Yeast Secretes High Amounts of Human Calreticulin without Cellular Stress

Round 1

Reviewer 1 Report

GENERAL COMMENTS:  

Zinkevičiūtė et al. use multiple separation techniques and mass spectrometry-based proteomics tools to profile recombinant protein expression of human calreticulin (CALR) in Saccharomyces cerevisiae, the Baker's yeast, at high yield (eg. hundreds of mg/L).  

The authors conduct a comparative proteomics study between 2 yeast strains:  i) a strain expressing recombinant CALR vs. 
ii) an empty control vector, pFGG3, apparently created in their laboratory [reference 25].  

A wide range of analytical tools are used in this comparative study, such as non-equilibrium pH gradient electrophoresis (NEPHGE)-based two-dimensional gel electrophoresis (2DE) and liquid chromatography-mass spectrometry (LC-MS).  This study represents a tremendous amount of work to elucidate the secretory mechanism(s) of the CALR protein, which remains unknown in humans.  The fact that the authors were able to replicate a discontinued NEPHAGE reagent is remarkable (and could be the topic of a separate technical paper); however, it is tangential to the study objective.  

The authors claim a secretion titer of 138-140 mg/L CALR, which is a lot of protein.

In reviewing this article, the following questions came to mind:  

01.  Regarding experimental design: rather than an "empty control vector," why would the wild-type strain (eg. not genetically modified) not be used along side the 2 other strains?  Use of the wild-type strain could show what collateral effects the pFGG3 plasmid may have relative to the wild-type proteome.  

02.  The authors go to great length to combine NEPHAGE (two-dimensional gel electrophoresis) with liquid chromatography tandem mass spectrometry (LC-MS/MS), but I am not certain that doing so is absolutely necessary.  If the authors provided more experimental MS detail, more details regarding proteome informatics (such as mass error tolerances searched with, etc.), and made their MS data available in a public repository (such as proteomeXchange/PRIDE), they may not have felt the need to perform so much work.  NOTE: from line 22 of the abstract onward thru the manuscript, "LC-MS" should be changed to "LC-MS/MS" as peptide ions were fragmented in a second MS experiment and searched for identification using via commercial software.  

03.  It is rather burdensome for the reader to ascertain that 2 disparate quantitative measures are used to estimate protein abundance: images from 2DE separations and TOP3, a type of label-free quantification from the MS experiment.  These details should be moved forward in the text-- and perhaps mentioned in the abstract-- to show how much work the authors performed.  I was able to "get" what exactly was done from reviewing Supplementary (or Spreadsheet S2), so refining that in the body of the text would be helpful to other readers.  

04.  The authors need to significantly revise and more clearly restate text from lines 490-496, which reads as follows: 

"Another 152 proteins were undetected in a part of samples due to the loss of peptides. We believe that this loss of peptides happened because of extremely low abundance or extensive overlap of peptides. Generally, the limiting factor in MS is sample preparation, which may be responsible for the missing proteins [84] (the samples were prepared identically as for the 2DE experiment). Low ionic strength and non-ionic detergents in the sample buffer result in hydrophobic and especially membrane, chromatin and ribosome protein loss [68,77]."

At a minimum, I have 2 problems with the "...undetected in a part of the samples due to the loss of peptides..." as:

01.  The TOTAL number of sample handling steps ALSO exacerbates the potential for peptide loss, which would include performing 2DE/NEPHAGE, then cutting out and digesting the gel spot with trypsin, etc.  

02.  When the authors reported peptides/proteins as "not detected" in these lines, I looked for where I could review their mass spectrometry (MS) methods and instrument parameters.  It is plausible that peptides may be present in a complex sample matrix, yet are not subjected to MS/MS fragmentation; and, as a result, are not detected via peptide-spectrum database searches.  A problem that I had was that these instrument parameters are NOT described in the Methods section & that the primary data are not posted to a repository in the public domain.   

At a minimum, the authors should report additional experimental details in the Methods section and resubmit for another round of peer review.  It would be ideal to also post primary data to a public MS data repository as well; however, in some circumstances, I understand that that is not a viable outcome.  


REQUESTED REVISIONS:  

01.  Please change use of "LC-MS" to "LC-MS/MS" to denote the appropriate experimental detail.  

02.  As written, the manuscript is clear and presented in a well-structured manner; however, it should be resubmitted with revisions.  

03.  Selected references need to be updated and/or added to.  While reference 44 was a seminal paper in yeast proteomics, it is >12 years old.  As appropriate, please either add and/or review the contents of the following 2 papers and update relevant citations:

Ho et al.  Unification of Protein Abundance Datasets Yields a Quantitative Saccharomyces cerevisiae Proteome.  Cell Systems.  6:192-205.e3.  28 FEB 2018.  DOI:  https://doi.org/10.1016/j.cels.2017.12.004

Lian et al.  Metabolic Engineering.  50: 85-108.  November 2018.  
DOI: https://doi.org/10.1016/j.ymben.2018.04.011

04.  The experimental design is appropriate to evaluate a "data-driven" hypothesis, which would required additional studies/papers in addition to this manuscript.  Given the results that are presented, it is not clear how the proteins identified from the yeast proteome "fit" with elucidating the secretion mechanism for CALR in either the yeast or human.  

05.  The authors do not provide sufficient detail to reproduce the mass spectrometry (MS) experiments and data subsequent analysis of LC-MS/MS data using the ProteinLynx Global SERVER software (PLGS, version 3.0.1, Waters Corporation).  

In terms of data analysis, it is routine practice to specify the following parameters for peptide-spectrum database searching: parent ion tolerance window (in Daltons or ppm); fragment ion tolerance window (in Daltons or ppm); number of missed tryptic cleavages per peptide; number of dynamic modifications per peptide, etc.  

In addition to an organism's predicted proteome, in this case the "UniprotKB/SwissProt Saccharomyces cerevisiae databases (2020-09-24) were used for protein identification" (lines 232-233), it is also increasingly routine to include the sequences of proteolytic enzymes (eg. trypsin), peptide and protein mass standards (eg. GluFib, which was used for mass lock in this study), as well as contaminant proteins (keratins, wool, etc.) to avoid false discovery/identification of peptides.  As a general practice, the authors should adopt this for future MS studies.  

Granted, this is very detailed and technical information that is MS-centric, it IS essential information required for reproducing the research and/or conducting a similar experiment on a comparable analytical platform.   

Furthermore, when the authors discuss protein abundance estimates between samples, they could more clearly state that protein abundance values are estimated from image intensity (taken from 2D-gels; NEPHAGE) vs. the label-free methods used for quantification from the MS experiment (eg. TOP3 intensity via ISOQuant, described in reference 31).  The distinction between the quantitative metrics used for protein abundance could be moved up in the text (perhaps to the abstract, too) to assist the readers of this article, as both the NEPHAGE/2DE images and TOP3 abundance calculations are, to some extent, complementary information.  Calculating 2 distinct quantitative metric(s), one at the protein level (2DE) and a second at the peptide level (TOP3), reflects a large amount of work the authors report on in this study.  

06.  Figures, tables, and images represent the data appropriately.  Statistical methods described in the text are routinely used for comparisons between experimental groups (eg. DeSeq2), though the latter is primarily used for comparative transcriptomics experiments.  

07.  Conclusions are consistent with evidence and arguments presented in the manuscript.  

08.  Ethics and data availability statements do not seem to be present, though the authors have not declared a conflict of interest.  Regarding data availability, it would have been remarkably helpful for the authors to have placed their raw mass spectrometry (MS) data in a publicly available repository, such as PRIDE, proteomeXchange, or another source in the public domain.  

Collectively, the authors performed a massive amount of work in this study & I would be happy to review this manuscript for a second time provided additional mass spectrometry details (and other revisions requested above) are included.  

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Although the paper addresses a rather interesting question related to the secretion mechanism of CALR, it requires substantial improvements before publishing. The focus of the paper is somewhat unclear as it looks as the authors wanted to obtain the high titer of CALR by expressing the gene in yeast, but then switched to studying the secretory mechanism which is definitely more interesting from the scientific point of view. However, the main problem of the work is that the latter study requires much more scrutiny in terms of several controls that are lacking. Without these controls, conclusions drawn are not justified.

Particular comments regarding the paper are the following:

- names of microorganisms should be written in italic

- lane 58 – what exactly was the control?

- lane 80 – what is the „simple eukaryotic cell“?

- lane 84 – 90 – the construction of the plasmid pFGAL7-CRT should be described in more details. Where exactly and how was the CALR gene inserted?

- the Discussion is generally too long and in a large part repeats results. Basically, the paper is very simple and contains of three parts. First, the 2DE technique has been  worked out using an in-house made mixture of ampholites. Then, the method was used for differential analysis protein patterns of yeasts overexpressing human CALR vs. the yeasts bearing empty vector. As this analysis could not reveal any differentially expressed proteins except superoxide dismutase, the differential analysis was performed using LC-MS. Discussion should follow results but not repeat them.

- A part of Discussion discusses the amount of calreticulin produced by the system used in this work and compares it with the previously used system. The paper, however, does not deal with the production of the protein (no optimization has been attempted, or the study of protein secretion under different conditions, etc.), thus the comparison is not necessarily experimentally sound, and the discussion is rather speculative.

- lanes 472-473 - The statement that „the elevated ROS levels induced the increased expression of SOD1” is a speculation and has not been confirmed in this work. It is particularly unsubstantiated because the regulation of other proteins usually regulated by the presence of ROS have not been observed.

 

- lanes 475 – 487 – this part of the discussion is meaningless since the authors have no results proving that more bands result from protein degradation in the first place and discussing different possibilities of mechanisms of the process that we do not know that exists at all seems senseless.

 

- lanes 521 and further – in this paragraph the authors try to explain differential expression of several proteins by carbon catabolite derepression. However, the growth regime was the same for the control, as well as for the CALR producing strain. Thus, there should be no difference in the proteins reflecting the growth conditions and the carbon source.

 

- lanes 553 and further – the paragraph discusses the lack of differential expression of secretion-related proteins when CALR is overexpressed. For drawing conclusions in this part, a positive control is missing. Authors should include the overexpression of another (control) protein in the same system and check if this would induce differences in the expression of secretion and cell protection proteins. Besides, a control in which a typical yeast secretory protein is expressed in the system the authors used would be required if they want to conclude that CALR behaves as a primarily secretory protein. Without such controls the conclusions drawn are not justified.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This comparative proteomics study (using NEPHGE-2DE and LC-MS) provides a whole list of differentially expressed proteins but unfortunately does not proof that any of these are important for the secretion of calreticulin (CALR). The study was designed in order to gain some insight on how CALR exits human cells using a simple eukaryotic system such as the yeast Saccharomyces cerevisiae, but the authors failed to identify new molecular players. In particular, the authors did not identify any protein involved in the secretory pathway nor any sign of cellular stress when yeast cells are secreting high levels of human recombinant CALR, using both NEPHGE-2D and LC-MS. Based on this, they concluded that CALR secretion in yeast does not induce cellular stress nor any secretory pathway. The authors reconstructed the carrier ampholyte (CA) composition for the NEPHGE-based first-dimension separation, claiming the unavailability of pre-made gel solutions. They compared their system to the commercial WITA gells and failed to show a good correlation between them. Still, the authors were confident about its use for 2DE proteomics.
Perharps, a good positive control (e.g. Pichia pastoris overexpressing xylanase A) could help in validating the system. The discussion is overly long and spends time speculating over how changes in protein expression map to particular pathways. The lack of functional assays to validate many of the pathways described (being SOD1 the only exception), and inclusion of relevant mutants (where possible) to assess their biological importance is a major limitation. Though, the authors were able to achieve the highest reported secretion titer for human CALR (~140 mg/L). The manuscript provides a lot of information for further research. 

Also, species names are not in italics across the manuscript (please see abstract lines 17 and 19; introduction lines 58 and 79; and references).
Abbreviations were not used in a consistent way (e.g. once they are introduced for the first time, they should be used from then on - please see for instance lines 40, 44 and 45). 
First three paragraphs in the “Plasmids, yeast strains, media and growth” section should be thoroughly revised. 
Formatting is also not consistent (please see line 577).
The use of words such as “believe” (line 491, 432 and 542) should be avoided - we are searching for truth, not beliefs. 

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

After further review, the 4th version of this manuscript (file name: cimb-1578965-peer-review-v4.pdf) is significantly improved from the original version that was submitted.  

The authors have appeared to follow suggestions regarding the addition of more experimental detail regarding methods-- particularly molecular biology & cloning & the LC-MS/MS methods-- which are standard practices in proteomics. 

This is shown on Line 232: "Raw data are available via the MassIVE repository with identifier MSV000088879."  By doing this, most MS (or proteomics) labs are able to evaluate the LC-MS/MS measurements, instrument settings, etc. -- even if the authors did not go into granular detail about this in a Supporting Methods document.  

Primary MS data  are available at the following URL:
https://massive.ucsd.edu/ProteoSAFe/dataset.jsp?task=7839799eca504221b2c414bb3713b615  

In addition to being a requirement for publication in this journal, primary MS data availability in a public repository has been a standard practice in the proteomics field for several years.

I hope that the authors remember this in future publications.  

In this Reviewer's position, this article may be accepted for publication.  The authors demonstrated and provided transparency in the more technical parts of the study from which their conclusions were based.  

Reviewer 2 Report

Although the authors have improved their manuscript in many technical details there are still several major problems I see. As previously mentioned the paper first focuses on the methodology (2D) but it does not bring any real novelty in this part and actually shows that the electrophoretic method was not appropriate for the case they describe. In a way, if this part were omitted it would not influence the scientific merit of the paper. In the second part that was in my opinion more important, there are some serious methodological flaws and important controls missing. For instance, it is from the methodological point incorrect to use experiments performed in another study and published in another paper as a control for experiments described here. Therefore, unfortunately I still cannot recommend this paper for publishing.

Author Response

Dear reviewer,

We noticed that the round 2 reviews are mixed up, and we are not sure which reviewer we are answering.

We agree that our claim that CALR behaves as a typical secretory protein was overstated. We also agree that this claim in the future should be more substantiated by further work including additional controls as reviewers suggested. However, we updated the manuscript and changed the conclusions. Nevertheless, we found that CALR secretion does not induce cellular stress and changes the proteome very little. We have proven this by quantitating the proteomes of the studied yeast cells on the protein level by NEPHGE-based 2DE (mostly high-abundant proteins) and on the peptide level by LC-MS^E. We also added a functional enrichment analysis of differentially expressed proteins identified by  LC-MS^E that supports our findings. This work not only sets the background for further studies in CALR secretion but also gives a look inside a phenomenon of stress-free production of recombinant proteins in yeast S. cerevisiae.

We invite the reviewers to reconsider our updated manuscript.

Reviewer 3 Report

The manuscript has been improved and some points clarified. However, some controls are still missing in order to sustain the conclusions of the study.

Author Response

Dear reviewer,

We noticed that the round 2 reviews are mixed up, and we are not sure which reviewer we are answering.

We agree that our claim that CALR behaves as a typical secretory protein was overstated. We also agree that this claim in the future should be more substantiated by further work including additional controls as reviewers suggested. However, we updated the manuscript and changed the conclusions. Nevertheless, we found that CALR secretion does not induce cellular stress and changes the proteome very little. We have proven this by quantitating the proteomes of the studied yeast cells on the protein level by NEPHGE-based 2DE (mostly high-abundant proteins) and on the peptide level by LC-MS^E. We also added a functional enrichment analysis of differentially expressed proteins identified by  LC-MS^E that supports our findings. This work not only sets the background for further studies in CALR secretion but also gives a look inside a phenomenon of stress-free production of recombinant proteins in yeast S. cerevisiae.

We invite the reviewers to reconsider our updated manuscript.