Microbial Community Profiling from Natural Whey Starter to Mozzarella among Different Artisanal Dairy Factories in Apulia Region (Italy)

Round 1

Reviewer 1 Report

1. The lack of line numbers makes the review difficult .
2.  Has the presence of Lactobacillus helveticus been confirmed by other molecular methods? Species identification of bacteria in the microbial community using NGS does not necessarily provide reliable results.
3.  Figs. 1A, 1B are poorly visible. The colors are too similar. 
4. The bioinformatics analysis should be extended using a Venn diagram and abundance Fold Change (expressed as log2 Fold Change) of bacterial generation. Thanks to these analyses, it will be possible to make a more in-depth comparison of the microbiomes of mozzarella and starters from selected factories.

 

Author Response

1. The lack of line numbers makes the review difficult .

Reply: Perfectly right. We uploaded a new revised version of the paper, with line numbers.

2. Has the presence of Lactobacillus helveticus been confirmed by other molecular methods? Species identification of bacteria in the microbial community using NGS does not necessarily provide reliable results.

Reply: Unfortunately, data discussed in this manuscript have been produced only from research purpose: thus, no other diagnostic and more accurate molecular techniques have been applied on the whole sample set. Although powerful and highly processive, we recognize that a 16S-rDNA NGS approach is not fully reliable (especially in species determination, as we highlight in lines 383-388). Literature seems to confirm that L. helveticus would be one of the main acidifying bacteria within different types of mozzarella. We further evidenced this issue from lines 388 to391. Thanks for the opportunity to clarify this aspect.

3. 1A, 1B are poorly visible. The colors are too similar.

Reply: Thanks for the remark. We removed the Figure 1A and leave the Figure 1B (new Figure 1), where relative abundance of bacterial genera other than Lactobacillus are reported. We hope this can give a more evident and clear idea of microbial composition.

4. The bioinformatics analysis should be extended using a Venn diagram and abundance Fold Change (expressed as log2 Fold Change) of bacterial generation. Thanks to these analyses, it will be possible to make a more in-depth comparison of the microbiomes of mozzarella and starters from selected factories.

Reply: Many thanks for the idea: we added Figure 4 with two Venn diagrams showing the results of the differential abundance analysis methods: genera that have been evidenced as “over-abundant” within the two sample group (“S” and “M”) are reported. On the statistical methodology: before implementing the coda-lasso, clr-lasso and Selbal approaches, we tried to apply more classical methods like the Deseq2 function within Phyloseq package, obtaining no significant results. We explain this aspect within lines 188-192.

Reviewer 2 Report

The current manuscript conducted the microbial community profiling from natural whey starter to Mozzarella among different artisanal dairy factories in Apulia Region (Italy). The topic is interesting. However, the data presenting and analysis is not sufficient. The authors tried to tell a complicated story with simple results. The results are as simple as the column figure illustrated, that Lactobacillus is the core acidifying component. It is not necessary to highlight these simple results with big figures or long discussion context. Let it be simple, since the valuable information is just as much as you presented in the figures whereas not in the long discussion. Detail comments are as following:

 

1. The abstract:

“There is a limited knowledge about microbial composition and diversity of artisanal mozzarella” This is impossible. The author did not do enough searching of references.

 

“Biodiversity was found quite similar between the whey and mozzarella sample pools, while a significant variability among production sites (factories) has been detected. Lactobacillus is the core acidifying component of the used starters, while some psychrophilic or contaminants bacteria appear in site-specific products.” These two sentences should be reversed according to the result sequence presented in the text. The “Lactobacillus” and all the genus names should be italic. Except for species, biodiversity, and the correlation analysis, there is no further information the authors could provide, is there? If so, why the main context is like this long?

 

2 In page3, paragraph “Library construction and sequencing”. Which company did the sequencing? It is not shame to present the company name, since we all know that you can’t own all the equipment to conduct the NGS. The NGS and the associated analysis is just common procedure for an NGS company. The authors provided long text for the detail procedure; it is actually not necessary. However, it is the authors’ choice. But at least, the company name should be presented. Is the “Lactic acid determination” also did by a company? Then, please provide the company name.

 

3. Fig.1 A and B should be combined. For either graph A or B, it is too big. The information is simple, which is the genus composition. To provide this simple information, it is not necessary to draw this half-page graph. Please combine the Fig1 A and B, and illustrate them in one figure and one page.

 

4. Fig.2 A and B has the same problem as the Fig. 1A and B. What does the Evenness mean? Is the Evenness similar parameter to the Chao and Shannon? If so, it is not necessary to present Fig.2 B again. What does Class M and S mean in these figures?

 

5. Except for the genus composition and biodiversity analysis, the authors only did the correlation analysis of “%lactic acid” and “Shannon diversity”. This is very limited. Why not do PCA or RDA to group the samples? Why not do more analysis? The authors should introduce new method and provide more information from the NGS data, otherwise this manuscript is as simple as the genus composition, biodiversity and correlation, nothing more.

Author Response

The current manuscript conducted the microbial community profiling from natural whey starter to Mozzarella among different artisanal dairy factories in Apulia Region (Italy). The topic is interesting. However, the data presenting and analysis is not sufficient. The authors tried to tell a complicated story with simple results. The results are as simple as the column figure illustrated, that Lactobacillus is the core acidifying component. It is not necessary to highlight these simple results with big figures or long discussion context. Let it be simple, since the valuable information is just as much as you presented in the figures whereas not in the long discussion. Detail comments are as following:

 

1. The abstract:

 

“There is a limited knowledge about microbial composition and diversity of artisanal mozzarella” This is impossible. The author did not do enough searching of references.

Reply: Amended. We re-formulated the concept by highlighting that, as reported in the Abstract, knowledge on this topic is increasing, probably due to the usage of Next Generation Techniques and the need for a more accurate microbiological control of PDO products.

 “Biodiversity was found quite similar between the whey and mozzarella sample pools, while a significant variability among production sites (factories) has been detected. Lactobacillus is the core acidifying component of the used starters, while some psychrophilic or contaminants bacteria appear in site-specific products.” These two sentences should be reversed according to the result sequence presented in the text. The “Lactobacillus” and all the genus names should be italic.

Reply: Thanks for the suggestion. Sentences within abstract have been reversed.

Except for species, biodiversity, and the correlation analysis, there is no further information the authors could provide, is there? If so, why the main context is like this long?

Reply: Actually, we tried to implement multiple different statistical and bioinformatic approaches on the sequencing and chemical dada (lines 167-205). First, we statistically evaluated performance of DNA extraction and sequencing throughput  for the two different matrices. Second, we provide read counts at every taxonomic level (collected in Supplementary Tables), focusing on “genus” composition in the main text. Third, we analysed alfa diversity by using four popular indexes (three metrics of species diversity, one for species evenness), and beta diversity by performing Principal Component Analyses (new Figure 3) and PERMANOVA tests (Supplementary File 1, Table 6). We checked also for a “Sampling time” effect on sample microbiological profile; we also tried to compute other alfa indices (for species evenness and dominance), but we find no statistical significant results.

After compositional and diversity analyses, we implemented three advanced differential abundance approaches (coda-lasso, clr-lasso, selbal), in order to detect which microbial genera, despite the extreme predominance of L. helveticus and the asymmetrical group sizes, significantly varied  between the two groups. We also tried other computational methods for microbial cross-correlation (i.e., genera with putative co-varying abundance profiles), but we did not find any significant results. The application of these techniques is quite novel in food metagenomics. 

Additionally, we tried to investigate the relationship among diversity, compositional and chemical data for the “M” group, that we collected/computed in the form of numerical vectors (although of limited sample size, n=18). Thus, we performed all possible pairwise correlations among the diversity distributions Shannon entropy, Inverse Simpson, Chao1, Pielou’s evenness versus the following chemical ones: % lactic acid, %lactose, %fat, moisture. We correlated the abundance profiles of each of the 43 detected genera (after ultra-rare and contaminant sequences removal) of “M” samples with the four chemical measurements, but we did not retrieve any significant correlation.

We are well aware of the verbosity and length of our manuscript. First, we made an effort to accurately describe in vitro and in silico methods in order to be transparent and open to research community and report all the significant results (plus all raw data within Supplementary File). Secondarily, we have to meet the journal strict requirements for an Original Article (at least 4,000 words).

 2 In page3, paragraph “Library construction and sequencing”. Which company did the sequencing? It is not shame to present the company name, since we all know that you can’t own all the equipment to conduct the NGS. The NGS and the associated analysis is just common procedure for an NGS company. The authors provided long text for the detail procedure; it is actually not necessary. However, it is the authors’ choice.

But at least, the company name should be presented. Is the “Lactic acid determination” also did by a company? Then, please provide the company name.

Reply: Thanks a lot for the suggestion. We performed sequencing and chemical analyses in our own facilities, as reported in lines 164-165 and 205-206. As abovementioned, we accurately describe the methodological part in order to be of support for researchers that conduct microbiological/chemical studies on fresh dairy products.

3. Fig.1 A and B should be combined. For either graph A or B, it is too big. The information is simple, which is the genus composition. To provide this simple information, it is not necessary to draw this half-page graph. Please combine the Fig1 A and B, and illustrate them in one figure and one page.

Reply: Thanks a lot. We opted to remove Figure 1A and pinpoint highly abundant genera (other than L. helveticus) in the new Figure 1 (previously named Figure 1B).

4. Fig.2 A and B has the same problem as the Fig. 1A and B. What does the Evenness mean? Is the Evenness similar parameter to the Chao and Shannon? If so, it is not necessary to present Fig.2 B again. What does Class M and S mean in these figures?

Reply: We re-sized the two Figures and provided better explanation of adopted short names. We explain the concept of species “evenness” (most popular mathematical conceptualization is Pielou’s evenness) in lines 178-182. Thus, Figure 1A is related to species diversity measures while 1B regards species evenness. Statistical comparisons (Wilcoxon tests) have been made for all measures. “Class M and S” indicate the samples classes “Mozzarella” and “Natural Whey Starters” or, briefly, “Starters”.

5. Except for the genus composition and biodiversity analysis, the authors only did the correlation analysis of “%lactic acid” and “Shannon diversity”. This is very limited. Why not do PCA or RDA to group the samples?

Reply: As abovementioned, we performed numerous pairwise correlations among numerical distributions of chemical and abundance profiles, as detailed in the Method section. Considering that we had only 18 data points for each distribution and un-matched samples, we were not able to perform more complex analyses. On the other hand, we performed PCA analysis coupled to PERMANOVA test to verify that: taken together, “M” microbiological profiles are similar to the “S” ones. However, microbiological profiles among factory sites are significantly dissimilar to each other. These results are presented as graphics (Figure 3) and Supplementary Table (ST6).

Why not do more analysis? The authors should introduce new method and provide more information from the NGS data, otherwise this manuscript is as simple as the genus composition, biodiversity and correlation, nothing more.

Reply: As described in the Method part, we attempted various bioinformatic methodologies on NGS data and provided all the significant results in the main text. Moreover, full results are also provided in the Supplementary File. Some of the computational methodologies (Selbal, coda-lasso, clr-lasso) are innovative tools for differential abundance analyses (please, look at DOI:10.1093/nargab/lqaa029 and 10.1128/mSystems.00053-18, for a detailed view of such methods), and have been added to compositional, diversity and correlation analyses.

As abovementioned, we made a great effort to implement other different techniques, obtaining no relevant results for some of them. Thus, we recognize that results are not extraordinarily novel or complex, but quite intriguing. The main one: independent dairy factory sites use their own autochthonous starter, that, interestingly, have the same core component, L. helveticus. Nonetheless, general microbial composition of starter and mozzarella samples is factory-specific. Another interesting point is the correlation between microbiological diversity and %lactic acid, even if calculated on a limited sample set of dairy product. We finally consider this research as a good starting point for a better understanding of Apulian artisanal mozzarella microbiology. Indeed, we are planning a more in-depth longitudinal microbiological survey for a selected factory (among the four ones described in this manuscript) and a larger inspection on other Apulian factories, using chemical acidification methods (see lines 420 – 428).

Round 2

Reviewer 2 Report

No further comments.