Identification of HIR, EDS1 and PAD4 Genes Reveals Differences between Coffea Species That May Impact Disease Resistance

Round 1

Reviewer 1 Report

In this manuscript, the HIR, EDS1, and PAD4 gene families in the three Coffea genomes were compared by the authors. Additionally, they compared the HIR, EDS1, and PAD4 genes' levels of gene expression following Hv/Uv inoculation. The phylogenetic analysis and gene structure sections are both quite thorough and easy to understand. However, I do have a few recommendations for the qRT-PCR analysis part.

1.     In figure 1A, the authors show the levels of gene expression for the HIR, EDS1, and PAD4 genes using the log₂ fold change. The region between (-1,1) has an expression level change that is less than 2 times, making certain conclusions useless, especially when some error bars in the y axis are longer than 1.

2.     In figure 1A, PAD4 gene expression levels are shown to increase at 6 hours after Hv inoculation, decrease at 12 hours, and then increase once more at 24 hours. How can this expression level change be explained? Can the authors test additional time points to determine what transpired from 6 to 24 hai?

3.     Could the authors try the same time points as the Hv inoculation for the Uv inoculation in Figure 1A?

4.     Would you kindly utilize fold change to illustrate the variations in gene expression level in Figure 2A?

 

5.     Do you have images taken under a microscope that depict the same conditions as the drawing images for Figures 1B and 2C?

Author Response

We are thankful to the reviewer for the careful revision that allows us to improve the final manuscript. Below we present answers to the questions raised.

In this manuscript, the HIR, EDS1, and PAD4 gene families in the three Coffea genomes were compared by the authors. Additionally, they compared the HIR, EDS1, and PAD4 genes' levels of gene expression following Hv/Uv inoculation. The phylogenetic analysis and gene structure sections are both quite thorough and easy to understand. However, I do have a few recommendations for the qRT-PCR analysis part.

1. In figure 1A, the authors show the levels of gene expression for the HIR, EDS1, and PAD4 genes using the log₂fold change. The region between (-1,1) has an expression level change that is less than 2 times, making certain conclusions useless, especially when some error bars in the y axis are longer than 1.

We only considered as upregulated the genes HIR4, PAD4 and EDS1. All genes presented a minimum value above 1, although we recognize that the standard deviation values are quite high which is a characteristic of our interaction since Hemileia vastatrix infection is quite asynchronous.  

2. In figure 1A, PAD4 gene expression levels are shown to increase at 6 hours after Hv inoculation, decrease at 12 hours, and then increase once more at 24 hours. How can this expression level change be explained? Can the authors test additional time points to determine what transpired from 6 to 24 hai?

Yes it is true, this pattern is quite interesting, an upregulation that 6 hour later is voided, and 12 hours is almost completely restored. However such a bimodal pattern has been observed before for this species for two different genes CaRLK and CaLOX13 (see Diniz et al. (2012)), We should also point out that we observe the same pattern for the CIFC H469/16 genotype.

3. Could the authors try the same time points as the Hv inoculation for the Uv inoculation in Figure 1A?

We could have tested 6, 12, and 24 hours after inoculation for both fungi, but the timeline considered is due to the fact that by 24hai the data obtained would not be informative for the non-host interaction since at this time the fungal structures are almost completely aborted or dead.

4. Would you kindly utilize fold change to illustrate the variations in gene expression level in Figure 2A?

We could compare it to the initial data point (18 hours after inoculation), and we did those charts, but since we wanted to show how the transcripts of the genes changed along the infection process we decided to use transcript abundance (normalized to the reference genes) instead.

5. Do you have images taken under a microscope that depict the same conditions as the drawing images for Figures 1B and 2C?

Previously we have published several pictures of the infection process in HDT and other compatible and incompatible interaction genotypes (Silva et al 1999; Silva et al 2002, Diniz et al 2012), and since the sampling used in this manuscript did not deviate from what was described before, we do not think it is necessary to show more pictures of the infection process. Instead, we include some schematic images of the infection process to help readers to relate gene expression data with the evolution of the infection process.

 

 

 

Reviewer 2 Report

After revision of the manuscript “Identification of HIR, EDS1 and PAD4 Genes in Coffea spp. reveals differences between Coffea species that can impact its resistance to diseases” I recommend publishing with major reviews. The specific comments and suggestions are listed below:

The manuscript presents relevant and interesting information, but it needs several writing reformulations, additional analysis and careful discussion of the results.

The aim of the work was to analyze gene expression related to host and non-host resistance in the hybrid Coffee genome HDT832/2, and find common genes that may be part of both types of resistance. Then, it compares these genes in coffee genomes and discusses possible implications for resistance. The research is made by a transcriptome analysis and two gene families are chosen: HIR and EDS1. The gene expression profiles in non-host and host responses are compared and results are confirmed by qRT-PCR. Also, the expression of the chosen genes is evaluated in four genotypes of Coffea arabica that show resistance or susceptibility to Hemileia vastatrix and the scenarios are compared. Finally, the studied genes are localized in chromosomes of the three available complete genomes of C. arabica, C. canephora and C. eugenoides and gene structure, transcripts and the primary organization of the predict proteins are compared. Conserved domains are analyzed and variations are discussed.

Major consideration and reviews:

The work starts with a transcriptome analysis, but no result of this analysis is cited or shown. I believe the authors are preparing a second manuscript, but it would be interesting to show or refer to what else they have found.

Then, the confirmation of gene overexpression was made by qRT-PCR, together with the study of the response against Uromyces vignae (data shown in Figure 1). This results need to be better analyzed. Statistical analysis to compare results is essential. It may be done by ANOVA followed by mean tests, for example. Also, it is not clear if biological replicates were performed, how many plants or leafs were used per sample and how many samples were used in each experiment (it must be included in materials and methods). It is mandatory to have replicates and biological repetitions. The same is missing in gene expression analysis in the C. arabica genomes (Data shown in Figure 2). The meaning of errors bars must also be written in figure legends

About methodology, we find in line 149 the session 2.2. “Light microscope observation of fresh tissues” but no image or data is clearly shown during results presentation. It would be interesting to include that as a supplemental material. Also, in line 358 it is described the disease progression and symptoms. It would be interesting to provide the pictures of the processes.

In figure 4, the authors show the genes location in the three species chromosomes. My suggestion is to show only the ones that have the studied genes, meaning 2, 4, 7 and 11. It would be easier to focus in the differences if the homeologous chromosomes were presented all side by side. Also, there is a discussion in the text about lac of synteny among genes or chromosomes (lines 409, 689 and 721). It should be show in the figure – a software as Mauve or similar can be used to made the analysis.

At the 3^rd session of results (line 341), it starts the comparison among genes, transcripts and predicted CDs. In figure 4 (page 14) would be informative to color the corresponding exons by the same color in each gene for comparison. It would help to visualize similarities and variations, especially for the gene PAD4 which have the greatest variation. In figure 5, when the predicted primary structures of proteins are shown, it is very difficult to visualize the similarities between CacPAD4, CcPAD4a and the other versions. Here, the synteny could be shown again to help the readers to visualize the similarity and divergence among the proteins. Regarding to this figure it is also important that the “EDS1” domain follows the same name written in the text, where it is called “EP” domain

In figure 6B (page 19) the resolution of the alignment must be improved. Also, the alignments of the supplemental figures are in low resolution that has to be improved

In session 3.7 (line 644), it would be much more informative to assembly the complete DNA sequence containing the genes and compare differences between them. This is necessary to give a better understanding of the role of the studied genes in rust resistance.

In line 606 it is written that is impossible to align all PAD4 proteins, but if we check on supplemental figures we see that the N-terminal of all proteins is very similar, so possible to be align. The separation between the two types of proteins may generate a better phylogenetic tree, but it is not impossible to align.

Finally, the paper discussion has to be reformulated (lines 767 to 785). The data indicates differences in PAD4 may be responsible for the resistance, but it was not proved – so, it has to be pointed and a strong possibility and other scenarios have to be discussed. It is also that needs to be changed in the title of the manuscript – the differences between the genes among the different species may impact resistance, but it was not shown that they actually do that.

Other important chances need to be done in writing involving some conceptual terminology:

Here are the most important ones:

In line 60, it is used “dominant genes”. There is no such thing. The resistance is dominant, because is conferred by dominant alleles. Genes are never dominant or recessive.

There is confusion with the term “homology”. Homologous sequences have common ancestry: they are or are not homologous. In a DNA sequence comparison we use identity to evaluate how close the sequences are. So, in line 283 and 406 it should be pointed a high identity between sequences. Also, in line 455, the genes share a 100% identity. See also lines 511, 519, 520 and 764. For protein comparison the similarity can also be discussed, and both identity and similarity are used for sequence compasison

Minor review:

Line 86 to 90 – it is confused what refers to the family HIR or to superfamily PID

Line 104 – EDS1, PAD4 and SAG101 belong to the EDS1 family – they are not one family

Line 118 to 120 – it is written that the EDS1 gene is only present in inoculated leaves – it is only expressed in these leaves but the gene is in the whole plant – the same is written in line 274.

Line 275 – the SAG101 gene was not expressed

Line 729 – talks about a terminal of a chromosome, it is 5’ or 3’ extremity?

Supplemental table: in the 3^rd line the primer name is missing

Author Response

We are thankful to the reviewer for the careful revision that allows us to improve the final manuscript. Below we present answers the questions raised.

After revision of the manuscript “Identification of HIR, EDS1 and PAD4 Genes in Coffea spp. reveals differences between Coffea species that can impact its resistance to diseases” I recommend publishing with major reviews. The specific comments and suggestions are listed below:

The manuscript presents relevant and interesting information, but it needs several writing reformulations, additional analysis and careful discussion of the results.

The aim of the work was to analyze gene expression related to host and non-host resistance in the hybrid Coffee genome HDT832/2, and find common genes that may be part of both types of resistance. Then, it compares these genes in coffee genomes and discusses possible implications for resistance. The research is made by a transcriptome analysis and two gene families are chosen: HIR and EDS1. The gene expression profiles in non-host and host responses are compared and results are confirmed by qRT-PCR. Also, the expression of the chosen genes is evaluated in four genotypes of Coffea arabica that show resistance or susceptibility to Hemileia vastatrix and the scenarios are compared. Finally, the studied genes are localized in chromosomes of the three available complete genomes of C. arabica, C. canephora and C. eugenoides and gene structure, transcripts and the primary organization of the predict proteins are compared. Conserved domains are analyzed and variations are discussed.

Major consideration and reviews:

1) The work starts with a transcriptome analysis, but no result of this analysis is cited or shown. I believe the authors are preparing a second manuscript, but it would be interesting to show or refer to what else they have found.

We used the data from the transcriptome analyses as an indication only and we focus on these two gene families that were interesting and highly connected with both broad-range resistance (EDS1 and PAD4) and hypersensitive response connected with NLRs (HIR and EDS1 interacting with PAD4), which are both present in the response of HDT to Hemileia vastatrix. So it was not our intention to describe in detail the genes found in the transcriptome study.

A reference to the fact that we use RNAseq data in a qualitative way was included.

Then, the confirmation of gene overexpression was made by qRT-PCR, together with the study of the response against Uromyces vignae (data shown in Figure 1). This results need to be better analyzed.

2) Statistical analysis to compare results is essential. It may be done by ANOVA followed by mean tests, for example.

The statistical analysis has been made and included in the manuscript.

3) Also, it is not clear if biological replicates were performed, how many plants or leafs were used per sample and how many samples were used in each experiment (it must be included in materials and methods). It is mandatory to have replicates and biological repetitions. The same is missing in gene expression analysis in the C. arabica genomes (Data shown in Figure 2). The meaning of errors bars must also be written in figure legends.

The requested information has been included

4) About methodology, we find in line 149 the session 2.2. “Light microscope observation of fresh tissues” but no image or data is clearly shown during results presentation. It would be interesting to include that as a supplemental material. Also, in line 358 it is described the disease progression and symptoms. It would be interesting to provide the pictures of the processes.

Previously we have published several pictures of the infection process in HDT and other compatible and incompatible interaction genotypes (Silva et al 1999; Silva et al 2002, Diniz et al 2012), and since the sampling used in this manuscript did not deviate from was described before we do not think it is necessary to show more pictures of the infection process. Instead, we include some schematic images of the infection process to help readers to relate gene expression data with the evolution of the infection process.

5) In figure 4, the authors show the genes location in the three species chromosomes. My suggestion is to show only the ones that have the studied genes, meaning 2, 4, 7 and 11. It would be easier to focus in the differences if the homeologous chromosomes were presented all side by side.

Thank you for the suggestion, we changed figure 3 and agree that it gives a better highlight of the differences.

6) Also, there is a discussion in the text about lac of synteny among genes or chromosomes (lines 409, 689 and 721). It should be show in the figure – a software as Mauve or similar can be used to made the analysis.

Synteny can be used in different contexts, and since we were using it just to describe the same localization in the chromosome, we decided to just use the word localization in the text instead of synteny.

7) At the 3^rd session of results (line 341), it starts the comparison among genes, transcripts and predicted CDs. In figure 4 (page 14) would be informative to color the corresponding exons by the same color in each gene for comparison. It would help to visualize similarities and variations, especially for the gene PAD4 which have the greatest variation.

Thank you for the suggestion, we colored the exons in figure 5 and agree that it gives a nice outline of the differences.

8) In figure 5, when the predicted primary structures of proteins are shown, it is very difficult to visualize the similarities between CacPAD4, CcPAD4a and the other versions.

Figure 5 was designed mainly to show the position of the protein conserved domains in EDS1 and PAD4 proteins, namely to show that Lipase 3 and EP domains are both missing in the predicted PAD4 protein from Coffea arabica, and where the EDS1 protein from Coffea canephora has a FH3 domain. We think that a schematic representation of the proteins fulfills both goals. Figure 5. Schematic representation of the conserved domains of EDS1 (A) and PAD4 (B).

9) Here, the synteny could be shown again to help the readers to visualize the similarity and divergence among the proteins.

As referred above the term synteny was used just to describe the same localization in the chromosome, so we decided to just use the word localization in the text instead of synteny.

10) Regarding to this figure it is also important that the “EDS1” domain follows the same name written in the text, where it is called “EP” domain

The EDS1 domain in the figure was corrected.

11) In figure 6B (page 19) the resolution of the alignment must be improved. Also, the alignments of the supplemental figures are in low resolution that has to be improved

We think that the quality was improved.

12) In session 3.7 (line 644), it would be much more informative to assembly the complete DNA sequence containing the genes and compare differences between them. This is necessary to give a better understanding of the role of the studied genes in rust resistance.

We do agree that the nucleotide sequence analysis in these genotypes and in a broader number of genotypes in CIFC collection would be very interesting, and, in the future, we intent to conduct such an analysis. However, even if the analysis presented does not permit the obtention of full sequences it gives an indication of the gene structure in those genotypes that could be helpful for the scientific community that is involved in disclosing tight mechanisms controlling plant immunity.

13) In line 606 it is written that is impossible to align all PAD4 proteins, but if we check on supplemental figures we see that the N-terminal of all proteins is very similar, so possible to be align. The separation between the two types of proteins may generate a better phylogenetic tree, but it is not impossible to align.

Most alignment programs have difficulties in aligning more than two sequences with very different sizes. We opted to divide the sequences in two groups because we think that it shows better the division between the two groups. The phylogenetic analysis was improved and more robust when we opted for the three groups. However, in supplementary data we show a phylogenetic tree that includes the complete EDS1 and PAD4 sequences.

14) Finally, the paper discussion has to be reformulated (lines 767 to 785). The data indicates differences in PAD4 may be responsible for the resistance, but it was not proved – so, it has to be pointed and a strong possibility and other scenarios have to be discussed. It is also that needs to be changed in the title of the manuscript – the differences between the genes among the different species may impact resistance, but it was not shown that they actually do that.

Several changes were made on the manuscript to clarify the main findings and their importance to the plant resistance.

Other important chances need to be done in writing involving some conceptual terminology:

Here are the most important ones:

15) In line 60, it is used “dominant genes”. There is no such thing. The resistance is dominant, because is conferred by dominant alleles. Genes are never dominant or recessive.

Thank you for your suggestion that contribute to the language clarification. Changed!

16) There is confusion with the term “homology”. Homologous sequences have common ancestry: they are or are not homologous. In a DNA sequence comparison we use identity to evaluate how close the sequences are. So, in line 283 and 406 it should be pointed a high identity between sequences. Also, in line 455, the genes share a 100% identity. See also lines 511, 519, 520 and 764. For protein comparison the similarity can also be discussed, and both identity and similarity are used for sequence comparison

Thank you for your suggestion that contribute to the language clarification. Changed!

Minor review:

17) Line 86 to 90 – it is confused what refers to the family HIR or to superfamily PID

Done

18) Line 104 – EDS1, PAD4 and SAG101 belong to the EDS1 family – they are not one family

Corrected.

19) Line 118 to 120 – it is written that the EDS1 gene is only present in inoculated leaves – it is only expressed in these leaves but the gene is in the whole plant – the same is written in line 274.

"present" was replaced by "express"

20) Line 275 – the SAG101 gene was not expressed

Done

21) Line 729 – talks about a terminal of a chromosome, it is 5’ or 3’ extremity?

Done.

22) Supplemental table: in the 3^rd line the primer name is missing

Done

 

 

Reviewer 3 Report

The manuscript of Tavares et al. “Identification of HIR, EDS1 and PAD4 Genes in Coffea spp. reveals differences between Coffea species that can impact its resistance to diseases“ shows results from transcriptomic analysis of resistant Coffea cultivar, RT-qPCR analysis and in-silico analysis of selected gene sequences. Based on these analyses authors claim that variability within EDS1/PAD4 gene family may present important attribute of pathogen resistance. The work belong to the very interesting and important field and may provide important results belonging to both, fundamental and application science. However, current form of the manuscript suffers to numerous more or less critical limitations which do not allow me to recommend presented work for publication. I will try to address the most critical issues in the following revision.

In general, the English language of the manuscript should be improved.

The first part of the manuscript describes RNAseq profiling. Authors try to show differentially regulated genes, but unfortunately, 454 methods used in presented work does not belong to the state-of-the-art techniques and should not be used for robust transcription quantification and there are no sample repetitions for some statistics evaluation. I recommend authors to present these data only as to show what transcripts were detected (qualitative analysis) and then select your genes of interest for further quantification by RT-qPCR. Authors should explain why they chose homologs of HIR, EDS and PAD but I understand that it is really hard based on such unreliable RNAseq database. Anyway, quality control of RNAseq data and some characterization of the assembly and mapping should be shown too. For example, it is not clear how many reads they obtained per sample. The number 877,790 presents a total number of reads from all samples (control and two inoculated) together?  My last minor comment to this part is that at line 286 authors claim that several contigs belong to the same gene, however, in the corresponding table, Gene IDs are different.

The second main section of results describes measurement of relative expression of selected genes (HIR, EDS gene families and PAD4) after inoculation by pathogen. As the RNAseq did not provide reliable data regarding gene expression, this should be the main source of robust data to show role of selected genes during plant response reaction. Unfortunately, the data showed here and not robust and the whole procedure suffers to several serious issues. Authors claim that they used three, respective two reference genes for normalization of expression. However, both equations do not reflect this. Furthermore, I found equation 2 to be probably wrong. E_CDAs should represent PCR efficiency of which gene? It is also not clear whether authors measured qPCR efficiencies for all used primers. If so, and if real efficiencies differed, it cannot be used as a base (proxy) powered by difference of Cts. Further questions and comments regarding quantification of transcription:

How many biological and technical replicates were measured for each sampletype?

Reference genes were selected based on which data? Authors cite RefFinder, but do not explain the data used as input.

qPCR data do not validate RNAseq data at all, but that should be fixed by changing the outcome of RNAseq to qualitative only.

There is no statistics included in the manuscript. Why?

In 3.2 (Figure 1), we see only three genes with some upregulation of pathogen inoculation: HIR4, PAD4 and EDS1 (very little upregulation). All other tested genes show significant downregulation. Even those upregulated ones show the interesting profile only very early after inoculation (6 hai). I find the huge drop in 12 hai very suspicious. I would recommend authors to measure this progress to better detail to make the result trustworthy. In optimal case, perform qPCR in samples collected 3 hai and 9 hai to confirm the overexpression of HIR4 and PAD4 at the specific time window potentially important for the course of pathogen attack.

In 3.3, transcript levels were measured in 18 hai and 1, 2, 3 dai whereas in previous experiment it was in 6, 12 and 24 hai. Why authors performed both experiments under different conditions? It is hard to compare both data. I also do not understand why there were no controls in the second experiment.

Also in 3.3 (Figure 2), certain transcripts shows strikingly similar pattern across specific genotype. For example, in CIFC H147, all genes but PAD4 show the highest transcription in 1 dai followed by drop in trans. level in 2 dai and then increases again in 3 dai. Similarly, in H469, all genes followed the same trend. Can we trust these patterns or do they developed due to some artefact problem in the experimental procedure, for example wrong reference genes.

To summarize, qPCR part of this manuscript should present the most important data and thus should be performed to the best level possible. Current state of the data is for me very hard to believe and have very little potential.

The third part of results is based on in silico analysis of selected genes in several Coffea genomes. While such analysis might be interesting, without proper functional experiment do not bring strong impact. I was pleased to read that authors validated some sequence by subsequent PCR and sequencing but I did not find this validation for the most important and interesting sequence - the CcEDS1 gene with unexpected FH2 region. Authors should add this validation to prove that the gene is really transcribed and expressed into such novel version of EDS1 protein. Since I am no expert in phylogenetic analyses, I will abstain from commenting this section further, but to my naive eye, it looks nice. But unfortunately, the clear link between in silico analysis and plant pathogen resistance is missing.

Finally, the whole manuscript is very hard to read (not only due to weak English). I would recommend authors to cut it and make it shorter, more compact and easier to read and understand. There are many empty paragraphs in introduction and discussion and several results with little novelty or impact for the quality of publication. Keep only important and interesting results. For better intelligibility, try naming each part of Results by stating the key point (results) of the specific paragraph, e.g. Presence of specific isoforms of resistance-related gene families detected by RNAseq analysis. Or something in similar logic. Such approach will also help authors to select which results are important (have some impact), which are potentially interesting datamining report (e.g. phylogeny) and which are just space-filling-ballast. J

Author Response

We are thankful to the reviewer for his recommendation and constructive comments. We have carefully considered the suggestions of the reviewer and made some changes on the manuscript.

The revisions are addressed point by point below:

The manuscript of Tavares et al. “Identification of HIR, EDS1 and PAD4 Genes in Coffea spp. reveals differences between Coffea species that can impact its resistance to diseases“ shows results from transcriptomic analysis of resistant Coffea cultivar, RT-qPCR analysis and in-silico analysis of selected gene sequences. Based on these analyses authors claim that variability within EDS1/PAD4 gene family may present important attribute of pathogen resistance. The work belong to the very interesting and important field and may provide important results belonging to both, fundamental and application science. However, current form of the manuscript suffers to numerous more or less critical limitations which do not allow me to recommend presented work for publication. I will try to address the most critical issues in the following revision.

In general, the English language of the manuscript should be improved.

A revision was done in order to correct the English language

The first part of the manuscript describes RNAseq profiling. Authors try to show differentially regulated genes, but unfortunately, 454 methods used in presented work does not belong to the state-of-the-art techniques and should not be used for robust transcription quantification and there are no sample repetitions for some statistics evaluation. I recommend authors to present these data only as to show what transcripts were detected (qualitative analysis) and then select your genes of interest for further quantification by RT-qPCR. Authors should explain why they chose homologs of HIR, EDS and PAD but I understand that it is really hard based on such unreliable RNAseq database. Anyway, quality control of RNAseq data and some characterization of the assembly and mapping should be shown too. For example, it is not clear how many reads they obtained per sample. The number 877,790 presents a total number of reads from all samples (control and two inoculated) together?  My last minor comment to this part is that at line 286 authors claim that several contigs belong to the same gene, however, in the corresponding table, Gene IDs are different.

The second main section of results describes measurement of relative expression of selected genes (HIR, EDS gene families and PAD4) after inoculation by pathogen. As the RNAseq did not provide reliable data regarding gene expression, this should be the main source of robust data to show role of selected genes during plant response reaction.

The transcriptome data analysis referred in the manuscript was done as one of the first approaches to NGS and the data were assembled without a reference genome, although the results allowed the identification of differentially expressed genes during host and non-host coffee-rust interaction. The qualitative analysis of the transcripts level uncovers the putative role of PAD4, EDS, and HIR gene families on the resistance responses. The identified genes were confirmed through blast search against Coffea canephora, C. eugenioides, and C. arabica genomes. It is important to note that the genome of HDT has not been sequenced.

The expression profiles used were those obtained by RT-qPCR.

Some improvements were done in the manuscript to clarify this question.

Unfortunately, the data showed here and not robust and the whole procedure suffers to several serious issues. Authors claim that they used three, respective two reference genes for normalization of expression. However, both equations do not reflect this. Furthermore, I found equation 2 to be probably wrong. E_CDAs should represent PCR efficiency of which gene? It is also not clear whether authors measured qPCR efficiencies for all used primers. If so, and if real efficiencies differed, it cannot be used as a base (proxy) powered by the difference of Cts. Further questions and comments regarding quantification of transcription:

Thank you for your comment, which allows us to clarify the methodology of gene expression quantification.

The fold change calculation were done accordingly to the Pfaffl method, the equation used is attached (Please see the attachment)

How many biological and technical replicates were measured for each sampletype?

This information was included in the M&M section

Reference genes were selected based on which data? Authors cite RefFinder, but do not explain the data used as input.

The input data used in RefFinder came from the RT-qPCR with cDNA of all samples in analysis, according to the software instructions. The final result output includes different methodology (Comparative DCt method, GeNorm, BestKeeper and NormFinder) and give an overall final ranking.

qPCR data do not validate RNAseq data at all, but that should be fixed by changing the outcome of RNAseq to qualitative only.

Our aim was to study the expression profiles of HIR, EDS1, and PAD4 genes by RT-qPCR since those genes were present in the transcriptome of HDT leaves inoculated with host and non-host rusts. We do not intend to validate the data from 454 pyrosequencing.

There is no statistics included in the manuscript. Why?

The statistical analysis was included

In 3.2 (Figure 1), we see only three genes with some upregulation of pathogen inoculation: HIR4, PAD4 and EDS1 (very little upregulation). All other tested genes show significant downregulation. Even those upregulated ones show the interesting profile only very early after inoculation (6 hai). I find the huge drop in 12 hai very suspicious. I would recommend authors to measure this progress to better detail to make the result trustworthy. In optimal case, perform qPCR in samples collected 3 hai and 9 hai to confirm the overexpression of HIR4 and PAD4 at the specific time window potentially important for the course of pathogen attack.

Indeed the results could seem very weird, but these profiles were observed in these HDT plants with other genes. It is important to notice that these interactions showed a pre-haustorial resistance (Diniz et al 2012) in which the fungal growth stops in the phase of penetration hypha and resistance responses (such as HR) were observed from the full appressorium or penetration hypha stages onwards.

In 3.3, transcript levels were measured in 18 hai and 1, 2, 3 dai whereas in previous experiment it was in 6, 12 and 24 hai. Why authors performed both experiments under different conditions? It is hard to compare both data. I also do not understand why there were no controls in the second experiment. To summarize, qPCR part of this manuscript should present the most important data and thus should be performed to the best level possible. Current state of the data is for me very hard to believe and have very little potential.

The experimental conditions were adapted to the different patterns of the fungal infection process observed in the plant-microbe interactions studied.

In the incompatible interaction HDT 832/2 - H. vastatrix, the fungus ceased its growth more frequently before haustorium formation (prehaustorial resistance). The nonhost resistance of this coffee genotype to Uromyces vignae was also expressed before haustorium formation. However, in both C. arabica-Hv incompatible interactions the fungus ceased its growth in more advanced stages of the infection than in HDT 832/2, being able to produce haustoria, particularly the genotype CIFC H469/16 (posthaustorial resistance). 

Also in 3.3 (Figure 2), certain transcripts shows strikingly similar pattern across specific genotype. For example, in CIFC H147, all genes but PAD4 show the highest transcription in 1 dai followed by drop in trans. level in 2 dai and then increases again in 3 dai. Similarly, in H469, all genes followed the same trend. Can we trust these patterns or do they developed due to some artefact problem in the experimental procedure, for example wrong reference genes.

The reference genes were validated by the bioinformatic tools and it is our conviction that the results presented are not due to any artifact but come from tricky differences, not completely disclosed, that are observed among the different interactions and are translated in different fungal growth and the simultaneous presence of a puzzling number of resistance symptoms in compatible interactions as well as a high degree of fungal development without the production of new uredospores in the incompatible interactions.

The third part of results is based on in silico analysis of selected genes in several Coffea genomes. While such analysis might be interesting, without proper functional experiment do not bring strong impact. I was pleased to read that authors validated some sequence by subsequent PCR and sequencing but I did not find this validation for the most important and interesting sequence - the CcEDS1 gene with unexpected FH2 region.

Authors should add this validation to prove that the gene is really transcribed and expressed into such novel version of EDS1 protein. Since I am no expert in phylogenetic analyses, I will abstain from commenting this section further, but to my naive eye, it looks nice. But unfortunately, the clear link between in silico analysis and plant pathogen resistance is missing.

The functional validation suggested here is of much interest and relevance. However, genetic transformation of coffee is cumbersome, regeneration of plants from transformed protoplasts is quite problematic, not to mention the added difficulties posed by the polyploid nature of Coffea arabica.

However, we consider that even without functional validation the data presented here is sufficiently interesting and useful to share with the scientific community working on coffee.

Finally, the whole manuscript is very hard to read (not only due to weak English). I would recommend authors to cut it and make it shorter, more compact and easier to read and understand. There are many empty paragraphs in introduction and discussion and several results with little novelty or impact for the quality of publication. Keep only important and interesting results. For better intelligibility, try naming each part of Results by stating the key point (results) of the specific paragraph, e.g. Presence of specific isoforms of resistance-related gene families detected by RNAseq analysis. Or something in similar logic. Such approach will also help authors to select which results are important (have some impact), which are potentially interesting datamining report (e.g. phylogeny) and which are just space-filling-ballast. J

Several changes were made to the manuscript to make it easier to read and understand.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

I don't see any experiments added or any figures significantly changed in the updated manuscript based on my suggestions. 

I advised the authors to repeat the qRT-PCR with more time points in my earlier review report. However, the authors reject doing so. You still need to do it even when another gene's similar results were reported in previous study by someone else. They are different genes, and the authors hope to publish a novel study. 

 

 

 

 

Author Response

Based on the authors’ suggestion, we had repeated and included more time points in the qPCR analysis. The new charts were included and the discussion was complete with the new data.

Reviewer 3 Report

I have read response of authors to my comments and I am quite sure that the manuscript was not improved to the level I would rank acceptable for publication in Agronomy. As I stated in the revision, authors should be given enough time to improve the manuscript. Controls are missing. Sequence of CcEDS1 should be verified. Additional RT-qPCR should be performed to validate the data and to make here published results more robust and trustworthy. All these requests were stated in my original revision.

Author Response

Controls are missing.

All controls were included in the analyses.

Sequence of CcEDS1 should be verified.

This sequence was reported by the Coffee Genome Hub project. In ours genotypes we were unable to confirm this sequence, and we do not have the sequenced genotype, so we could not verified the existence of this sequence, however, it is important to note (we included this reference in the discussion) that this gene structure has been reported in other species, namely Vitis.

Additional RT-qPCR should be performed to validate the data and to make here published results more robust and trustworthy. All these requests were stated in my original revision.

Based on the reviewer’s’ suggestion, we had repeated and included more time points in the qPCR analysis. The new charts were included, and the discussion was complete with the new data.

 

Round 3

Reviewer 1 Report

I appreciate the authors' correspondence. The added results were included in their updated manuscript. However, they might have forgotten replacing the Figure 1. I would advise making a minor revision and accepting this manuscript after the proper figure has been substituted.

 

Author Response

In the manuscript revision we have uploaded the new figures, however we did not included the new figures in the manuscript.

The new manuscript with all the new figures is being uploaded.