Using network analysis as a tool for treating human skin ecological disorders

Abstract

This invention discloses a method for identifying bacterial strains suitable for treating human skin dysbiosis. The method includes the step of performing network analysis by computer to determine the correlation between the bacterial strain and at least a second bacterial strain in dysbiosis and non-dysbiosis, wherein the correlation differs between the dysbiosis and the non-dysbiosis, and wherein the correlation means positive or negative correlation. Further, the network is generated by co-occurrence analysis of the abundance of the bacterial strain and the second bacterial strain, the co-occurrence analysis being performed by DNA sequencing based on 16S rRNA amplicon or whole genome sequencing.

Description

Technical Field

[0001] This invention relates to the field of human microbiome, and particularly to the field of network analysis for treating ecological imbalances affecting human skin. Background Technology

[0002] The structure and function of the human microbiome are largely influenced by microbe-microbe and microbe-host interactions. Understanding these interactions is crucial for understanding microbiome function. Network theory is a method for modeling complex biological systems such as microbiomes, where members interact in multiple ways. Complementing traditional approaches focused on isolated individual microbial members, network theory provides a holistic approach to multi-microbe interactions within the microbiome community and its interaction with its host. Network-based analysis allows researchers to model and analyze complex multi-microbe interactions within a single network.

[0003] In typical association-based network models of microbial systems, network nodes represent taxa, and edges represent positive or negative associations between taxa within a defined population. Microbiome network analysis can include describing the structural characteristics of microbial communities, as well as key network topologies such as central nodes, connectors, and modules. Microbiome community building is driven by dynamic ecological and evolutionary processes. Applying network theory to microbiome research can help reveal the microbial relationships necessary for community building or stability, as well as potential pathogenic effects on the host. Known levels of interaction within a community can predict ecological stability and resilience. Sudden changes in the abundance of a particular microbial genus may signal a transition from a healthy state to a diseased state, or vice versa.

[0004] Extensive methods have been used to construct microbiome networks. The most popular approach is based on correlation techniques to discover significant pairwise associations by calculating correlation coefficients such as the Pearson correlation coefficient or the Spearman nonparametric rank correlation coefficient. However, these methods have limitations with microbiome data, including inappropriate correlations due to compositional factors and severe lack of dynamics due to the quality of zero counts. The focus on correlation-based analysis has led to the development of compositionally robust methods. SparCC (Sparse Correlations for Compositional Data) is a methodology that uses linear Pearson correlations between logarithmically transformed components to infer associations in compositional data. SPIEC-EASI (Sparse InversE Covariance Estimation for Ecological Association Inference) is another statistical approach for inferring microbial ecological networks, combining data transformations developed for compositional data analysis with a graphical model inference framework, assuming the underlying ecological association network is sparse.

[0005] WO 14205088 A1 (Prodermiq) discloses a method for characterizing the microbiome of a subject's skin or subcutaneous tissue. The method includes a) obtaining a sample containing multiple microorganisms from the subject's skin or subcutaneous tissue; and b) analyzing and classifying the multiple microorganisms from (a) to characterize the subject's microbiome. In some embodiments, the method further includes comparing the subject's microbiome with a reference microbiome, or generating a microbiome profile of the subject, or identifying a disease or disorder that the subject has or is at risk of developing, or providing the subject with a personalized treatment plan. In various embodiments, the reference microbiome is classified as having a healthy profile, and the similarity between the subject's microbiome and the reference microbiome identifies the subject's microbiome as having a healthy profile. Alternatively, the reference microbiome is classified as having a disease or disorder or being at risk of having a disease or disorder, and the similarity between the subject's microbiome and the reference microbiome identifies the subject's microbiome as having a disease or disorder or being at risk of having a disease or disorder.

[0006] NAKATSUJI T. Antimicrobials from human skin commensal bacteria protect against Staphylococcus aureus and are deficient in atopic dermatitis. SciTransl Med. 2017 Feb 22; 9(378) discloses that the microbiome can promote or disrupt human health by influencing both adaptive and innate immune functions. The authors tested whether bacteria commonly found on human skin participate in host defense by killing Staphylococcus aureus (S. aureus), a pathogen commonly found in patients with atopic dermatitis (AD) and a significant factor exacerbating the disease. High-throughput screening for antimicrobial activity against Staphylococcus aureus was performed on isolates of coagulase-negative Staphylococcus (CoNS) collected from the skin of healthy and AD subjects. CoNS strains with antimicrobial activity were common in the normal population but rare in AD subjects. The low frequency of antimicrobial active strains was associated with Staphylococcus aureus colonization. Antimicrobial activity was identified as previously unknown antimicrobial peptides (AMPs) produced by CoNS species, including Staphylococcus epidermidis and Staphylococcus hominis. These AMPs are strain-specific, potent, and selectively kill Staphylococcus aureus and synergize with human AMP LL-37. Application of these CoNS strains to mice confirmed their in vivo defensive function relative to the application of inactive strains. Notably, reintroduction of antimicrobial CoNS strains to human subjects with Alzheimer's disease (AD) reduced Staphylococcus aureus colonization. These findings demonstrate how symbiotic skin bacteria fight pathogens and show how dysbiosis of the skin microbiome can contribute to disease.

[0007] US 2018311144 A (NATURACOSMETICOS SA) discloses a probiotic cosmetic preparation for skin cosmetic treatment, which contains a microbiome and preferably betaine and / or prebiotics.

[0008] WO 11022660 A1 (Vedanta Biosciences) discloses methods for diagnosing and treating microbiome-related diseases or improving health, which use interaction network parameters to analyze interaction networks between microorganisms and between microorganisms and the host to identify important (e.g., “highly relevant”) organisms or molecules, as determined by various network parameters. The provided methods include and extend beyond correlation to use these important (e.g., “highly relevant”) organisms or molecules as targets for regulation or as therapeutic agents to improve health. The publication also discloses products containing microbiome modulators, probiotics, or other therapeutic agents derived from these important “highly relevant” organisms or molecules to improve health.

[0009] This publication only considers the relevance under one condition.

[0010] However, the inventors believe that following such a method cannot help identify all potential target organisms that may intervene in the subjects. Changes in network parameters between dysbiosis and non-dysbiosis conditions can help identify some more potential intervention targets. Summary of the Invention

[0011] According to the present invention, a method for identifying bacterial strains suitable for treating human skin dysbiosis is disclosed. The method includes a step of performing network analysis via computer to determine whether the bacterial strain is correlated with at least a second bacterial strain in both dysbiosis and non-dysbiosis conditions, wherein the correlation varies between the dysbiosis and non-dysbiosis conditions, and wherein the correlation is either positive or negative. Further, the network is generated by co-occurrence analysis of the abundance of the bacterial strain and the second bacterial strain, the co-occurrence analysis being performed by DNA sequencing based on 16S rRNA amplicon sequencing or whole-genome sequencing. Detailed Implementation

[0012] As used herein, the term “comprising” encompasses the terms “substantially consisting of” and “consisting of”. When using the term “comprising,” the listed steps or options need not be exhaustive. Unless otherwise specified, a range of values ​​expressed in the form “x to y” should be understood to include both x and y. When specifying any range of values ​​or quantities, any particular upper limit value or quantity may be associated with any particular lower limit value or quantity. Except in examples and comparative experiments, or where otherwise expressly indicated, all figures should be understood to be modified by the word “about.” Unless otherwise specified, all percentages and ratios contained herein are by weight. As used herein, unless otherwise specified, the indefinite article “a” or “an” and its corresponding definite article “described” mean at least one, or one or more. Various features of the invention referenced in the foregoing sections are suitably applied to other sections with necessary modifications. Thus, a feature specified in one section may be suitably combined with features specified in other sections. The addition of any section headings is merely for convenience and is not intended to limit this disclosure in any way. The embodiments are intended to illustrate the invention and are not intended to limit the invention to those embodiments alone.

[0013] The term "microbiota" generally refers to all microorganisms found in association with higher organisms, such as humans. Organisms belonging to the human microbiota can generally be classified into bacteria, archaea, yeasts, and single-celled eukaryotes, as well as viruses and various parasites such as worms.

[0014] The term "microbiome" generally refers to all microorganisms, their genetic factors (genomes), and the environmental interactions found in relation to higher organisms such as humans.

[0015] The term "symbiont" refers to an organism that is generally harmless to its host and can also establish a mutually beneficial relationship with it. The human body contains approximately 100 trillion symbionts, and it has been shown that the number of these organisms is 10 times greater than the number of human cells.

[0016] The term "microbial-derived component" refers to a component that is composed of, originates from, or is produced by members of a microbial community. Components can be, for example, microorganisms, microbial proteins, microbial secretions, or microbial fractions.

[0017] The term "network" refers to the structural representation of components (components of host or microbial origin) that describes the relationships between components through various methods.

[0018] The term "node" refers to the endpoint or intersection of a network's graphical representation. It is an abstraction of elements such as organisms, proteins, genes, transcripts, or metabolites.

[0019] The term "edge" refers to a link between two nodes. A link is an abstraction of the connection between nodes, such as the covariance between nodes.

[0020] The term "highly relevant organism" refers to a key functional member of the microbiome that has edge connections with a large number of nodes in the network. For example, a microbial species can perform biotransformation of multiple metabolites and thus may affect host metabolism and host health.

[0021] The term “metagenomics” refers to genomic techniques that study microbial communities directly in their natural environment without the need to isolate and culture individual species in a laboratory.

[0022] Network analysis has long been present in many fields of computational biology and bioinformatics, including genomics, proteomics, and metabolomics, but it has not yet been widely applied to microbial community research. Network analysis can uncover potential functional relationships between microbial individuals through interactions between them. This method is based on species abundance used in traditional ecological analyses, and interspecies interactions are included in the analysis. Several studies have successfully applied network analysis to human microbial communities. Traditional network properties (such as network density) largely ignore information about nodes and edges in the network. In microbial community interaction networks, each node represents a species or an operational taxonomic unit (OTU), while edges represent interactions between two species or OTUs. In community interaction networks, negative relationships between two entities indicate resistance or inhibition, while positive relationships indicate synergy. These interactions are present in the realization of the microbial network's function. Greater importance, as well as the occurrence and development of diseases, will affect the realization of network function, leading to changes in interspecies interactions within the network.

[0023] According to a first aspect of the present invention, a method for determining whether a bacterial strain is a probiotic suitable for treating human skin dysbiosis is disclosed, the method comprising performing a network analysis to determine whether the bacterial strain is correlated with at least a second bacterial strain in both dysbiosis and non-dysbiosis conditions, wherein the bacterial strain is considered suitable if there is a difference in the correlation between the dysbiosis and the non-dysbiosis conditions.

[0024] Preferably, the difference in correlation between the dysbiosis condition and the non-dysbiosis condition manifests as a shift from high correlation to low correlation, or vice versa. More preferably, the correlation is high in the dysbiosis condition and low in the non-dysbiosis condition. Alternatively, the correlation is low in the dysbiosis condition and high in the non-dysbiosis condition. Preferably, the difference in correlation is at least 40%. More preferably, the correlation is at least the top 5% of the total identified bacterial strains in the non-dysbiosis skin. Even more preferably, the correlation is at least the top 5% of the total identified bacterial strains in the non-dysbiosis skin, and the difference in correlation between the non-dysbiosis state and the dysbiosis state is at least 40%. More preferably, in this case, the dysbiosis condition is acne.

[0025] According to the method of the present invention, correlation means either positive or negative correlation. Therefore, in one aspect of the invention, correlation means positive correlation. Alternatively, correlation means negative correlation.

[0026] Particularly preferred are bacterial strains and the second bacterial strain that are strains found on human skin.

[0027] Network Analysis

[0028] The preferred approach is to perform network analysis using Sparse InversEcovariance estimation in SPIEC-EASI version 0.1.2. SPIEC-EASI (Sparse InversE Covariance Estimation for Ecological Association Inference) is a statistical method for inferring microbial ecological networks from amplicon sequencing datasets, addressing both of these issues. SPIEC-EASI combines data transformations developed for component data analysis with a graphical model inference framework, assuming the underlying ecological association network is sparse. To reconstruct the network, SPIEC-EASI relies on algorithms for sparse neighborhood and inverse covariance selection.

[0029] Alternatively, the analysis can be performed using any other equivalent technique or computer program.

[0030] Additionally, Cytoscape (version 3.5.1) is preferably used to visualize the OTU network and calculate network correlations. Cytoscape is an open-source software platform for visualizing molecular interaction networks. The Cytoscape core distribution provides a set of basic features for data integration, analysis, and visualization. Any other alternative software program can also be used here. Preferably, network analysis is set at the OTU (Operational Taxonomic Unit), ASV (Action Script Viewer), species, or genus level. More preferably, network analysis is set at the OTU level.

[0031] More preferably, the network analysis is generated by co-occurrence analysis of the abundance of the first and second bacterial strains, the co-occurrence analysis being performed by DNA sequencing via 16S rRNA amplicon or whole genome sequencing.

[0032] Preferably, the bacterial strain is derived from at least one of the genera Staphylococcus, Streptococcus, Microbacterium, Methyloversatilis, Deinococcus, Moraxella, or Acinetobacter.

[0033] Similarly, the preferred second bacterial strain is derived from the genera *Acidovorax*, *Actinomyces*, *Bacillus*, *Chryseobacterium*, *Corynebacterium*, *Fusobacterium*, *Staphylococcus*, *Streptococcus*, *Microbacterium*, *Methylobacterium*, *Methyloversatilis*, and others. The bacterial strain contains at least one of the genera *Deinococcus*, *Micrococcus*, *Moraxella*, *Neisseria*, *Paracoccus*, *Prevotella*, *Pseudomonas*, *Sphingomonas*, *Acinetobacter*, or *Cutibacterium*, wherein if the genus of the second bacterial strain is the same as that of the bacterial strain, the species are different; and wherein if the genus and species are the same, the strains are different.

[0034] Preferably, the dysbiosis condition includes at least one of acne, dandruff, dry skin, aging skin, pigmented skin, or inflammation. More preferably, it is acne.

[0035] More preferably, when the dysbiosis is acne, the bacterial strain is from the genera Staphylococcus, Streptococcus, Microbacterium, Methyloversatilis, Deinococcus, Moraxella, or Acinetobacter.

[0036] More preferably, the bacterial strain from the genus Staphylococcus is Staphylococcus hominis or Staphylococcus epidermidis.

[0037] The invention will now be explained with the aid of non-limiting embodiments.

[0038] Example

[0039] Example 1:

[0040] The procedures followed are disclosed below.

[0041] Facial buffer wash samples (approximately 4 ml) were obtained from 35 acne-affected subjects and 32 non-acne-affected subjects using an 18 mm diameter column. The buffer samples were stored at -20°C prior to analysis. Microbial precipitates in the samples were washed away by centrifugation at 4°C (10 min / 13,000 rpm, Eppendorf 5810R). The precipitates were resuspended in 180 μl of fresh enzyme lysis buffer (20 mg / ml lysozyme (Sigma L6876), at 20 mM Tris·Cl, pH 8.0, 2 mM sodium EDTA, and 1.2%...). (X-100). Incubate the mixture at 37°C for 30 minutes, then add 25 μl of proteinase K and 200 μl of Buffer AL (ethanol-free). Incubate further at 56°C for 30 minutes.

[0042] Next, 100 μl of acid-washed glass beads (Sigma, G8772) were added to the tube and homogenized using MP Biomedicals FastPrep-24 (5 m / s, 45 sec, twice). Then, DNA was extracted from the sample using a DNA extraction kit (Qiagen, DNeasy Blood & Tissue kit, 69506) according to the manufacturer's instructions. The variable regions V1-V3 of the 16S rRNA gene in the microbial DNA were sequenced for bacterial classification. First, the 16S V1-V3 genes were amplified using primer pairs recommended by the sequencing company (forward: AGAGTTTGATYMTGGCTCAG, reverse: ATTACCGCGGCTGCTGG). A 50 μl PCR reaction system consisted of 1 μl (10 μM) of each forward and reverse primer, 5 μl of 10x PCR buffer, 2 μl of MgCl2 (50 mM), 1.5 μl of dNTP mixture (10 mM), 2 μl of DNA template, 0.2 μl of Platinum Taq DNA polymerase (Invitrogen), and 35.3 μl of molecular-grade water (Sigma, W4502). For the V1-V3 regions, samples were amplified using the following parameters: 94 °C for 2 min, 20 cycles: 94 °C for 30 s, 65 °C (decreasing by 0.5 °C per cycle) for 1 min; 10 cycles: 94 °C for 30 s, 55 °C for 1 min; and 72 °C for 5 min. To reduce PCR amplification bias, each sample was tested in triplicate. Library construction was performed using a two-step PCR method. The second round of PCR (8 cycles) was performed by BGI using fusion primers with double-indexed sequences and adaptors. Products were purified using Ampure beads. Library quantity and quality were analyzed using Bioanalyzer (Agilent Technologies). Sequencing was performed on an Illumina Miseq PE300 platform using only qualified libraries.

[0043] Using Vsearch v1.9.6, the denoised sequencing data were clustered into Operational Taxonomic Units (OTUs), with a cluster identity of 0.97 and a minimum cluster size of 10. The Least Common Ancestor (LCA) methodology was used to taxonomically classify the OTUs against the SILVA, NCBI, RDP, DDBJ, Greengenes, CAMERA, EzBioCloud, and EMBL databases.

[0044] The OTU table and associated representative sequences (selected as the most abundant sequences within the OTU clusters) were used as input to QIIME

[46] version 1.9.1 (Quantitative Insights into Microbial Ecology), an open-source software package for analyzing complex microbial communities. The output of QIIME was used as input to a phyloseq object in R. For genus networks, the OTU (operational taxonomic unit) table was merged into the genus level before subsequent analyses. The OTU prevalence threshold for OTU network analysis was set to at least 50%.

[0045] Network analysis was performed using Sparse Inverse Covariance Estimation (SICO) in SPIEC-EASI version 0.1.2. This method estimates the inverse covariance matrix from sequencing data (Gaussian graphical model). A sparse network was generated using the lasso method. Neighborhood and covariance selection methods were employed, including Meinshausen and Buhlmann (MB) and StARS (Stability Approach to Regularization Selection), with a maximum threshold of 0.01 for λ.

[0046] Use Cytoscape (version 3.5.1) to visualize the OTU network and calculate network correlations.

[0047] For the purposes of this experiment, Staphylococcus and Cutibacterium are the most abundant genera in the healthy (non-dysbiosis) and acne (dysbiosis) microbiomes.

[0048] To further explore the interactions between these two genera, a subnetwork of OTUs was constructed using Staphylococcus and Cutibacterium.

[0049] As summarized in Table 1, the correlations of several OTUs change in subnetworks between acne and non-acne microbiomes.

[0050] OTU 722, the most abundant OTU of Propionibacterium acnes, is a central node in acne. Its relative abundance did not change significantly between acne and healthy conditions, but its correlation with other bacteria was lower in healthy conditions (score = 4), but the correlation changed to a higher correlation in acne conditions (score = 8).

[0051] We hypothesize that the increased correlation is a potential indicator of the pathogenicity of this strain in acne. In contrast, *Staphylococcus epidermidis* (OTU 24) is generally considered a commensal bacterium on the skin. Its relative abundance increases in acne conditions, but its correlation with other bacteria decreases (ranging from 10 to 6), suggesting that *Staphylococcus epidermidis* may have a different ecological role than *Propionibacterium acnes* in both acne and healthy skin.

[0052] We also observed that Staphylococcus hominis had a high correlation with other bacteria in the genus Staphylococcus and the Cutibacterium subnetwork.

[0053] OTU 2719 and OTU 227 of Staphylococcus aureus were highly correlated in healthy skin conditions (scores = 9 and 10, respectively), but their correlation decreased in acne or dysbiosis conditions (scores = 5 and 2, respectively).

[0054] Subnetwork analysis reveals the possible ecological roles of microorganisms, which may not always be reflected in their relative abundance.

[0055] Table 1

[0056]

[0057] Note: In Table 1, *P<0.05

[0058] Staphylococcus hominis is considered a minor species in the skin microbiome (average relative abundance <1%). Therefore, its functional relevance may be overlooked if findings from conventional microbiome analyses are followed. However, based on the data indicated in Table 1, Staphylococcus hominis may be a beneficial bacterium in acne-prone skin.

[0059] Example 2

[0060] By following the method disclosed in Example 1 with appropriate modifications, the relative abundance and extent of other bacteria that may be beneficial in treating acne were observed. The findings are summarized in Table 2.

[0061] Table 2

[0062]

[0063] *P<0.05

Claims

1. A method for identifying probiotic strains suitable for treating human skin dysbiosis, comprising the step of performing network analysis via computer to determine the correlation between the bacterial strain and at least a second bacterial strain in dysbiosis and non-dysbiosis conditions, wherein: (i) The correlation differs between the said ecological disorder symptoms and the said non-ecological disorder symptoms. (ii) The correlation is lower in the ecological disorder symptoms and higher in the non-ecological disorder symptoms; (iii) The correlation refers to a positive or negative correlation, and (iv) The network is generated by co-occurrence analysis of the abundance of the bacterial strain and the second bacterial strain, the co-occurrence analysis being performed by DNA sequencing or whole-genome sequencing based on 16S rRNA amplicon, in which each node represents a species or an operational taxonomic unit (OTU), and the edges represent interactions between two species or OTUs.

2. The method of claim 1, wherein the correlation difference is at least 40%.

3. The method according to claim 1 or 2, wherein during the network analysis, the level of Operational Taxonomy Unit (OTU), Action Script Reader (ASV), and species or genus is set.

4. The method according to claim 1 or 2, wherein the bacterial strain and the second bacterial strain are strains found on human skin.

5. The method according to claim 1 or 2, wherein the bacterial strain is derived from at least one of Staphylococcus, Streptococcus, Microbacterium, Methyloversatilis, Deinococcus, Moraxella, or Acinetobacter.

6. The method according to claim 5, wherein the second bacterial strain is derived from the genera *Acidovorax*, *Actinomyces*, *Bacillus*, *Chryseobacterium*, *Corynebacterium*, *Fusobacterium*, *Staphylococcus*, *Streptococcus*, *Microbacterium*, *Methylobacterium*, and *Methyloversitia*. The bacterial strain is selected from at least one of the following genera: *Lis*, *Deinococcus*, *Micrococcus*, *Moraxella*, *Neisseria*, *Paracoccus*, *Prevotella*, *Pseudomonas*, *Sphingomonas*, *Acinetobacter*, or *Cutibacterium*, wherein if the genus of the second bacterial strain is the same as that of the bacterial strain, the species are different; and wherein if the genus and species are the same, the strains are different.

7. The method according to claim 1 or 2, wherein the ecological disorder includes at least one of acne, dandruff, dry skin, aging skin, pigmented skin, or inflammation.

8. The method of claim 7, wherein when the dysbiosis is acne, the bacterial strain is derived from Staphylococcus, Streptococcus, Microbacterium, Methyloversatilis, Deinococcus, Moraxella, or Acinetobacter.

9. The method of claim 8, wherein the bacterial strain from the genus Staphylococcus is Staphylococcus hominis or Staphylococcus epidermidis.

Images