Organization Across a Caragana korshinskii Plantation Chronosequence: Edaphic Gradients, Hierarchical Filtering and Domain-Specific Assembly
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe manuscript presents an interesting study of changes in the soil microbiome along a chronosequence of 5-, 10-, 20-, and 30-year-old stands of Caragana korshinskii, a species used for land restoration in arid climates. A major strength of this study, compared with other published studies, is that, in addition to the aforementioned age gradient, it incorporates a spatial gradient (from bulk soil, through the rhizosphere, to roots) and considers both bacterial and fungal soil communities, which exhibit distinct patterns of community organization.
The article is written concisely. A range of advanced statistical techniques was used for data analysis, and the results are presented in complex and aesthetically appealing figures. In my opinion, however, for such a complex study, placing the Materials and Methods section at the end of the manuscript is not entirely appropriate. Although this format is used by some journals, here, following a general introduction, the authors immediately expose the reader to a rather technical description of the results and to the aforementioned complex figures. Proper interpretation of these results requires the reader to first become familiar with the study design and methodology. In my view, this is somewhat at odds with the editorial philosophy of this journal, where reference numbering follows the order in which citations appear in the text, while obtaining information about the details of the study requires the reader to move between different sections of the manuscript. I am also somewhat surprised by the apparent mixture of formatting styles associated with the journals Forests and Plants.
I have several comments concerning the study design that may have implications for the subsequent parts of the manuscript.
Please clarify the exact sampling procedure for the “Soil 10 cm” and “Soil 30 cm” treatments. Do these labels refer to the depth below the soil surface at which the samples were collected, or to the distance from the plant/root system? Please provide the vertical and horizontal position of each soil sample relative to the shrub and its roots, as well as the range of soil depths represented by each sample.
Please also specify the depth from which the soil–root samples were collected for the characterization of the “rhizosphere” and root-associated samples.
The authors should also clarify how the samples were collected and what the spatial sampling design was. Does each age class correspond to a single plot containing five selected shrubs? Such a design has important implications for the scope of subsequent inference, because age class would then be confounded with site, while each shrub would represent only a subsample rather than an independent replicate.
Please report the number of samples collected, whether these were the same samples from which DNA was extracted, and how the samples were pooled or otherwise aggregated.
Looking at the figures, I wonder whether simplifying their presentation and improving readability could be achieved by retaining the labels 5Y, 10Y, 20Y, and 30Y rather than introducing the additional G1–G4 symbols.
In the PCA of soil chemical properties, which forms the basis for defining the PCA-derived edaphic axis (ERGI), the authors do not report which of the analyzed soil properties contributed most strongly to the observed distribution of points in the ordination. The concentrations of some elements increased with stand age, whereas those of others decreased. The environmental vectors shown in Fig. 1C appear relatively short, raising concerns about how strongly the measured edaphic variables are represented in the displayed ordination space. Please report the PCA loadings for each environmental variable.
I would also consider retaining the term “PCA-derived edaphic axis” or “Edaphic Gradient Index” in the description. Alternatively, the term “restoration” should be more clearly justified.
The Results section is largely framed around a sequential interpretation in which stand age is associated with edaphic reconfiguration, which in turn is interpreted as contributing to microbiome reorganization. However, given the chronosequence design, the data primarily demonstrate that microbiome composition differs among stands representing different stages of the chronosequence. They do not, by themselves, provide sufficient evidence to conclude that stand development caused the observed microbiome reorganization. In particular, a chronosequence represents a space-for-time substitution and therefore does not directly demonstrate temporal changes within the same stand.
The authors themselves acknowledge that “because distance-based tests can be affected by group dispersion, these PERMANOVA results should be interpreted with the corresponding PERMDISP diagnostics.” They also state in the Methods that PERMDISP was performed. However, I could not identify the specific PERMDISP results in the Results section. This is an important omission, as significant differences in multivariate dispersion could contribute to the observed PERMANOVA effects. In such a case, the interpretation of the PERMANOVA results requires caution, because significant differences may reflect differences in within-group dispersion in addition to, or rather than, differences in community centroids.
A key link appears to be missing from the current analysis: the direct relationship between edaphic conditions and microbiome composition. The study demonstrates an association between stand age and changes in soil chemistry, as well as between stand age and changes in microhabitat and microbial community structure. However, it does not directly demonstrate that variation in soil chemistry is associated with, or explains, the observed changes in microbial community composition. This is particularly important given that the title and the overall narrative of the manuscript attribute microbiome reorganization to “edaphic reconfiguration.” Therefore, the proposed role of edaphic reconfiguration currently appears to be more interpretative than directly supported by the presented analyses. If this relationship is central to the study's conceptual framework, the authors should consider providing an analysis that directly evaluates the association between edaphic variables and microbial community composition.
Overall, I consider the study potentially publishable, but substantial revisions are required to strengthen the statistical support and, most importantly, to ensure that the conclusions remain appropriately aligned with what can be inferred from the study design and analyses. I therefore recommend major revision.
Author Response
We thank the reviewer for the careful and constructive evaluation. We agree that the main issues concern reproducibility of the sampling design, the inferential limits of a chronosequence, explicit reporting of dispersion diagnostics, and the previously overstated connection between edaphic change and microbiome reorganization. We have substantially reorganized and rewritten the manuscript accordingly.
Comment 1. For a complex study, Materials and Methods should precede Results; the manuscript also appears to mix Forests and Plants formatting.
Response: We agree. The Materials and Methods section has been moved immediately after the Introduction so that the sampling design and analytical framework are available before the Results. The sections have been renumbered throughout. We also standardized the journal header to Forests and removed the remaining Plants header wording from the supplied file.
Revision: Structure changed to 1. Introduction → 2. Materials and Methods → 3. Results → 4. Discussion → 5. Conclusions; journal header standardized.
Comment 2. Clarify exactly what “Soil 10 cm” and “Soil 30 cm” mean, including vertical depth, horizontal position relative to shrubs/roots, and sampled depth ranges.
Response: We agree and have clarified the terminology from the field records. “Soil 10 cm” and “Soil 30 cm” refer to horizontal distance from the sampled root system, not vertical depth. Fine roots were located at 25–30 cm below the soil surface, and non-rhizosphere soil was collected at the same 25–30 cm vertical depth but 10 cm or 30 cm horizontally from the target roots. The revised Methods now state both the vertical depth and horizontal position explicitly.
Revision: Revised Section 2.1, Study site and sampling strategy.
Comment 3. Please specify the depth from which rhizosphere and root-associated samples were collected.
Response: Agreed. The revised Methods now specify that the fine-root system, rhizosphere soil and operational Root compartment were all sampled at 25–30 cm soil depth. This also makes clear that the Soil 10 cm/30 cm labels describe horizontal distance rather than depth.
Revision: Revised Section 2.1.
Comment 4. Clarify the spatial sampling design and whether each age class corresponds to one plot with five shrubs; if so, age is confounded with site and shrubs are subsamples rather than independent replicates.
Response: We agree that this is the most important design limitation. We have now confirmed that each stand-age class was represented by one plantation established in a different year, within which five shrubs were randomly sampled. The five shrubs are therefore within-stand spatial subsamples, not independent stand-level replicates, and stand age is fully confounded with plantation/site. We have revised the title, Abstract, Methods, Results, Discussion and Conclusions accordingly. Stand-age-related statistics are now treated as descriptive within this specific chronosequence and are not interpreted as replicated population-level age effects or direct temporal changes.
Revision: Revised title, Abstract, Section 2.1, Sections 3.1–3.3, Discussion limitations, and Conclusions.
Comment 5. Report the number of samples collected, whether these were the samples used for DNA extraction, and whether/how samples were pooled.
Response: We agree. The field design comprised three sampling compartments—Root, Rhizosphere and root-surrounding soil—with the root-surrounding soil represented at two horizontal positions (10 and 30 cm from the located root system). Therefore, each shrub yielded four operational microbiome sample types: Soil 10 cm, Soil 30 cm, Rhizosphere and Root compartment. Five shrubs were sampled in each plantation, giving five unpooled subsamples per operational sample type, 20 microbiome samples per plantation age class and 80 samples in total. Every collected microbiome sample was subjected to a separate DNA extraction, so the number of DNA extractions was identical to the number of field samples. No field samples or DNA extracts were pooled; only triplicate PCR products from the same DNA sample were pooled during library preparation. This hierarchy is now stated explicitly in the Methods.
Revision: Revised Section 2.1.
Comment 6. Use 5Y, 10Y, 20Y and 30Y rather than the additional G1–G4 notation to improve readability.
Response: We agree with the reviewer’s preference. The revised title, text and figure captions use the stand ages directly rather than interpreting G1–G4 as an additional biological classification. Because the embedded figures in the supplied DOCX are raster images, the tracked manuscript flags Figures 3 and 4 for regeneration from the original plotting source with 5Y/10Y/20Y/30Y labels. The same convention should be applied consistently to the remaining source figures before resubmission.
Revision: Terminology standardized throughout the revised text and captions; raster artwork regeneration is flagged in the tracked manuscript.
Comment 7. Report PCA loadings for the soil variables and address the relatively short environmental vectors in Fig. 1C.
Response: We agree. The revised Methods now state explicitly that the loadings of SWC, TN, SOM, EC, pH and TP on PC1 and PC2 must be reported, and the Results/Figure 1 legend directs the reader to the full loading table. The actual loading coefficients are not present in the supplied manuscript or figure at a precision that can be recovered reliably, so exact values are flagged for insertion from the original PCA output rather than reconstructed from the graphic.
Revision: Revised Sections 2.3 and 3.1 and Figure 1 legend (exact loading values required from the original PCA output).
Comment 8. Consider retaining “PCA-derived edaphic axis” or “Edaphic Gradient Index”; the term “restoration” requires stronger justification.
Response: We agree and have removed the term “restoration” from the index. The former ERGI is now called the PCA-derived Edaphic Gradient Index (EGI), and the manuscript explicitly states that EGI is an author-derived descriptive summary of multivariate soil chemistry, not a validated restoration index.
Revision: Revised Abstract, Sections 2.3 and 3.1, Figure 1 legend, and Discussion.
Comment 9. The chronosequence demonstrates differences among stands, not temporal causation; wording such as stand development “reorganized” the microbiome is too causal.
Response: We fully agree. This concern became even more important after clarifying that one plantation represents each age class. We therefore changed the title to a chronosequence-based descriptive formulation and removed causal language throughout. The manuscript now states explicitly that the design is a space-for-time substitution, that age is confounded with plantation/site, and that the five shrubs are within-stand subsamples. Accordingly, the results document differences across the four sampled plantations rather than demonstrating temporal development or a replicated stand-age effect.
Revision: Title, Abstract, end of Introduction, Sections 3.1–3.3, Section 4, and Conclusions.
Comment 10. PERMDISP was mentioned in Methods but its actual results were not reported; PERMANOVA must be interpreted together with dispersion diagnostics.
Response: We agree. The revised Results now contains a dedicated PERMDISP sentence immediately after the PERMANOVA results and explains why dispersion must be considered when interpreting centroid separation. The exact F statistics and permutation P values are not included in the supplied manuscript, so they are explicitly flagged for insertion from the original betadisper/PERMDISP output. We also expanded the Methods to identify the vegan functions used for PERMANOVA and dispersion testing, subject to verification against the archived analysis script.
Revision: Revised Sections 2.4, 3.2 and 4.
Comment 11. The manuscript does not directly demonstrate that soil chemistry is associated with or explains microbial community composition, although the title/narrative attributes reorganization to edaphic change.
Response: We agree. The title no longer attributes microbiome change mechanistically to edaphic reconfiguration, and the Discussion now distinguishes co-occurring edaphic and microbial patterns from a demonstrated environment–community causal relationship. Given both the matched-subset limitation and the one-plantation-per-age design, we consider this conservative reframing more appropriate than overinterpreting an additional constrained ordination. If fully matched sample-level environmental and ASV matrices are available, a direct exploratory dbRDA/RDA can still be added, but it would not resolve the age–site confounding and would therefore be interpreted cautiously.
Revision: Revised title, Sections 3.1 and 4; optional direct environment–community analysis flagged for addition if the matched data are available.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsReview for: Stand Age Reorganizes Caragana korshinskii Root-Associated Microbiomes through Edaphic Reconfiguration, Hierarchical Filtering and Domain-Specific Assembly
By Su et al.
The authors present an interesting study on how soil- and root-associated bacterial and fungal communities differ across shrub stands of different ages. They examine several potential mechanisms underlying these patterns, including edaphic reconfiguration, hierarchical filtering, and domain-specific assembly. The central aim of the study is to disentangle the complex drivers of bacterial and fungal richness across compartments—bulk soil, the rhizosphere, and the endosphere—as well as across stand-age categories. The authors clearly describe the observed statistical patterns in bacterial and fungal communities, while also acknowledging that further analyses are needed to assess potential functional or causal relationships. Overall, the manuscript is well written, the findings are interpreted cautiously, and the applied genomic and statistical methods are appropriate. However, I recommend adding more information on background of different indices/metrics used and how these were calculated, which software packages (+ versions) were used, and adding relevant references to the highlighted paragraphs below and to the reference list.
Specific comments:
Line 19: Write ASV in full as it is the first time this abbreviation is mentioned in the text.
Line 84: The abbreviation ERGI should be written in full here (= environmental restoration gradient index).
Line 91: Caragana korshinskii – please change font to italics.
Figure 1b – Explain abbreviations (e.g., SWC, TN, SOM, etc.) in the figure legend to make it easier for the reader to follow.
Lines 107 – 113: I recommend adding somewhere in this paragraph that the patterns of decreasing ASV and bacterial soil-pool richness was similar across restoration stages (Fig. 2b).
Figures 3 + 4: In these figures, you are using G1 to G4 for the different restoration stages while in Fig. 2 you are using age categories. I recommend using the same age categories in Fig. 3 and 4 as I believe this is easier to follow for the audience. Also, in Fig. 2, you have pooled J10 + J30 while in Fig. 3 + 4, these remain separate. Please add an explanation why this was done.
Lines 197 – 202 and lines 293 - 294: Please use italics for species/genera names.
Line 370: Please provide references for the two primer pairs (i.e., 799F/1115R and ITS1F/ITS2R) here and add full references to reference list.
Lines 391 – 397: Which programs or R modules/functions were used to do the described analyses (e.g., PCA, Spearman rank correlation, and Kruskal-Wallis tests)? Please provide names for these and version number(s) as well as relevant references here and in the reference list.
Lines 406 – 414: Which programs or R modules/functions were used to do the described analyses (e.g., observed richness, Shannon diversity, and PERMANOVA)? Please provide names for these and version number(s) as well as relevant references here and in the reference list.
Lines 416–417: I recommend adding brief explanations of βNTI and RCbray to make these metrics more accessible to readers. For example:
Beta-Nearest Taxon Index (βNTI) quantifies whether the phylogenetic relatedness among taxa in different communities deviates from what would be expected under a random null model, thereby helping to infer the strength and direction of deterministic selection. Bray-Curtis-based Raup-Crick (RCbray) uses taxonomic abundance data to assess whether community dissimilarity is greater or lower than expected by chance, particularly in cases where deterministic selection is weak or absent.
Lines 424 – 427: In addition to the previous comment, I recommend adding short explanations what the different categories indicate. For example (but please check for accuracy):
Line 425: heterogenous selection (shows greater phylogenetic turnover than expected. This means variable selection (different environmental pressures) causes the communities to filter for different lineages).
Line 425: homogeneous selection (shows less phylogenetic turnover than expected. This means homogeneous selection (similar environmental pressures) drives both communities to share closely related taxa)
Line 426: dispersal limitation (movement between sites is restricted, making communities different)
Line 427: homogenizing dispersal (high movement/migration between sites makes communities alike)
Line 427 – line 428: undominated (driven by weak selection, weak dispersal, diversification, or pure ecological drift).
Lines 415 – 429: Which programs or R modules/functions were used to do the described analyses? Please provide names for these and version number(s) as well as relevant references here and in the reference list. Also, provide relevant references for the indices used. For betaMNTD, I suggest Stegen et al. (2013):
Stegen, J. C., Lin, X., Fredrickson, J. K., Chen, X., Kennedy, D. W., Murray, C. J., ... & Konopka, A. (2013). Quantifying community assembly processes and identifying features that impose them. The ISME journal, 7(11), 2069-2079.
Lines 433-434: Which programs or R modules/functions were used to calculate IndVal? Please provide names for these and version number(s) as well as relevant references here and in the reference list.
Line 443: Which programs or R modules/functions were used to calculate Pairwise Spearman correlations? Please provide names for these and version number(s) as well as relevant references here and in the reference list.
Author Response
We thank the reviewer for the positive assessment and for the detailed methodological suggestions. We have expanded the Methods to make the indices and software workflows more accessible and reproducible, while avoiding the introduction of unverified software-version information.
Comment 1. Line 19: spell out ASV at first mention.
Response: Corrected. The Abstract now reads “amplicon sequence variants (ASVs)” at first use.
Revision: Revised Abstract.
Comment 2. Line 84: write ERGI in full.
Response: We revised this more substantially in response to Reviewer 1. “ERGI” has been replaced by “PCA-derived Edaphic Gradient Index (EGI)” and is defined at first use. The word “restoration” was removed from the index name to avoid implying that the PCA axis is an independently validated restoration metric.
Revision: Revised Sections 2.3 and 3.1.
Comment 3. Line 91: italicize Caragana korshinskii.
Response: Corrected. The species name and genus/species names introduced in the revised text are italicized.
Revision: Throughout the revised manuscript.
Comment 4. Figure 1B: define SWC, TN, SOM and other abbreviations in the legend.
Response: Corrected. The Figure 1 legend now defines SWC, TN, SOM, EC and TP; pH is retained as the standard chemical notation.
Revision: Revised Figure 1 legend.
Comment 5. Lines 107–113: mention that the decline in ASV/soil-pool richness was similar across restoration stages.
Response: Agreed. We added a sentence noting that the direction and magnitude of soil-to-root ASV contraction were broadly similar across the four stand-age classes, supporting persistent hierarchical compartment filtering rather than an age-specific effect.
Revision: Revised Section 3.1, paragraph describing Figure 2.
Comment 6. Figures 3 and 4 should use the same age labels as Figure 2; also explain why Soil 10 cm and Soil 30 cm are pooled in Figure 2 but separate in Figures 3 and 4.
Response: We agree. The revised text and captions use direct age labels (5Y, 10Y, 20Y and 30Y), and the raster figures are flagged for regeneration from source. We also added an explicit methodological explanation: Soil 10 cm and Soil 30 cm are combined only in the filtering analysis to define the upstream soil reservoir for retention/filtering calculations, whereas they remain separate in beta-diversity and null-model analyses to retain depth/position-specific community structure.
Revision: Revised Section 2.4 and Figure 3/4 legends; figure-source regeneration flagged.
Comment 7. Lines 197–202 and 293–294: italicize species/genera names.
Response: Corrected. The candidate bacterial and fungal genera in the Results and Discussion are italicized, including Kineococcus, Massilia, Rhizobium, Promicromonospora, Nocardia, Pseudomonas, Mucor, Neosetophoma, Mortierella, Clonostachys, Exophiala, Tricellula and Papiliotrema.
Revision: Revised Sections 3.3 and 4.
Comment 8. Line 370: provide references for primers 799F/1115R and ITS1F/ITS2R.
Response: Corrected. We added references for 799F (Chelius and Triplett, 2001), 1115R (Redford et al., 2010), ITS1F (Gardes and Bruns, 1993), and the commonly used ITS2 reverse primer source (White et al., 1990). Because the manuscript names the reverse primer “ITS2R” without providing its sequence, we also flagged the exact primer name/sequence for verification against the laboratory record before final submission.
Revision: Revised Section 2.2; four primer references added to the reference list.
Comment 9. Lines 391–397: specify programs/R packages/functions and versions for PCA, Spearman correlation and Kruskal–Wallis tests.
Response: We agree. Section 2.3 now identifies the analyses as R-based and explicitly requests the exact R version and functions/packages from analysis script.
Revision: Revised Section 2.3.
Comment 10. Lines 406–414: specify software/modules/functions for observed richness, Shannon diversity and PERMANOVA.
Response: Expanded. Observed richness is now operationally defined as the number of non-zero ASVs per sample. The revised Methods specify vegan 2.6-4 and the relevant functions for Shannon diversity (diversity), Bray–Curtis dissimilarity (vegdist), PERMANOVA (adonis2), and dispersion testing (betadisper plus permutation testing). The authors should verify these function calls against the archived script before resubmission.
Revision: Revised Section 2.4.
Comment 11. Lines 416–417: briefly explain βNTI and RCbray.
Response: Added. βNTI is now defined as the standardized deviation of observed βMNTD from a phylogenetic null expectation, and RCbray is described as an abundance-based Raup–Crick metric used to evaluate taxonomic dissimilarity when |βNTI| does not indicate dominant selection.
Revision: Revised Section 2.5.
Comment 12. Lines 424–428: explain heterogeneous selection, homogeneous selection, dispersal limitation, homogenizing dispersal and undominated processes.
Response: Added. Each category is now defined in ecological terms immediately after the decision thresholds. We retained conservative wording so that these categories are interpreted as null-model assignments rather than direct measurements of environmental pressure or dispersal rate.
Revision: Revised Section 2.5.
Comment 13. Lines 415–429: specify software/packages/functions for the null-model analysis and add references such as Stegen et al. (2013).
Response: Expanded. The revised section cites Stegen et al. (2012, 2013) and adds the picante reference (Kembel et al., 2010) for phylogenetic-community calculations. The exact package version and number of null randomizations are flagged for confirmation from the original analysis environment because these details are not contained in the supplied manuscript.
Revision: Revised Section 2.5 and References.
Comment 14. Lines 433–434: specify how IndVal was calculated, with software/version and references.
Response: Expanded. The revised Methods identify the indicspecies/IndVal framework and add De Cáceres and Legendre (2009) to the reference list. The exact package version and function (e.g., multipatt()) are flagged for verification. We also retain the important result that no ASV passed FDR-adjusted significance, so the candidate-indicator analysis remains explicitly exploratory.
Revision: Revised Section 2.6 and References.
Comment 15. Line 443: specify software/modules/functions for pairwise Spearman correlations.
Response: Expanded. The network Methods now state that pairwise Spearman correlations were calculated in R and request the exact function/package/version and multiple-testing implementation from the archived analysis script. The correlation threshold and FDR criterion are retained, and the manuscript continues to describe the networks as statistical association graphs rather than direct interactions.
Revision: Revised Section 2.6.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe manuscript has been substantially modified and improved, which should be strongly appreciated. Most of the weaknesses have been addressed to the extent permitted by the adopted research design. However, I wonder whether constructing four separate networks based on n = 5 samples per age class is statistically too unstable, and whether these results should instead be presented as an exploratory analysis or additionally validated through network stability analyses.
Please carefully review the entire text and standardize the terminology and formatting. Within a relatively short section of the manuscript, different terms are used to describe the same concept, such as “age/site classes,” “stand-age classes,” and “age class/site.” Please use the terminology consistently throughout the manuscript.
I noticed an encoding error in Fig. 1E affecting the unit “g kg−1” in the revised version. The unit is correctly displayed in the working version of the manuscript.
I think it would be worthwhile to add one sentence to the Materials and Methods explaining why the authors decided to collect samples from a depth of 25–30 cm.
Author Response
Comment 1: The manuscript has been substantially modified and improved, which should be strongly appreciated. Most of the weaknesses have been addressed to the extent permitted by the adopted research design. However, I wonder whether constructing four separate networks based on n = 5 samples per age class is statistically too unstable, and whether these results should instead be presented as an exploratory analysis or additionally validated through network stability analyses.
Response: We agree with the reviewer that n = 5 samples per plantation age/site class is insufficient to regard four separately inferred Spearman association networks as statistically stable estimates of microbial network structure. This concern is well supported by the network-inference literature. Berry and Widder (2014) recommended using at least approximately 25 samples for microbial co-occurrence networks, and a recent review similarly noted that approximately 25–30 samples per group is a reasonable minimum for correlation-based microbiome network analysis. More recent resampling work has further shown that Spearman-network attributes, including node number and average links, can be strongly sample-size dependent and often require substantially larger sample sets to stabilize. We therefore chose the reviewer’s first suggested solution: the four networks are retained only as an exploratory, hypothesis-generating analysis.
Response: We did not present bootstrap or leave-one-out resampling of the same five observations as a formal validation, because such procedures cannot create additional independent biological information and could give an unwarranted impression of network stability at this very small effective sample size. Instead, we have substantially softened the network interpretation throughout the manuscript. Network edge counts, mean degree, modularity, edge-type proportions, and hub/connector assignments are now described only as descriptive patterns obtained under identical filtering and thresholding rules. We no longer treat the apparent 10-year increase in retained connectivity as a statistically validated temporal transition, nor do we interpret hub/connector taxa as validated keystone microorganisms.
Comment 2: Please carefully review the entire text and standardize the terminology and formatting. Within a relatively short section of the manuscript, different terms are used to describe the same concept, such as “age/site classes,” “stand-age classes,” and “age class/site.” Please use the terminology consistently throughout the manuscript.
Response: Thank you for identifying this inconsistency. We performed a manuscript-wide terminology and formatting audit. Because each chronological age group is represented by one plantation and plantation age is therefore fully confounded with site, we now consistently use the term “plantation age/site class(es)” when referring to the 5-, 10-, 20-, and 30-year groups. The four soil–root sampling categories are consistently referred to as “microhabitats.” We removed interchangeable uses of “stand-age classes,” “age class/site,” “stage,” and “niche” when these terms referred to the same grouping variables.
Response: We also standardized scientific formatting throughout the manuscript, including italicization of Caragana korshinskii and genus/species names, capitalization of the operational “Root compartment” category, spacing between values and units, use of R², βNTI and RCbray notation, multiplication and minus signs, and consistent figure-panel and supplementary-table references.
Comment 3: I noticed an encoding error in Fig. 1E affecting the unit “g kg−1” in the revised version. The unit is correctly displayed in the working version of the manuscript.
Response: Thank you for noticing this production/encoding error. We restored the correctly rendered unit from the working version and replaced the affected Figure 1 artwork in the revised manuscript. The unit is now displayed as “g kg−1” in Figure 1E. We also checked the remaining labels and units in Figure 1 to ensure that no additional encoding artifacts were introduced during figure export or document conversion.
Comment 4: I think it would be worthwhile to add one sentence to the Materials and Methods explaining why the authors decided to collect samples from a depth of 25–30 cm.
Response: We agree and have added the requested rationale in Section 2.1. The 25–30 cm interval was selected primarily on the basis of preliminary field excavation at the study sites, where fine roots suitable for standardized root and rhizosphere sampling were consistently encountered in this layer. This choice is also ecologically reasonable for C. korshinskii: previous field studies show that a substantial proportion of its root system is concentrated in shallow subsurface soil, and Liu et al. (2022) specifically reported root-hair-rich branches at approximately 30–40 cm and used that observation to define rhizosphere sampling depth. We therefore clarified that the chosen interval targeted an active root-associated layer while maintaining the same vertical sampling position across all four plantations.
Author Response File: Author Response.pdf
