a settled rule, and systematic comparative analyses will be required to directly evaluate which of these features best predict conservation across functional categories.” Added ref Necsulea A, Soumillon M, Warnefors M, Liechti A, Daish T, Zeller U, Baker JC, Grützner F, Kaessmann H. The evolution of lncRNA repertoires and expression patterns in tetrapods. Nature. 2014 Jan 30;505(7485):635-40.

 

 

 

1. Bias in dataset assembly and group comparisons (p.6).

The proposed comparison of “metabolism-related” versus “neurobiology-related” lncRNAs depends on assigning lncRNAs to function via GO/pathway annotations. But lncRNA annotation is sparse and strongly biased by research focus (brain is commonly overrepresented). This can create annotation/ascertainment bias. Please describe how you will (or did) control for annotation depth and publication bias when assembling the two groups (for example: using uniform, evidence-based inclusion criteria, matching on annotation density or expression breadth, or performing sensitivity analyses).

Response:
We thank the reviewer for highlighting the potential for annotation and publication bias in assembling metabolism- versus neurobiology-related lncRNA groups. We agree that lncRNA functional annotation is currently sparse and unevenly distributed, with brain-related transcripts often overrepresented due to research focus.

We deeply thank the reviewer for pointing this. Another reviewer raised similar points, pointing out the shortcomings of our strategy. To answer both reviewers, we have substantially expanded the ‘Proposed Strategy for Empirical Evaluation’ section by integrating multi-layer conservation metrics, structural constraint considerations, expression conservation, annotation bias controls, and specific conserved lncRNA examples. We also took into account the challenge that neuro-associated lncRNAs are disproportionately studied, leading to annotation depth bias.

 We have also added a Figure (Figure 2) to illustrate these points, and the experimental strategy.

 

Manuscript revision lines 179-250

«4. Proposed Strategy for Empirical Evaluation

A major unresolved question in evolutionary transcriptomics is whether lncRNAs involved in deeply conserved physiological systems (such as metabolism) experience stronger evolutionary constraint than those associated with more lineage-adaptive systems (such as neurobiology). Addressing this requires an empirical framework that goes beyond primary-sequence identity and instead integrates multiple layers of conservation, including genomic context, regulatory architecture, structural constraints, and expression preservation.

4.1 Defining functional lncRNA sets across biological domains

The first step is the assembly of high-confidence lncRNA sets associated with metabolism and neurobiology. Rather than relying solely on tissue-enriched expression, inclusion should be based on functional evidence, such as experimentally validated roles in metabolic regulation (e.g., lncRNAs implicated in insulin signaling, lipid homeostasis, or mitochondrial pathways) or neurodevelopmental processes (e.g., synaptic regulation, neuronal differentiation). Curated annotation resources such as GENCODE, LNCipedia, and NONCODE provide transcript models, while pathway association can be derived from Gene Ontology, KEGG, and Reactome-based enrichment. Importantly, functional assignments must be supported either by mechanistic studies or multiple independent datasets, reducing the risk of annotation noise.

4.2 Multi-layered conservation metrics beyond nucleotide identity

A central point of the “lncRNA conservation paradox” is that functional constraint may not be reflected in linear sequence similarity. Therefore, comparative analyses should incorporate several orthogonal conservation layers:

• Primary sequence constraint, quantified using PhastCons or PhyloP scores, remains informative for short conserved motifs or exon cores.
• Promoter and enhancer conservation may better capture selective pressure acting on transcriptional regulation rather than RNA sequence itself.
• Splice-site and exon–intron architecture conservation can indicate preserved transcript processing, even when exon sequences diverge.
• Syntenic conservation, defined by preserved genomic neighborhood, is particularly important for rapidly evolving lncRNAs lacking clear orthologs.

Such approaches allow classification of lncRNAs into distinct evolutionary modes: sequence-conserved, regulatory-conserved, structure-conserved, or lineage-specific.

4.3 Structural constraint as a major driver of lncRNA conservation

Many lncRNAs act through RNA secondary or tertiary structures rather than through encoded peptides or strict motif. Thus, conservation should also be evaluated at the structural level using covariation-based methods and comparative folding approaches. For example, deeply conserved lncRNAs such as MALAT1 and NEAT1 retain conserved structural domains despite modest overall sequence conservation, suggesting selection on RNA architecture. Integrating structure-aware conservation metrics may therefore reveal hidden constraint missed by standard alignments.

4.4 Expression conservation and functional analogy across species

Because many lncRNAs are tissue-specific, expression conservation provides an additional axis of evolutionary constraint. Comparative transcriptomic atlases across vertebrates can be used to assess whether orthologous loci exhibit conserved developmental timing, tissue restriction, or stimulus responsiveness. Notably, conservation of expression patterns may persist even in cases where direct orthology is unclear, supporting the possibility of functional analogs rather than strict sequence orthologs.

4.5 Controlling for annotation bias and domain-specific overrepresentation

A major challenge is that neuro-associated lncRNAs are disproportionately studied, leading to annotation depth bias. Metabolism-related lncRNAs may appear less conserved simply because fewer have been characterized. To mitigate this, metabolism- and neuro-lncRNA groups should be matched for transcript length, exon count, expression breadth, and publication density. Sensitivity analyses under alternative inclusion thresholds can further test robustness.

4.6 Expected outcomes and interpretation

If metabolism-associated lncRNAs exhibit higher conservation across multiple layers (promoter, synteny, structure, expression), this would support the hypothesis that essential homeostatic pathways impose stronger evolutionary constraint on regulatory noncoding transcripts. Conversely, if both groups show similarly weak sequence constraint but differ in regulatory or structural preservation, this would suggest that lncRNA evolution is shaped less by pathway category and more by molecular mechanism of action (guide, scaffold, decoy) and interaction with conserved protein complexes.  Overall, this framework provides a testable strategy to resolve whether evolutionary conservation in lncRNAs reflects biological domain, mechanistic requirement, or regulatory architecture. »

 

 

 

Minor points

Reduce reliance on secondary reviews; include primary comparative genomics and functional studies where possible.

 

We agree with the reviewer and added more original references : Carninci et  al (1) and Kasparov et al. (2),  Boyer et al, Bower et al, Blume et al. (see ref 5-7, please.)

 

When making comparative claims, provide explicit examples or quantitative evidence rather than broad statements.

 

We agree with the reviewer ; We hope that the added paragraphs (see above) will answer to his remarks.

 

Clarify methods for any proposed dataset assembly, including inclusion/exclusion criteria and how functional attribution will be handled.

Response:
We thank the reviewer for this insightful suggestion. We have clarified our proposed methods for dataset assembly. Specifically, metabolism- and neurobiology-associated lncRNAs will be selected using curated annotation resources (e.g., GENCODE) and functional evidence from the literature and pathway databases. Inclusion criteria require either experimentally validated functional annotation or strong computational prediction supported by multiple independent studies. LncRNAs with ambiguous or conflicting functional assignments will be excluded. Functional attribution will be based on pathway enrichment, GO terms, or tissue-specific expression patterns, ensuring consistent and evidence-based group assignment. Annotation depth and expression breadth will be recorded to facilitate sensitivity analyses and control for potential bias.

Manuscript revision (Section 4 lines 181-193):

“To assemble metabolism- and neurobiology-associated lncRNA groups, we suggest to use curated databases (e.g., GENCODE) and literature-based evidence. Inclusion requires experimental validation or computational predictions supported by multiple independent studies. LncRNAs with ambiguous or conflicting annotations would be excluded. Functional attribution would rely on pathway enrichment, GO terms, and tissue-specific expression patterns. Annotation depth and expression breadth would be recorded to allow normalization and sensitivity analyses, reducing potential bias in group comparisons.”

 

 

Reviewer 2 Report

Comments and Suggestions for Authors

This review by Laurent Metzinger and Valérie Metzinger-Le Meuth poses the question of whether metabolic-related long non-coding RNAs (lncRNAs) might be more conserved than their brain-related counterparts, despite evidence that may suggest otherwise. This is certainly an important question that warrants careful consideration and more explicit discussion.

I fully agree with the authors that a more thorough investigation of lncRNAs features, extending beyond primary sequence conservation alone predictive of their function is required. 

To what might need clarification:

As mentioned, current data show that sequence conservation by itself is not a reliable metric of functionality. However, relying on selected examples such as MALAT1, NEAT1, and MEG3, which the authors themselves note represent notable exceptions, may be somewhat confusing, especially for those readers who are not deeply familiar with the field. These lncRNAs are broadly studied and implicated in cancer and many other biological contexts and therefore may not be representative of any specific lncRNA population. Moreover, a small number of well-studied examples is insufficient to draw conclusions that contradict broader trends in lncRNA evolution. This should perhaps be clarified.

"Given their abundance (with lncRNAs estimated at sixty thousands of transcripts, greatly outnumbering the 2500 miRNAs"

I am not sure whether this number is correct. If so, please state / cite the source that proves it and indicate the relevant species. As a note, according to Gencode v49 the number of human lncRNAs transcripts is 191,079 https://www.gencodegenes.org/human/stats.html.

Suggestion:

I also think the authors could more briefly discuss difficulties in determine which lncRNAs are truly functional, as estimates vary substantially depending on whether they are derived from computational predictions or experimental evidence. For example, work from the FANTOM Consortium used multiple metrics to assess functionality and suggested that approximately 60–70% of lncRNAs may be potentially functional (Hon CC, Ramilowski JA, et al., Nature, 2017; 543:199–204). In contrast, subsequent experimental-based estimates from the same consortium placed this number in a considerably lower range, around 20–30% in human fibroblasts (Ramilowski JA, Yip CW, et al., Genome Research, 2020; 30:1060–1072). This apparent discrepancy may, however, simply reflect the limited scope of current experimental evidence, as lncRNA function depends not only on cell type but also on cell state, disease context, cellular activation, and environmental stimuli. Addressing these complexities will likely require more thorough and systematic computational approaches, as proposed here, in combination with broader experimental frameworks. 

Specific text points:

I am confused by certain wording, as it seems to imply the opposite to what is written in the text.

“Limitations of Predicting Conservation Based on Biological Function.”

Limitations of Predicting Biological Function Based on Conservation?

“This apparent evolutionary paradox questions how functional relevance relates to conservation”

This apparent evolutionary paradox questions how conservation relates to functional relevance?

Potentially missing references:

“Moreover, ncRNAs appear to act as key regulators of chromatin states, epigenetic trajectories, and transcriptional networks, frequently in cooperation with proteins harboring intrinsically disordered regions, which dominate known gene-regulatory protein classes.”

I recommend adding references here. Ideally a few.

“Some lncRNAs even appear to arise de novo from previously untranscribed genomic regions, giving rise to lineage-specific regulatory transcripts.:

I recommend adding a reference here.

 

Comments on the Quality of English Language

Although this is primarily an editorial issue, please ensure that the final manuscript is carefully and consistently formatted, including citation placement and spacing. E.q. "cellular homeostasis, and diseases.(1)" vs "providing compelling evidence that ncRNAs contribute directly to human pathology(2)."

In addition, please check typographical consistency in reference formatting—for instance, in “MALAT1 (5).” the period appears to be bolded.

Author Response

Reviewer 2

This review by Laurent Metzinger and Valérie Metzinger-Le Meuth poses the question of whether metabolic-related long non-coding RNAs (lncRNAs) might be more conserved than their brain-related counterparts, despite evidence that may suggest otherwise. This is certainly an important question that warrants careful consideration and more explicit discussion.

I fully agree with the authors that a more thorough investigation of lncRNAs features, extending beyond primary sequence conservation alone predictive of their function is required. 

 

Answer : We thank reviewer for his interest in our paper, and his justified and clever remarks, and have tried to our best to implement his insightful suggestions. We believe that the revised manuscript is stronger, thanks in great part to him, and thank him deeply.

 

To what might need clarification:

As mentioned, current data show that sequence conservation by itself is not a reliable metric of functionality. However, relying on selected examples such as MALAT1, NEAT1, and MEG3, which the authors themselves note represent notable exceptions, may be somewhat confusing, especially for those readers who are not deeply familiar with the field. These lncRNAs are broadly studied and implicated in cancer and many other biological contexts and therefore may not be representative of any specific lncRNA population. Moreover, a small number of well-studied examples is insufficient to draw conclusions that contradict broader trends in lncRNA evolution. This should perhaps be clarified.

 

Answer  We fully agree with the reviewer’s clever remark and have integrated it into our manuscript in the relevant place, lines 99-102.

Manuscript revision

« Thus, examples such as MALAT1 and NEAT1, which are broadly studied and implicated in cancer and many other biological contexts, may not be representative of any specific lncRNA population. Moreover, a small number of well-studied examples is insufficient to draw conclusions that contradict broader trends in lncRNA evolution.»

"Given their abundance (with lncRNAs estimated at sixty thousands of transcripts, greatly outnumbering the 2500 miRNAs"

I am not sure whether this number is correct. If so, please state / cite the source that proves it and indicate the relevant species. As a note, according to Gencode v49 the number of human lncRNAs transcripts is 191,079 https://www.gencodegenes.org/human/stats.html.

 

We agree and apologize as we took outdated numbers ; The relevant number and URL are now in the revised version. Lines 56-58

Manuscript revision

« Given their abundance (with lncRNAs estimated at 191,079, greatly outnumbering the 7563 short RNAs, GENCODE version 49, https://www.gencodegenes.org/human/stats.html)... »

 

Suggestion:

I also think the authors could more briefly discuss difficulties in determine which lncRNAs are truly functional, as estimates vary substantially depending on whether they are derived from computational predictions or experimental evidence. For example, work from the FANTOM Consortium used multiple metrics to assess functionality and suggested that approximately 60–70% of lncRNAs may be potentially functional (Hon CC, Ramilowski JA, et al., Nature, 2017; 543:199–204). In contrast, subsequent experimental-based estimates from the same consortium placed this number in a considerably lower range, around 20–30% in human fibroblasts (Ramilowski JA, Yip CW, et al., Genome Research, 2020; 30:1060–1072). This apparent discrepancy may, however, simply reflect the limited scope of current experimental evidence, as lncRNA function depends not only on cell type but also on cell state, disease context, cellular activation, and environmental stimuli. Addressing these complexities will likely require more thorough and systematic computational approaches, as proposed here, in combination with broader experimental frameworks. 

 

Answer We thank the reviewer for his clever suggestions, that significantly improved our introduction. We have adapted them in our manuscript with our deepest thanks, lines 59-71. Relevant citations were added to the manuscript.

Manuscript revision

« There are difficulties in determining which lncRNAs are functional, as estimates vary substantially depending on whether they are derived from computational predictions or experimental evidences. For example, work from the FANTOM Consortium used multiple metrics to assess functionality and suggested that approximately 60–70% of lncRNAs may be potentially functional (Hon CC, Ramilowski JA, et al., Nature, 2017; 543:199–204). In contrast, subsequent experimental-based estimates from the same consortium placed this number in a considerably lower range, around 20–30% in human fibroblasts (Ramilowski JA, Yip CW, et al., Genome Research, 2020; 30:1060–1072). This apparent discrepancy may, however, simply reflect the limited scope of current experimental evidences, as a given lncRNA’s function depends not only on cell type but also on cell state, disease context, cellular activation, and environmental stimuli. Addressing these complexities will likely require more thorough and systematic computational approaches, as proposed here, in combination with broader experimental frameworks. »

 

Specific text points:

I am confused by certain wording, as it seems to imply the opposite to what is written in the text.

“Limitations of Predicting Conservation Based on Biological Function.”

Limitations of Predicting Biological Function Based on Conservation?

 

Answer We agree fully with the reviewer and thank him for his clever suggestion, that we implemented in the manuscript.

Revision

In the abstract, « This apparent evolutionary paradox questions the limitations of predicting biological function based on conservation, »

 

Potentially missing references:

“Moreover, ncRNAs appear to act as key regulators of chromatin states, epigenetic trajectories, and transcriptional networks, frequently in cooperation with proteins harboring intrinsically disordered regions, which dominate known gene-regulatory protein classes.”

I recommend adding references here. Ideally a few.

 

Answer We agree, and have added relevant original research citations at the end of this sentence. Boyer et al, Bower et al, Blume et al, and a review Mattick et al. (see ref 4-7, please.)

 

“Some lncRNAs even appear to arise de novo from previously untranscribed genomic regions, giving rise to lineage-specific regulatory transcripts:

I recommend adding a reference here. Ruiz-Orera et al. NAR Genomics Bioinforma. 1 avr 2019;1(1):e2‑e2.

 

Answer We agree, and have added a relevant citation at the end of this sentence. Ruiz-Orera et al. NAR Genomics Bioinforma. 1 avr 2019;1(1):e2‑e2.

 

Comments on the Quality of English Language

Although this is primarily an editorial issue, please ensure that the final manuscript is carefully and consistently formatted, including citation placement and spacing. E.q. "cellular homeostasis, and diseases.(1)" vs "providing compelling evidence that ncRNAs contribute directly to human pathology(2)."

In addition, please check typographical consistency in reference formatting—for instance, in “MALAT1 (5).” the period appears to be bolded.

 

Answer We thank the reviewer for having spotted these formatting mistakes. We have corrected them.

 

 

Reviewer 3 Report

Comments and Suggestions for Authors

The work by Metzinger and Metzinger-Le Meuth presents a review on the long non-coding RNA (lncRNA) evolution at different levels of conservation and in different domains of regulation, with the goal to resolve the apparent paradox between the low sequence conservation of lncRNAs and their functional criticality. They proposed a comparative framework for evaluating evolutionary conservation of lncRNAs engaged in different processes (i.e., metabolism and neurology). In addition, they discussed the relevance of lncRNA conservation in light of precise medicine.

Comments and suggestions:

1. There are several issues with the graphical elements. First, only one figure is provided, and this figure is about the biogenesis and the mechanisms of lncRNAs regulation, which is relevant but not the core of this review, that is, the evolutionary conservation constraints of lncRNAs. Second, details of how lncRNAs regulate gene expression, chromatin structure, protein interaction and localization are missing. The authors may consider adding detailed schematics to these aspects instead of simple words such as “Chromatin modifiers” or “4. ceRNA”. In summary, the figure in the current version looks crude and needs to be improved significantly.
2. The type of the current submission is “review”, however, the work reads more like an opinion or a perspective. Despite the fact that the authors proposed a framework to evaluate the lncRNA evolutionary conservation paradox, a critical problem and an important topic, the manuscript suffers from a severe lack of details and depth. For instance, page 6, “Proposed strategy for empirical evaluation”. This section is the heart of the work, but only reads like a “Future work” section. It has a to-do list (e.g., what lncRNA related database/dataset to use, what conservation metrics to apply, which genomic regions and pathways to compare), but without any results or preliminary data to support the authors’ point of view, nor does it provide fundamental insights into the lncRNA biology. The authors are expected to strengthen this section significantly.
3. Section “Therapeutic Implications”. This part also reads wonky. If I understand correctly, the theme of this review is the evolutionary origin of the lncRNAs conservation, and I do not see how therapeutics can straightforwardly fit. It may be better to discuss the therapeutic potential of lncRNAs from the motif/sequence/structure conservation perspective. Anyway, the relatedness between lncRNAs and drugs should be still in the context of evolution. Another issue is similar to the previous comment. This part also reads rather superficial and the authors are expected to elaborate with more details.

Author Response

Reviewer3

The work by Metzinger and Metzinger-Le Meuth presents a review on the long non-coding RNA (lncRNA) evolution at different levels of conservation and in different domains of regulation, with the goal to resolve the apparent paradox between the low sequence conservation of lncRNAs and their functional criticality. They proposed a comparative framework for evaluating evolutionary conservation of lncRNAs engaged in different processes (i.e., metabolism and neurology). In addition, they discussed the relevance of lncRNA conservation in light of precise medicine.

Answer : We thank reviewer for his justified and clever remarks, and have tried to our best to implement his insightful suggestions. We believe that the revised manuscript is stronger, thanks in great part to him, and thank him deeply.

 

Comments and suggestions:

1. There are several issues with the graphical elements. First, only one figure is provided, and this figure is about the biogenesis and the mechanisms of lncRNAs regulation, which is relevant but not the core of this review, that is, the evolutionary conservation constraints of lncRNAs. Second, details of how lncRNAs regulate gene expression, chromatin structure, protein interaction and localization are missing. The authors may consider adding detailed schematics to these aspects instead of simple words such as “Chromatin modifiers” or “4. ceRNA”. In summary, the figure in the current version looks crude and needs to be improved significantly.

Answer We agree with the reviewer, and have altered the figure 1, in a more detailed fashion, and have added an abbreviation list. We hope that our figure is now suitable. We have added a figure 2, that summarizes the experimental procedure suggested in our article, as discussed below.

 

1. The type of the current submission is “review”, however, the work reads more like an opinion or a perspective. Despite the fact that the authors proposed a framework to evaluate the lncRNA evolutionary conservation paradox, a critical problem and an important topic, the manuscript suffers from a severe lack of details and depth. For instance, page 6, “Proposed strategy for empirical evaluation”. This section is the heart of the work, but only reads like a “Future work” section. It has a to-do list (e.g., what lncRNA related database/dataset to use, what conservation metrics to apply, which genomic regions and pathways to compare), but without any results or preliminary data to support the authors’ point of view, nor does it provide fundamental insights into the lncRNA biology. The authors are expected to strengthen this section significantly.

 

Answer We deeply thank the reviewer for pointing out the shortcomings of our strategy. Another reviewer raised similar points. To answer this, we have substantially expanded the ‘Proposed Strategy for Empirical Evaluation’ section by integrating multi-layer conservation metrics, structural constraint considerations, expression conservation, annotation bias controls, and specific conserved lncRNA examples. We have added a Figure (Figure 2) to illustrate these points.

 

Response

«4. Proposed Strategy for Empirical Evaluation (Revised)

A major unresolved question in evolutionary transcriptomics is whether lncRNAs involved in deeply conserved physiological systems (such as metabolism) experience stronger evolutionary constraint than those associated with more lineage-adaptive systems (such as neurobiology). Addressing this requires an empirical framework that goes beyond primary-sequence identity and instead integrates multiple layers of conservation, including genomic context, regulatory architecture, structural constraints, and expression preservation.

4.1 Defining functional lncRNA sets across biological domains

The first step is the assembly of high-confidence lncRNA sets associated with metabolism and neurobiology. Rather than relying solely on tissue-enriched expression, inclusion should be based on functional evidence, such as experimentally validated roles in metabolic regulation (e.g., lncRNAs implicated in insulin signaling, lipid homeostasis, or mitochondrial pathways) or neurodevelopmental processes (e.g., synaptic regulation, neuronal differentiation). Curated annotation resources such as GENCODE, LNCipedia, and NONCODE provide transcript models, while pathway association can be derived from Gene Ontology, KEGG, and Reactome-based enrichment. Importantly, functional assignments must be supported either by mechanistic studies or multiple independent datasets, reducing the risk of annotation noise.

4.2 Multi-layered conservation metrics beyond nucleotide identity

A central point of the “lncRNA conservation paradox” is that functional constraint may not be reflected in linear sequence similarity. Therefore, comparative analyses should incorporate several orthogonal conservation layers:

• Primary sequence constraint, quantified using PhastCons or PhyloP scores, remains informative for short conserved motifs or exon cores.
• Promoter and enhancer conservation may better capture selective pressure acting on transcriptional regulation rather than RNA sequence itself.
• Splice-site and exon–intron architecture conservation can indicate preserved transcript processing, even when exon sequences diverge.
• Syntenic conservation, defined by preserved genomic neighborhood, is particularly important for rapidly evolving lncRNAs lacking clear orthologs.

Such approaches allow classification of lncRNAs into distinct evolutionary modes: sequence-conserved, regulatory-conserved, structure-conserved, or lineage-specific.

4.3 Structural constraint as a major driver of lncRNA conservation

Many lncRNAs act through RNA secondary or tertiary structures rather than through encoded peptides or strict motif. Thus, conservation should also be evaluated at the structural level using covariation-based methods and comparative folding approaches. For example, deeply conserved lncRNAs such as MALAT1 and NEAT1 retain conserved structural domains despite modest overall sequence conservation, suggesting selection on RNA architecture. Integrating structure-aware conservation metrics may therefore reveal hidden constraint missed by standard alignments.

4.4 Expression conservation and functional analogy across species

Because many lncRNAs are tissue-specific, expression conservation provides an additional axis of evolutionary constraint. Comparative transcriptomic atlases across vertebrates can be used to assess whether orthologous loci exhibit conserved developmental timing, tissue restriction, or stimulus responsiveness. Notably, conservation of expression patterns may persist even in cases where direct orthology is unclear, supporting the possibility of functional analogs rather than strict sequence orthologs.

4.5 Controlling for annotation bias and domain-specific overrepresentation

A major challenge is that neuro-associated lncRNAs are disproportionately studied, leading to annotation depth bias. Metabolism-related lncRNAs may appear less conserved simply because fewer have been characterized. To mitigate this, metabolism- and neuro-lncRNA groups should be matched for transcript length, exon count, expression breadth, and publication density. Sensitivity analyses under alternative inclusion thresholds can further test robustness.

4.6 Expected outcomes and interpretation

If metabolism-associated lncRNAs exhibit higher conservation across multiple layers (promoter, synteny, structure, expression), this would support the hypothesis that essential homeostatic pathways impose stronger evolutionary constraint on regulatory noncoding transcripts. Conversely, if both groups show similarly weak sequence constraint but differ in regulatory or structural preservation, this would suggest that lncRNA evolution is shaped less by pathway category and more by molecular mechanism of action (guide, scaffold, decoy) and interaction with conserved protein complexes.  Overall, this framework provides a testable strategy to resolve whether evolutionary conservation in lncRNAs reflects biological domain, mechanistic requirement, or regulatory architecture. »

 

 

1. Section “Therapeutic Implications”. This part also reads wonky. If I understand correctly, the theme of this review is the evolutionary origin of the lncRNAs conservation, and I do not see how therapeutics can straightforwardly fit. It may be better to discuss the therapeutic potential of lncRNAs from the motif/sequence/structure conservation perspective. Anyway, the relatedness between lncRNAs and drugs should be still in the context of evolution. Another issue is similar to the previous comment. This part also reads rather superficial and the authors are expected to elaborate with more details.

Answer We agree totally with the reviewer that this transition between paragraphs 4 and 5 is too abrupt. We have thus added a sub-chapter to explicitly state why the evolutionary/conservation discussion matters for therapy

Manuscript Revision.

We inserted at the start of paragraph 5 (lines 207-217), the following lines, in order to help to smoothen the transition.

« Taken together, these considerations highlight that evolutionary conservation in lncRNAs cannot be inferred solely from broad functional categories such as metabolism or neurobiology, but instead reflects a complex interplay of structural constraints, genomic context, and molecular mechanism. Importantly, this issue is not only of theoretical interest for evolutionary transcriptomics, but also carries direct translational relevance. Indeed, understanding which lncRNAs are conserved, which are lineage-specific, and which molecular features are maintained across species has major implications for functional prioritization, experimental modeling, and ultimately therapeutic targeting. As efforts to manipulate the non-coding transcriptome advance, integrating evolutionary and mechanistic insight will be essential for identifying robust disease-associated lncRNAs and for designing interventions that achieve both specificity and efficacy in vivo. So, »

 

 

 

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript primarily presents opinions but does not develop sufficiently informative or evidence-based insights. In its current form, it reads more like a research proposal than a scholarly article. Therefore, I think it is suitable  for substantial revision.

Author Response

a seen with editor, nothing to reply here.

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have overhauled the overall structure of the manuscript, and it reads with  much more flow now. There are a couple of minor issues remaining:

1. Line 161 and 188. There are two Section 4: please correct the numbering.
2. Line 210-222. As previously commented, technical depth is desirable in this section  (i.e. “4.2 Multi-layered conservation metrics beyond nucleotide identity”) because  this is the core methodology section. For example, “Splice-site and exon–intron architecture conservation”. It would be much better if the authors can provide info such as the specific metric/tool that can be used to fulfill this task, just like “PhastCons” and “PhyloP” scores as the authors mentioned in the first bullet point.
3. Line 224-232, “…conservation should also be evaluated at the structural level using covariation-based methods and comparative folding approaches.” Like the previous comment, the authors are recommended to provide more details regarding the specific metrics/methods/software tools to evaluate the structural level conservation. Similar comments apply to the next subsection (“4.4 Expression conservation and functional analogy across species”) as well.
4. Literature references are lacking for metrics/methods such as “PhastCons”, “PhyloP”, etc.

Author Response

Reviewer

 

The authors have overhauled the overall structure of the manuscript, and it reads with  much more flow now.

 

Answer : We thank the reviewer for his kind remark and the profound interest he showed or our manuscript. He helped us to profoundly ameliorate it.

 

There are a couple of minor issues remaining:

1. Line 161 and 188. There are two Section 4: please correct the numbering.

We thank deeply the reviewer for having spotted this oversight. The sections have been renumbered accordingly. There is now one section 4, followed by sections 5 (metrics) and 6 (conclusion).

 

1. Line 210-222. As previously commented, technical depth is desirable in this section  (i.e. “4.2 Multi-layered conservation metrics beyond nucleotide identity”) because  this is the core methodology section. For example, “Splice-site and exon–intron architecture conservation”. It would be much better if the authors can provide info such as the specific metric/tool that can be used to fulfill this task, just like “PhastCons” and “PhyloP” scores as the authors mentioned in the first bullet point.

 

Answer :

We thank the reviewer for this helpful suggestion. We agree that Section 5.2 (which has been renumbered, see above) would benefit from greater methodological specificity, particularly given its central role in outlining the proposed comparative framework. We have therefore expanded this section to include concrete examples of computational metrics and software tools that can be used to assess splice-site conservation, exon–intron architecture preservation, synteny, and regulatory conservation. These now include tools such as LiftOver, CESAR, MAFFT/Clustal-based splice-site alignment, GERP++, and genome-wide synteny mapping approaches. References relevant ot these new tools and metrics have been added accordingly

The sub chapter now reads :

5.2 Multi-layered conservation metrics beyond nucleotide identity

A central point of the “lncRNA conservation paradox” is that functional constraint may not be reflected in linear sequence similarity. Therefore, comparative analyses should incorporate several orthogonal conservation layers:

• Primary sequence constraint can be quantified using PhastCons and PhyloP scores derived from multi-species alignments (27), as well as GERP++ scores to detect rejected substitutions (28). These metrics allow detection of conserved elements within exons or short functional motifs.
• Promoter and enhancer conservation may better capture selective pressure acting on transcriptional regulation rather than RNA sequence itself. They may be assessed using cross-species alignment of regulatory regions combined with chromatin-state conservation derived from comparative ATAC-seq or ChIP-seq datasets. Tools such as LiftOver enable mapping of promoter coordinates across genomes (29), while conservation of transcription factor binding sites can be evaluated using position-weight matrix (PWM) analysis across aligned promoter regions.
• Splice-site and exon–intron architecture conservation can indicate preserved transcript processing, even when exon sequences diverge. They can be evaluated by comparing exon boundaries across species using genome alignments (e.g., via Ensembl Compara). Tools such as CESAR (Coding Exon-Structure Aware Realigner) and splicing-aware aligners allow assessment of conserved splice donor/acceptor sites (30). Conservation of intron phase, exon number, and transcript isoform structure provides evidence of selective constraint even when exon sequences diverge.
• Syntenic conservation, defined by preserved genomic neighborhood, is particularly important for rapidly evolving lncRNAs lacking clear orthologs. It can be analyzed using pairwise or multi-species synteny blocks derived from Ensembl Compara or MCScanX. For rapidly evolving lncRNAs lacking clear sequence orthology, conserved flanking protein-coding genes can provide evidence of positional orthology (31,32).

Such approaches allow classification of lncRNAs into distinct evolutionary modes: sequence-conserved, regulatory-conserved, structure-conserved, or lineage-specific.

 

 

 

1. Line 224-232, “…conservation should also be evaluated at the structural level using covariation-based methods and comparative folding approaches.” Like the previous comment, the authors are recommended to provide more details regarding the specific metrics/methods/software tools to evaluate the structural level conservation. Similar comments apply to the next subsection (“4.4 Expression conservation and functional analogy across species”) as well.

Answer :

We appreciate this important comment. In response, we have substantially expanded Sections 5.3 and 5.4 to provide concrete examples of structural conservation tools (e.g., RNAz, EvoFold, R-scape, Infernal, and covariance models) as well as transcriptomic comparison approaches for expression conservation (e.g., cross-species RNA-seq normalization, TPM-based comparisons, Tau specificity index, correlation-based conservation scoring, and single-cell cross-species integration frameworks).

. References relevant ot these tools have been added accordingly

The sub chapters 5.3 end 5.4 now read :

5.3 Structural constraint as a major driver of lncRNA conservation

Many lncRNAs act through RNA secondary or tertiary structures rather than through encoded peptides or strict motif. Thus, conservation should also be evaluated at the structural level using covariation-based methods and comparative folding approaches. For example, deeply conserved lncRNAs such as MALAT1 and NEAT1 retain conserved structural domains despite modest overall sequence conservation, suggesting selection on RNA architecture. Integrating structure-aware conservation metrics may therefore reveal hidden constraint missed by standard alignments. Because many lncRNAs act through RNA secondary or tertiary structures rather than primary sequence motifs, conservation should be evaluated at the structural level using comparative and covariation-based approaches. Tools such as RNAz and EvoFold detect thermodynamically stable and evolutionarily conserved RNA secondary structures across multiple alignments (33). Covariation analysis using R-scape can identify compensatory base changes indicative of selection acting on RNA structure (1,34). Infernal and covariance models (CMs) further enable structure-aware homology searches that detect conserved structural domains even when linear sequence similarity is weak.

Comparative folding strategies, integrating minimum free energy predictions with cross-species alignments (e.g., via MAFFT or Clustal Omega), can reveal structurally constrained domains embedded within otherwise rapidly evolving transcripts (35,36). Such structure-informed metrics may uncover hidden evolutionary constraint that is not detectable through conventional nucleotide identity thresholds.

 

5.4 Expression conservation and functional analogy across species

Because many lncRNAs are tissue-specific, expression conservation provides an additional axis of evolutionary constraint. Comparative transcriptomic atlases across vertebrates can be used to assess whether orthologous loci exhibit conserved developmental timing, tissue restriction, or stimulus responsiveness. Notably, conservation of expression patterns may persist even in cases where direct orthology is unclear, supporting the possibility of functional analogs rather than strict sequence orthologs. Cross-species RNA-seq datasets can be normalized using TPM-based approaches or variance-stabilizing transformations to enable interspecies comparisons. Conservation of expression can then be quantified using correlation-based metrics across matched tissues, developmental stages, or physiological conditions.  Tissue specificity can be assessed using indices such as the Tau specificity index, enabling comparison of breadth of expression across species (37). In addition, emerging cross-species single-cell transcriptomic integration frameworks (e.g., ortholog-guided clustering or mutual nearest neighbor approaches) allow evaluation of whether lncRNA loci exhibit conserved cell-type–restricted expression patterns.

Importantly, expression conservation may persist even when strict sequence orthology is unclear, raising the possibility of functional analogs rather than direct orthologs. Integrating positional conservation, structural features, and cross-species expression similarity therefore provides a more comprehensive view of lncRNA evolutionary constraint.

 

1. Literature references are lacking for metrics/methods such as “PhastCons”, “PhyloP”, etc.

 

Answer  We thank deeply the reviewer for having spotted this oversight. The relevant reference has been added to the text accordingly. Ramani R, Krumholz K, Huang YF, Siepel A. PhastWeb: a web interface for evolutionary conservation scoring of multiple sequence alignments using phastCons and phyloP. Bioinformatics. 1 juill 2019;35(13):2320‑2. doi:10.1093/bioinformatics/bty966 PubMed PMID: 30481262; PubMed Central PMCID: PMC6596881.

 

As we have suggested the use of more tools in 5.3 and 5.4, according to the reviewer’s insightful query, we added relevant references for all of them.

 

Ramani R, Krumholz K, Huang YF, Siepel A. PhastWeb: a web interface for evolutionary conservation scoring of multiple sequence alignments using phastCons and phyloP. Bioinformatics. 1 juill 2019;35(13):2320‑2. doi:10.1093/bioinformatics/bty966 PubMed PMID: 30481262; PubMed Central PMCID: PMC6596881.

Davydov EV, Goode DL, Sirota M, Cooper GM, Sidow A, Batzoglou S. Identifying a high fraction of the human genome to be under selective constraint using GERP++. PLoS Comput Biol. 2 déc 2010;6(12):e1001025. doi:10.1371/journal.pcbi.1001025 PubMed PMID: 21152010; PubMed Central PMCID: PMC2996323.

Park KJ, Yoon YA, Park JH. Evaluation of Liftover Tools for the Conversion of Genome Reference Consortium Human Build 37 to Build 38 Using ClinVar Variants. Genes. 26 sept 2023;14(10):1875. doi:10.3390/genes14101875 PubMed PMID: 37895222; PubMed Central PMCID: PMC10606611.

Sharma V, Hiller M. Coding Exon-Structure Aware Realigner (CESAR): Utilizing Genome Alignments for Comparative Gene Annotation. Methods Mol Biol. 2019;1962:179‑91. doi:10.1007/978-1-4939-9173-0_10 PubMed PMID: 31020560.

Herrero J, Muffato M, Beal K, Fitzgerald S, Gordon L, Pignatelli M, et al. Ensembl comparative genomics resources. Database J Biol Databases Curation. 2016;2016:bav096. doi:10.1093/database/bav096 PubMed PMID: 26896847; PubMed Central PMCID: PMC4761110.

Wang Y, Tang H, Debarry JD, Tan X, Li J, Wang X, et al. MCScanX: a toolkit for detection and evolutionary analysis of gene synteny and collinearity. Nucleic Acids Res. avr 2012;40(7):e49. doi:10.1093/nar/gkr1293 PubMed PMID: 22217600; PubMed Central PMCID: PMC3326336.

Washietl S, Pedersen JS, Korbel JO, Stocsits C, Gruber AR, Hackermüller J, et al. Structured RNAs in the ENCODE selected regions of the human genome. Genome Res. juin 2007;17(6):852‑64. doi:10.1101/gr.5650707 PubMed PMID: 17568003; PubMed Central PMCID: PMC1891344.

Rivas E. RNA covariation at helix-level resolution for the identification of evolutionarily conserved RNA structure. PLoS Comput Biol. juill 2023;19(7):e1011262. doi:10.1371/journal.pcbi.1011262 PubMed PMID: 37450549; PubMed Central PMCID: PMC10370758.

Katoh K, Rozewicki J, Yamada KD. MAFFT online service: multiple sequence alignment, interactive sequence choice and visualization. Brief Bioinform. 19 juill 2019;20(4):1160‑6. doi:10.1093/bib/bbx108 PubMed PMID: 28968734; PubMed Central PMCID: PMC6781576.

Sievers F, Higgins DG. The Clustal Omega Multiple Alignment Package. Methods Mol Biol. 2021;2231:3‑16. doi:10.1007/978-1-0716-1036-7_1 PubMed PMID: 33289883.

Palmer D, Fabris F, Doherty A, Freitas AA, de Magalhães JP. Ageing transcriptome meta-analysis reveals similarities and differences between key mammalian tissues. Aging. 11 févr 2021;13(3):3313‑41. doi:10.18632/aging.202648 PubMed PMID: 33611312; PubMed Central PMCID: PMC7906136.