A modular lentiviral system for multiplexed gene perturbation and functional analysis suggests interdependence of hormone receptors in breast cancer growth in vivo

Article activity feed

1. 

   Review Commons

   Aug 5, 2026

   Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

   Learn more at Review Commons

   ---

   Reply to the reviewers

   Reviewer #1 (Evidence, reproducibility and clarity (Required)):

   Summary: The manuscript describes a modular, doxycycline-inducible lentiviral vector platform that enables conditional overexpression, RNAi-mediated knockdown, and CRISPR-based perturbations, combined with fluorescent and luminescent reporters for multiplexed tracking of cell populations. Using this system, the authors perform pooled, competitive in vitro and in vivo assays, focusing on hormone receptor dependencies (ER, PR, AR) in MCF7 breast cancer cells. The key biological conclusion is that AR depletion has minimal effects in vitro but significantly impairs growth in vivo, and that …

   Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

   Learn more at Review Commons

   ---

   Reply to the reviewers

   Reviewer #1 (Evidence, reproducibility and clarity (Required)):

   Summary: The manuscript describes a modular, doxycycline-inducible lentiviral vector platform that enables conditional overexpression, RNAi-mediated knockdown, and CRISPR-based perturbations, combined with fluorescent and luminescent reporters for multiplexed tracking of cell populations. Using this system, the authors perform pooled, competitive in vitro and in vivo assays, focusing on hormone receptor dependencies (ER, PR, AR) in MCF7 breast cancer cells. The key biological conclusion is that AR depletion has minimal effects in vitro but significantly impairs growth in vivo, and that combined hormone receptor knockdown leads to synergistic growth suppression in a xenograft model.

   Major comments:

   1. Strength of evidence supporting the key conclusions: The technical demonstration of the vector platform is generally convincing, particularly the modular design and the feasibility of multiplexed fluorescent tracking. However, the biological conclusions are only partially supported by the data. The claim that hormone receptor interdependence, and in particular AR dependence, is revealed specifically in vivo rests on a single cell line (MCF7), a limited number of animals, and a small set of shRNAs-some of which appear to have weak or no functional impact in vitro. As such, the conclusions should be clearly qualified as preliminary and context-specific, rather than presented as generalizable insights into hormone receptor biology. In particular, the strong concluding statements (final paragraph of the manuscript) should be toned down to reflect: the limited number of models tested, the absence of mechanistic insight, and the reliance on RNAi-based perturbations without rescue experiments.

   We thank the reviewer for this important point. We agree that reliance on a single cell line, limited animal numbers, and possible variability in shRNA performance warranted more cautious framing of the biological conclusions. We have added an explicit caveat paragraph before the closing summary stating that these findings should be regarded as proof of concept, generated in a single breast cancer cell line (MCF7) using a limited number of animals and a small panel of shRNAs whose individual potency was not exhaustively benchmarked. The AR-dependency finding, in particular, is based on RNA interference alone, without orthogonal rescue experiments or mechanistic follow-up, and should be interpreted as preliminary and context-specific rather than as a generalizable feature of AR biology. We have also softened the closing sentence of the manuscript from “accelerates the preclinical development...” to “may help inform future preclinical strategies for hormone-sensitive breast cancer and beyond, pending validation in additional models.”

   1. Claims that require qualification or revision: Several claims appear overstated relative to the data provided: a. The assertion that the system enables robust temporal control of perturbations is not supported by quantitative data on leakiness, induction kinetics, or stability of editing over time.

   Thank you for raising these points. To address the concerns, we repeated the experiments, including time-course analyses and DNA sequencing with the inducible CAS9 and qRNA, and now present the data in Figure 2. These results show the induction kinetics of the system and address the leakiness of the double-inducible CRISPR/CAS9 system. Because DNA editing is a permanent change to the DNA, we assume that the stability of the edit over time will also be permanent (see Figure 2 and Supplemental Table 2).

   The claim that pooled multiplexed perturbation "reveals" discrepancies between in vitro and in vivo AR function is based on a narrow experimental scope and should be reframed as an illustrative example rather than a definitive finding. Thank you for pointing this out. We have revised and softened the claims by removing “reveals” and replacing it with “suggests” throughout the manuscript.

   Statements implying reduced off-target effects due to temporal regulation are speculative and should be removed unless supported by data. Thank you for pointing this out. We have removed those parts from the manuscript and revised the relevant sentences to avoid speculation about off-target effects, which are largely determined by the design of gRNAs and shRNAs.

   Additional experiments essential to support the paper (limited and realistic): Only minimal additional experiments are required to support the manuscript as it stands: a. Quantification of inducibility and leakiness of the dual Tet CRISPR system (e.g., untreated vs. dox-treated control in Fig. 2C, and time-course of editing efficiency).

   Thank you for the suggestion. We have now performed this experiment and included the data (see the new Figure 2G).

   Quantification of FUCCI reporter outputs (cell-cycle phase distributions) rather than representative images alone. Thank you for the suggestion. We repeated the experiment and quantified the FUCCI reporter output (see Figure 3).

   Reproducibility and methodological clarity: Several aspects of the methodology require clarification to ensure reproducibility: a. Lentiviral titers and recombination rates are not reported. Given the complexity and size of the constructs, this information is essential. We have added more detailed descriptions of the experimental procedures to the Methods section and have repeated the Cas9 transduction experiments, including their corresponding results and viral titers (see Figure 2 and Supplementary Figure 6).

   It is unclear how background editing is prevented in the dual Tet CRISPR system, since both Cas9 and gRNA are present in the same cells and may exhibit basal expression. We use serum from South America, where standard antibiotic treatment is less common than in the United States, reducing the likelihood that doxycycline is present and minimizing basal expression. The rationale for using double-inducible systems is to minimize basal expression of both components and thereby reduce the chance that either one reaches levels sufficient for editing. We have also made the underlying design logic explicit in the manuscript: because unwanted editing requires both Cas9 and gRNA to be simultaneously present above a functional threshold, and each is independently repressed through a distinct tetracycline-responsive mechanism (Tet-On for Cas9, Tet-Off for gRNA), the probability of coincident leaky expression of both components is substantially lower than for either component alone. This is now linked to the empirically measured background editing rate (~4% over 11 days without doxycycline, by ONT sequencing; Fig. 2G).

   The manuscript does not adequately address integration-based leakiness of doxycycline-inducible systems in lentiviral backbones, especially compared to transposon-based approaches. Thanks for this point. While we have not directly compared our system with transposon-based approaches, we have added a brief section on this in the Discussion to address this point explicitly. Because our system relies on polyclonal, antibiotic- or FACS-selected populations rather than single-cell-validated clones, some degree of integration site–dependent leakiness in the noninduced state cannot be fully excluded and may contribute to the low background editing (~4%) we observe over extended culture. We contrast this with transposon-based delivery systems (piggyBac, Sleeping Beauty), which have distinct genomic integration profiles relative to lentivirus, and note that doxycycline-inducible piggyBac constructs have, in some contexts, achieved tighter regulation with undetectable basal leakiness in vivo. This may reflect the absence of viral LTR-associated regulatory elements and the lentiviral bias toward active chromatin. We discuss chromatin insulators, safe-harbor-targeted integration, and transposon-based delivery as potential strategies to further reduce leaky background expression in future iterations of the platform.

   The description of how MCF7-luciferase cells are used to generate lentiviral vectors is confusing and must be clarified. We have updated this in the revised manuscript and hope it is clearer now.

   Replication and statistical analysis: The statistical treatment of pooled competition data is insufficiently detailed. It is unclear how many animals, glands, or technical replicates contribute to each comparison. The manuscript does not clearly report the fraction of cells recovered for single, double, and triple knockdowns. Multiple comparisons and normalization strategies are not consistently explained. We apologize if these sections were difficult to follow and have substantially expanded them in the manuscript. Because a recurring concern with any pooled, barcoded competition assay is that the fluorescent tag, lentiviral integration site, or clonal origin of a population could itself influence engraftment or proliferation independently of the intended genetic perturbation, we now normalize each doxycycline-treated population to its own untreated (−Dox) baseline and then express it relative to the equivalently normalized NT population within the same tumor (Fig. 5C, F). We also report the numbers of animals and glands contributing to each comparison directly in the Fig. 5 legend, and we justify this normalization approach in more detail in a new paragraph in the Discussion.

   Minor comments: a. Comparison to existing Gateway-compatible systems (e.g. the John Doench system) would help contextualize the technical advance.

   Thank you. We have now added a paragraph to the Discussion that compares our platform with existing Gateway-based and CRISPR/Cas9 lentiviral resources, including the Broad Institute's Genetic Perturbation Platform co-developed by Doench and colleagues (Brunello, Dolcetto, Calabrese). This paragraph clarifies that our system is complementary to these genome-scale discovery tools rather than competing with them: it is designed for hypothesis-driven, combinatorial, multiplexed perturbation studies with tight temporal control and validated ex vivo and in vivo applications, rather than for large-scale single-modality screening.

   Page 5, line 6: "The ..." should be "It ...". Thank you for the comment and we have corrected this. c. P2A sequences are not self-cleaving; the manuscript should correctly state that the ribosome skips peptide bond formation between the last two amino acids.

   Thank you for the comment and we have corrected this.

   Fig. 2C lacks an untreated control. Thank you for this. Instead of using an untreated control, we used two different guides targeting USP14 and USP7 to validate the inducible gRNA with constitutively expressed Cas9. For the knockdown controls, each gene served as the control for the other in the corresponding experiments, demonstrating that knockdown could be achieved with the specific inducible gRNA.

   Fig. 2E (AkaLuc panel): the label "Reagent" is incorrect and overly broad; a clearer and more consistent nomenclature is needed. Thank you for bringing this to our attention; we have corrected it in the revised manuscript. The panel has also been moved and now appears as Fig. 3B, with the axis relabeled “Ctrl”/“+Substrate” in place of “Reagent.” The purpose of fluorescence-based tracking of cell populations is not clearly explained; the rationale should be explicitly stated. We have now added an explicit rationale at the start of this Results subsection: multiplexed fluorescence barcoding is used to pool several genetically distinct cell populations and track them side by side within the same well, culture, or animal. This allows different genotypes to be compared under identical experimental conditions rather than across separate parallel experiments. This internally controlled design reduces confounding variability arising from well-to-well, batch-to-batch, or animal-to-animal differences and increases the statistical power obtained per experiment. It is particularly valuable in vivo, where it substantially reduces the number of animals required, in keeping with the 3Rs principles.

   Claims that hormone supplementation rescues AR depletion are not supported, particularly given the lack of a clear AR-dependent phenotype in prior assays (e.g. Fig. S3A). Thank you for bringing this up. We have rewritten this section and clarified that it behaves more like the non-targeting control shRNA (NT).

   The large difference in proliferation between NT cells {plus minus} doxycycline in Fig. 4B is concerning and may reflect a normalization or analysis error; this should be addressed. Thank you for pointing this out. We have now included additional raw data for clarity and addressed the issue accordingly, so it should not be misinterpreted in the future. Figures would benefit from clearer legends specifying n, statistical tests, and normalization procedures. Thank you for pointing this out; we have addressed it in the revised manuscript. Reviewer #1 (Significance (Required)):

   Nature and significance of the advance: This work represents a technical advance, rather than a conceptual or biological one. The modular lentiviral platform for inducible perturbations and multiplexed fluorescent tracking is potentially useful, particularly for pooled in vivo competition assays where reducing animal use is desirable.

   Context within existing literature: Inducible lentiviral shRNA and CRISPR systems, as well as fluorescent barcoding strategies, are well established. The main contribution here is the integration of these elements into a single, modular framework. However, the manuscript would benefit from clearer comparison to existing systems (including Gateway-based and inducible CRISPR platforms) and discussion of known limitations of lentiviral Tet systems.

   Audience: The work will primarily interest I) cancer biologists performing functional genetic screens, ii) researchers developing or applying genetic perturbation tools, and iii) laboratories interested in pooled in vivo assays. The biological findings regarding hormone receptor interdependence are likely of more limited interest unless further validated.

   Reviewer expertise: My expertise includes functional cancer genomics, lentiviral and CRISPR-based perturbation systems, in vitro and in vivo genetic screening approaches, as well as bioinformatics. I have sufficient expertise to evaluate the technical platform and biological conclusions presented.

   Reviewer #2 (Evidence, reproducibility and clarity (Required)):

   Summary This study presents a novel, modular lentiviral platform integrating inducible overexpression, shRNA knockdown, and CRISPR/Cas9 editing within a single system. Its core innovation lies in the versatile design, utilizing Gateway cloning to enable rapid exchange of 14 fluorescent proteins, 3 luminescent reporters, and selection markers, facilitating high-content phenotyping. A dual doxycycline-inducible architecture ensures precise temporal control, minimizing off-target effects. Methodologically, it introduces a robust fluorescence barcoding strategy, allowing multiplexed tracking of up to nine distinct genetic perturbations in pooled assays. When applied to breast cancer intraductal xenografts, this revealed critical in vivo receptor interdependencies-such as the essential roles of AR, ER and PR, in tumor growth. By enabling combinatorial genetic analysis within single animals, this technology significantly reduces experimental variability and animal usage by up to eightfold, advancing both the efficiency of functional genomics and adherence to ethical research standards.

   Major comments

   1. Despite using an inducible system, shRNA and CRISPR components may still have off-target effects. In the F4 study, was whole-genome sequencing or transcriptomic analysis performed to assess unintended perturbations of non-target genes?

   Because we did not aim to conduct detailed mechanistic studies or present shRNA as a new technology, and because the hairpins used are not novel and have already been described in published studies, we did not attempt to identify or test potential off-target genes in this work.

   Barcode stability: Are the fluorescent barcodes stably expressed during long-term in vivo culture? Is there a risk of silencing or loss that could affect the reliability of long-term tracking?

   The in vivo studies we conducted lasted more than 30 days. The cells were first validated by qPCR and Western blot, generating transgenic cells that were then expanded for all replicates and validations, including the in vivo experiments. The barcodes are driven by the EF1α promoter, which is not expected to be susceptible to methylation and, in theory, should support expression in long-term in vivo studies even beyond the 30-day duration used here.

   Cell-cell interference: In mixed transplantation experiments, could different genotypes influence each other through paracrine signaling or competition for resources, leading to observed growth phenotypes that are not entirely cell-autonomous?

   Thank you for raising this important point. We observed similar ex vivo effects on growth reduction in single experiments (except for AR in vitro) as we did in vivo with the mixed population. In all in vivo studies, whether mixed or single, the cancer cells injected into a mouse are never cell-autonomous. That is also why one performs in vivo studies using genetic perturbations: to demonstrate that cellular dependencies on genes are not in vitro artifacts, to show that certain genes are important in an in vivo setting, and to reveal how a gene affects the microenvironment in vivo, not just ex vivo. Because tumors are inherently heterogeneous, one could also argue that a mixture of different genotypes allows us to study tumor evolution with greater insight into different gene losses.

   Applicability to non-coding genes or weak-effect genes: This platform relies on observable phenotypic changes for screening. Is it sensitive enough for genes with weak effects or functional redundancy in regulatory networks?

   Thank you for the interesting question. In theory, the approach should be sensitive enough to use with genes that do not cause a growth defect. The functional readouts should then be further tailored to this purpose, using reporter assays, biomarker assays, or alternative sequencing-based readouts to study different gene-specific consequences. The barcodes can still be used to separate cells by flow cytometry and then to analyze them with the assay of choice.

   Are there differences in the induction efficiency of different shRNA or CRISPR components? Could this lead to biases in certain barcode signals, thereby affecting the accuracy of competitive growth analysis?

   Thank you for the question. We did not detect any induction efficiency effects that would bias the barcoding, as we used hairpins and gRNA guides that have been validated or previously used in the literature. However, this does not mean such bias could not occur in some cases.

   Because we also use a “no doxycycline” control, one can not only study the intercell grafting efficacy of any transgenic cell but also estimate the theoretical growth based on NT and non-dox samples. This can normalize any calculations if such bias affects the dynamics of the hairpin or gRNA. Inclusion of the no-dox control accounts for this directly: each doxycycline-treated population is normalized to its own untreated (−Dox) baseline before being expressed relative to the similarly normalized NT population within the same tumor (Fig. 5C, F). We have now implemented and reported this normalization in the manuscript, with the rationale explained in a new Discussion paragraph, rather than treating it only as a theoretical mitigation.

   Are there variations in the packaging efficiency of different shRNA or CRISPR components into lentiviral vectors? Does this affect viral titer? How is copy number consistency achieved in cells after lentiviral transduction? Has the knockdown efficiency been compared between cells transduced with 3 shRNAs and those with single shRNA? Thank you for the questions. For the shRNA and gRNA transductions, the plasmids are approximately the same size, and we have not observed any differences in transduction efficiency between these vectors. Purity and other factors during vector isolation usually have a more pronounced effect on transduction efficiency.

   The vectors that are difficult to transduce are those encoding CAS9 itself; in particular, inducible CAS9 is more challenging to transduce, and achieving a good viral titer is important for efficient transduction. We repeated an experiment and transduced the cells with different viral titers and found that the number of cells surviving antibiotic selection correlated with titer, but after expansion they expressed similar levels of CAS9.

   Since we do not use monoclonal cells and instead work with polyclonal transgenic lines, we assume that random integration should occur in a similar way as in the NT control, and therefore we did not karyotype the cells or analyze copy numbers. The knockdown efficiency was similar when comparing single and triple knockdown. We have now incorporated this explanation directly into the Methods section of the manuscript, stating explicitly that all shRNA and gRNA expression plasmids are of comparable size with no consistent differences in packaging or transduction efficiency, that titer is more strongly influenced by plasmid purity and preparation than by insert identity, and that knockdown efficiency assessed by Western blot was comparable between single, double, and triple knockdown lines (Fig. 4E–G), indicating that combinatorial transduction with multiple shRNAs did not measurably compromise silencing efficiency per target.

   Minor comments

   1. The labeling "NT" in Data F3 and Supplementary Data SF3 is unclear. Do they refer to the same condition? Is DOX added in the "NT" condition in SF3? Thank you for pointing this out. NT refers to non-targeting shRNA cells, and they are treated the same as the other cells in all panels. We have updated the figure legend to clarify this.

   Cost-effectiveness ratio: Although animal use is reduced, is the cost of constructing the multiplexed barcoded viral library significantly higher than traditional methods? Is its overall economic feasibility suitable for large-scale screening? The cost of creating any shRNA or gRNA is only negligibly higher when generating different guides or hairpins and producing them with multiple barcodes, especially compared with in vivo study and animal costs. That is why we based our calculations not on construct costs but on animal costs and, more importantly, on the reduction in the number of animals needed.

   For large-scale screening, the system can distinguish only among nine barcodes, gene targets, and all their combinations, so it is not currently designed for settings in which more than nine genes are targeted.

   Reviewer #2 (Significance (Required)):

   The most important significance of the study is integrating multiple technologies into one system enabling high-content phenotyping, which may facilitate the discovery of new biological pathways.

   Read the original source
2. 

   Review Commons

   Aug 5, 2026

   Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

   Learn more at Review Commons

   ---

   Referee #2

   Evidence, reproducibility and clarity

   Summary

   This study presents a novel, modular lentiviral platform integrating inducible overexpression, shRNA knockdown, and CRISPR/Cas9 editing within a single system. Its core innovation lies in the versatile design, utilizing Gateway cloning to enable rapid exchange of 14 fluorescent proteins, 3 luminescent reporters, and selection markers, facilitating high-content phenotyping. A dual doxycycline-inducible architecture ensures precise temporal control, minimizing off-target effects. Methodologically, it introduces a robust fluorescence barcoding strategy, allowing multiplexed tracking of up to nine distinct genetic perturbations …

   Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

   Learn more at Review Commons

   ---

   Referee #2

   Evidence, reproducibility and clarity

   Summary

   This study presents a novel, modular lentiviral platform integrating inducible overexpression, shRNA knockdown, and CRISPR/Cas9 editing within a single system. Its core innovation lies in the versatile design, utilizing Gateway cloning to enable rapid exchange of 14 fluorescent proteins, 3 luminescent reporters, and selection markers, facilitating high-content phenotyping. A dual doxycycline-inducible architecture ensures precise temporal control, minimizing off-target effects. Methodologically, it introduces a robust fluorescence barcoding strategy, allowing multiplexed tracking of up to nine distinct genetic perturbations in pooled assays. When applied to breast cancer intraductal xenografts, this revealed critical in vivo receptor interdependencies-such as the essential roles of AR, ER and PR, in tumor growth. By enabling combinatorial genetic analysis within single animals, this technology significantly reduces experimental variability and animal usage by up to eightfold, advancing both the efficiency of functional genomics and adherence to ethical research standards.

   Major comments

   1. Despite using an inducible system, shRNA and CRISPR components may still have off-target effects. In the F4 study, was whole-genome sequencing or transcriptomic analysis performed to assess unintended perturbations of non-target genes?
   2. Barcode stability: Are the fluorescent barcodes stably expressed during long-term in vivo culture? Is there a risk of silencing or loss that could affect the reliability of long-term tracking?
   3. Cell-cell interference: In mixed transplantation experiments, could different genotypes influence each other through paracrine signaling or competition for resources, leading to observed growth phenotypes that are not entirely cell-autonomous?
   4. Applicability to non-coding genes or weak-effect genes: This platform relies on observable phenotypic changes for screening. Is it sensitive enough for genes with weak effects or functional redundancy in regulatory networks?
   5. Are there differences in the induction efficiency of different shRNA or CRISPR components? Could this lead to biases in certain barcode signals, thereby affecting the accuracy of competitive growth analysis?
   6. Are there variations in the packaging efficiency of different shRNA or CRISPR components into lentiviral vectors? Does this affect viral titer? How is copy number consistency achieved in cells after lentiviral transduction? Has the knockdown efficiency been compared between cells transduced with 3 shRNAs and those with single shRNA?

   Minor comments

   1. The labeling "NT" in Data F3 and Supplementary Data SF3 is unclear. Do they refer to the same condition? Is DOX added in the "NT" condition in SF3?
   2. Cost-effectiveness ratio: Although animal use is reduced, is the cost of constructing the multiplexed barcoded viral library significantly higher than traditional methods? Is its overall economic feasibility suitable for large-scale screening?

   Significance

   The most important significance of the study is integrating multiple technologies into one system enabling high-content phenotyping, which may facilitate the discovery of new biological pathways.

   Read the original source
3. 

   Review Commons

   Aug 5, 2026

   Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

   Learn more at Review Commons

   ---

   Referee #1

   Evidence, reproducibility and clarity

   Summary:

   The manuscript describes a modular, doxycycline-inducible lentiviral vector platform that enables conditional overexpression, RNAi-mediated knockdown, and CRISPR-based perturbations, combined with fluorescent and luminescent reporters for multiplexed tracking of cell populations. Using this system, the authors perform pooled, competitive in vitro and in vivo assays, focusing on hormone receptor dependencies (ER, PR, AR) in MCF7 breast cancer cells. The key biological conclusion is that AR depletion has minimal effects in vitro but significantly impairs growth in vivo, and that combined hormone receptor knockdown leads to …

   Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

   Learn more at Review Commons

   ---

   Referee #1

   Evidence, reproducibility and clarity

   Summary:

   The manuscript describes a modular, doxycycline-inducible lentiviral vector platform that enables conditional overexpression, RNAi-mediated knockdown, and CRISPR-based perturbations, combined with fluorescent and luminescent reporters for multiplexed tracking of cell populations. Using this system, the authors perform pooled, competitive in vitro and in vivo assays, focusing on hormone receptor dependencies (ER, PR, AR) in MCF7 breast cancer cells. The key biological conclusion is that AR depletion has minimal effects in vitro but significantly impairs growth in vivo, and that combined hormone receptor knockdown leads to synergistic growth suppression in a xenograft model.

   Major comments:

   1. Strength of evidence supporting the key conclusions: The technical demonstration of the vector platform is generally convincing, particularly the modular design and the feasibility of multiplexed fluorescent tracking. However, the biological conclusions are only partially supported by the data. The claim that hormone receptor interdependence, and in particular AR dependence, is revealed specifically in vivo rests on a single cell line (MCF7), a limited number of animals, and a small set of shRNAs-some of which appear to have weak or no functional impact in vitro. As such, the conclusions should be clearly qualified as preliminary and context-specific, rather than presented as generalizable insights into hormone receptor biology. In particular, the strong concluding statements (final paragraph of the manuscript) should be toned down to reflect: the limited number of models tested, the absence of mechanistic insight, and the reliance on RNAi-based perturbations without rescue experiments.
   2. Claims that require qualification or revision: Several claims appear overstated relative to the data provided:

   a. The assertion that the system enables robust temporal control of perturbations is not supported by quantitative data on leakiness, induction kinetics, or stability of editing over time.

   b. The claim that pooled multiplexed perturbation "reveals" discrepancies between in vitro and in vivo AR function is based on a narrow experimental scope and should be reframed as an illustrative example rather than a definitive finding.

   c. Statements implying reduced off-target effects due to temporal regulation are speculative and should be removed unless supported by data.

   1. Additional experiments essential to support the paper (limited and realistic): Only minimal additional experiments are required to support the manuscript as it stands:

   a. Quantification of inducibility and leakiness of the dual Tet CRISPR system (e.g., untreated vs. dox-treated control in Fig. 2C, and time-course of editing efficiency).

   b. Quantification of FUCCI reporter outputs (cell-cycle phase distributions) rather than representative images alone.

   1. Reproducibility and methodological clarity: Several aspects of the methodology require clarification to ensure reproducibility:

   a. Lentiviral titers and recombination rates are not reported. Given the complexity and size of the constructs, this information is essential.

   b. It is unclear how background editing is prevented in the dual Tet CRISPR system, since both Cas9 and gRNA are present in the same cells and may exhibit basal expression.

   c. The manuscript does not adequately address integration-based leakiness of doxycycline-inducible systems in lentiviral backbones, especially compared to transposon-based approaches.

   d. The description of how MCF7-luciferase cells are used to generate lentiviral vectors is confusing and must be clarified.

   1. Replication and statistical analysis: The statistical treatment of pooled competition data is insufficiently detailed. It is unclear how many animals, glands, or technical replicates contribute to each comparison. The manuscript does not clearly report the fraction of cells recovered for single, double, and triple knockdowns. Multiple comparisons and normalization strategies are not consistently explained.

   Minor comments:

   a. Comparison to existing Gateway-compatible systems (e.g. the John Doench system) would help contextualize the technical advance.

   b. Page 5, line 6: "The ..." should be "It ...".

   c. P2A sequences are not self-cleaving; the manuscript should correctly state that the ribosome skips peptide bond formation between the last two amino acids.

   d. Fig. 2C lacks an untreated control.

   e. Fig. 2E (AkaLuc panel): the label "Reagent" is incorrect and overly broad; a clearer and more consistent nomenclature is needed.

   f. The purpose of fluorescence-based tracking of cell populations is not clearly explained; the rationale should be explicitly stated.

   g. Claims that hormone supplementation rescues AR depletion are not supported, particularly given the lack of a clear AR-dependent phenotype in prior assays (e.g. Fig. S3A).

   h. The large difference in proliferation between NT cells {plus minus} doxycycline in Fig. 4B is concerning and may reflect a normalization or analysis error; this should be addressed.

   i. Figures would benefit from clearer legends specifying n, statistical tests, and normalization procedures.

   Significance

   Nature and significance of the advance: This work represents a technical advance, rather than a conceptual or biological one. The modular lentiviral platform for inducible perturbations and multiplexed fluorescent tracking is potentially useful, particularly for pooled in vivo competition assays where reducing animal use is desirable.

   Context within existing literature: Inducible lentiviral shRNA and CRISPR systems, as well as fluorescent barcoding strategies, are well established. The main contribution here is the integration of these elements into a single, modular framework. However, the manuscript would benefit from clearer comparison to existing systems (including Gateway-based and inducible CRISPR platforms) and discussion of known limitations of lentiviral Tet systems.

   Audience: The work will primarily interest I) cancer biologists performing functional genetic screens, ii) researchers developing or applying genetic perturbation tools, and iii) laboratories interested in pooled in vivo assays. The biological findings regarding hormone receptor interdependence are likely of more limited interest unless further validated.

   Reviewer expertise: My expertise includes functional cancer genomics, lentiviral and CRISPR-based perturbation systems, in vitro and in vivo genetic screening approaches, as well as bioinformatics. I have sufficient expertise to evaluate the technical platform and biological conclusions presented.

   Read the original source
4. 

   Version published to 10.64898/2025.12.07.692832 on bioRxiv

   Dec 9, 2025