Does microbiome-associated disease affect the inter ... · 8/9/2019 · 2 Abstract Space is a...

1

Does microbiome-associated disease affect the inter-subject heterogeneity of human microbiome?

Zhanshan (Sam) Ma12* Lianwei Li1

1Computational Biology and Medical Ecology Lab 5 State Key Laboratory of Genetic Resources and Evolution

Kunming Institute of Zoology Chinese Academy of Sciences

Kunming, China 2Center for Excellence in Animal Evolution and Genetics 10

Chinese Academy of Sciences Kunming, China

*For all correspondence: [email protected]

15

Running Head: Does MAD affect the inter-subject heterogeneity?

Keywords: Human microbiome associated disease (MAD); community spatial heterogeneity

(CSH); Taylor’s power law extensions (TPLEs); diversity-disease relationship (DDR);

heterogeneity-disease relationship (HDR). 20

Importance It has been argued that lack of consideration of scales beyond individual and ignoring of microbe

dispersal are two crucial roadblocks in preventing deep understanding of the heterogeneity of human

microbiome. Assessing and interpreting the spatial distribution (dispersal) of microbes explicitly are 25

particularly challenging, but implicit approaches such as Taylor’s power law (TPL) can still be

effective. Here, we investigate the relationship between human microbiome-associated diseases

(MADs) and the inter-subject microbiome heterogeneity, or heterogeneity-disease relationship

(HDR), by harnessing the power of TPL extensions and by analyzing a big dataset of 25 MAD studies

covering all five major microbiome habitats and majority of the high-profile MADs including 30

obesity and diabetes. We postulate that HDR reveals evolutionary characteristics because it utilizes

the TPL parameter that implicitly characterizes spatial behavior (dispersion), which is primarily

shaped by microbe-host co-evolution and is more robust against disturbances including diseases than

the traditional diversity-disease relationship (DDR).

35

certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was notthis version posted August 9, 2019. ; https://doi.org/10.1101/730747doi: bioRxiv preprint

https://doi.org/10.1101/730747

2

Abstract Space is a critical and also challenging frontier in human microbiome research. It has been argued

that lack of consideration of scales beyond individual and ignoring of microbe dispersal are two

crucial roadblocks in preventing deep understanding of the heterogeneity of human microbiome.

Assessing and interpreting the spatial distribution (dispersal) of microbes explicitly are particularly 40

challenging, but implicit approaches such as Taylor’s power law (TPL) can still be effective and offer

significant insights into the heterogeneity in abundance and distribution of human microbiomes.

Here, we investigate the relationship between human microbiome-associated diseases (MADs) and

the inter-subject microbiome heterogeneity, or heterogeneity-disease relationship (HDR), by

harnessing the power of TPL extensions and by analyzing a big dataset of 25 MAD studies covering 45

all five major microbiome habitats and majority of the high-profile MADs including obesity and

diabetes. Our HDR analysis revealed that in approximately 10%-17% of the cases, disease effects

were significant—the healthy and diseased cohorts exhibited statistically significant differences. In

majority of the MAD cases, the microbiome was sufficiently resilient to endure the disturbances of

MADs. Furthermore, comparative analysis with traditional DDR (diversity-disease relationship) 50

results is presented. We postulate that HDR reveals evolutionary characteristics because it utilizes

the TPL parameter that implicitly characterizes spatial behavior (dispersion), which is primarily

shaped by microbe-host co-evolution and is more robust against disturbances including diseases,

while diversity in DDR analysis is primarily an ecological-scale characteristic and is less robust

against diseases. Nevertheless, both HDR and DDR cross-verified remarkable resilience of human 55

microbiomes against MADs.

Introduction

The term “heterogeneity” literally means the quality or state of being diverse in character or content.

From its literal meaning, heterogeneity may even be considered as the other side of evenness coin 60

or as a proxy of biodiversity. Nevertheless, much of the formal and quantitative studies of

heterogeneity in ecology has been performed separately from diversity research. For example, in

population ecology, traditionally the terms heterogeneity, aggregation, dispersion, patchiness can be

used interchangeably, and all of which were often used to characterize the spatial distribution of

biological population. For this reason, the terms spatial heterogeneity and spatial distribution are 65

often used interchangeably in population ecology; in the former, the spatial information is referred

implicitly and, in the later, the spatial information is referred explicitly. Therefore, to avoid confusion,

the terms spatial heterogeneity (aggregation, dispersion, patchiness, contagiousness) should be


https://doi.org/10.1101/730747

3

more accurate in ecology. Note that temporal heterogeneity can also be defined and measured. In the

temporal context, temporal heterogeneity is essentially a proxy of (temporal) stability of population 70

or community (Ma 2015, Oh et al. 2016, Li & Ma 2019). In this study, our focus is spatial

heterogeneity.

When the concept of heterogeneity is constrained by the spatial arrangement (distribution) or

temporal (variation), its difference with diversity or evenness becomes clear. Obviously, traditional 75

diversity concept and its various metrics do not usually deal with spatial arrangement (e.g., Chao et

al. 2014). We are only aware of two exceptions for the lack of spatial information. One is the concept

of beta-diversity, which implicitly consider spatial information but is obviously far less convenient

in processing spatial information than the heterogeneity concept. Another exception is to do with the

concept of dominance (Ma & Ellison 2018, 2019), which is briefly discussed below. 80

Spatial heterogeneity is a fundamental property of any natural ecosystems including human

microbiomes. The spatial distribution of microbes on or in our bodies certainly possesses

far-reaching implications to our health and diseases. Space is not only the last, but also a critical

frontier in human microbiome research. In ecology, two most important scales that the spatial 85

heterogeneity exhibits are the population and community. Of course, to display spatial heterogeneity,

there must be at least two individuals in the ecosystem, which implies that heterogeneity is a group

(cohort, population, community, etc) characteristic. Obviously, the concept of heterogeneity is also

widely used in other fields of biology (e.g., genetic heterogeneity and landscape heterogeneity).

90

Theoretically, the abundance and distribution of species are among the most fundamental themes of

both theoretical population and community ecology. The concept of heterogeneity, especially the

spatial heterogeneity, is one of the most successful characterizations of species distribution and

abundance (Cohen & Meng 2015, Cohen & Saitoh 2016, Ma 2015, Reuman et al. 2014, 2017; Taylor

1984, 2019, Taylor et al., 1983, 1988). Practically, being able to predict the heterogeneity scaling 95

(change) over space and/or time, and to understand how the scaling is influenced by environmental

disturbances such as diseases is of critical significance.

At population scale, aggregation or dispersion are preferred terms and can be used interchangeably

with heterogeneity. In population ecology, aggregation can be measured quantitatively with 100


https://doi.org/10.1101/730747

4

Taylor’s (1961) power law, which achieved a rare status of ecological law (Taylor 2019) as further

explained below shortly. In community level or multi-species assembly context, we use the term to

refer to the uneven or heterogeneous nature of species abundances among different species within

a community and/or between communities. At community scale, community spatial heterogeneity

can be quantitatively measured with Taylor’s power law extensions (TPLEs), which were proposed 105

by Ma (2015) via extending Taylor’s (1961) power law from population to community level. On the

surface, it appears that the community heterogeneity can be considered as the other side of diversity

(evenness) “coin.” However, as explained previously, community spatial heterogeneity deals with

spatial information, therefore it is necessary to reiterate that the two sides of the same coin may indeed

possess different patterns. In other words, the heterogeneity analysis cannot be replaced by diversity 110

analysis, and the former can offer unique insights, which traditional diversity analysis may not.

In microbial ecology, the spatial distribution of microbes (whether it is the distribution of pathogens

or symbionts within our body) is, by no means, less important than in the ecology of plants and

animals in nature. The spatial distribution of microbes within our bodies or the intersubject 115

heterogeneity (differences) can influence the competition, coexistence, dispersal, and spread of

microbes within or between our bodies, and should have far reaching significance to our healthy and

diseases. In fact, the term “heterogeneity” has frequently occurred in the mission statements of both

the HMP (Human Microbiome Project) and HMP2 or iHMP (integrative HMP) of US-NIH, referring

to various aspects of the human microbiome from inter-subject heterogeneity in microbiome 120

composition to the heterogeneity in disease treatment response (HMP Consortium 2012, iHMP

Consortium 2019), which highlighted the significance of the heterogeneity in human microbiome

research. Nevertheless, methodological studies on the heterogeneity have been relatively rare and

studies that explicitly measure the relationship between microbiome heterogeneity and diseases

effects are still few. In the present study, we fill this gap by reanalyzing a big dataset of 16s-rRNA 125

metagenomic datasets of 25 MAD (microbiome associated diseases) studies collected from existing

literature, which cover all five major human microbiome habitats (gut, oral, skin, lung and vaginal)

and include the majority of the high-profile MADs such as IBD, obesity, diabetes, BV, and

neuron-degenerative diseases, and therefore of wide representative of the field.

130

Indeed, the heterogeneity of human microbiome is remarkable. For instance, it was found that no gut

bacterial strains of moderate abundance (>0.5% of the microbial community) were shared among all

individuals of human twins (Turnbaugh & Gordon 2009, Miller et al. 2019). However,

characterizing, quantifying and interpreting the heterogeneity of human microbiomes are rather


https://doi.org/10.1101/730747

5

challenging (HMP Consortium 2012, iHMP Consortium 2019). For another example, Falony et al. 135

(2016) collected 503 host factors possibly influencing human gut microbiome across nearly 4000

individuals, but interpreted merely 16.4% of the variation among individuals. Miller et al. (2019)

suggested that two roadblocks are responsible for the lack of explicability to the heterogeneity of

human microbiome. First, the analysis scale was usually limited to individual subjects (hosts).

Second, the microbial dispersal across individual hosts was often ignored. Directly supporting Miller 140

et al. (2019) arguments are rare but certainly exist. For example, Rothschild et al. (2018) revealed

that whether subjects lived together or not was the strongest predictor of microbiome variance,

indicating that transmission (among hosts and between hosts and their environment) could be a

critical driver of microbiome heterogeneity. Taylor’s power law, which achieved a rare status of

classic laws in macrobial ecology (Taylor 2019) and was recently extended to community ecology 145

of microbes (Ma 2015, Li & Ma 2019, Oh 2016), offers a powerful tool to remove the roadblocks

for better understanding the heterogeneity of human microbiomes because of its inherent

connections with spatial distribution and dispersion (Taylor 1983, 1984, 2019).

The objective of this study is therefore two-fold. First, we investigate the relationship between the 150

inter-subject heterogeneity and major human MADs, determining whether or not diseases have

significant influences on the inter-subject heterogeneity by harnessing the power of TPLEs in

measuring community spatial heterogeneity, as explained above. Second, we compare the

relationship between the heterogeneity-disease relationship (HDR) explored in this study with an

early meta-analysis by Ma et al. (2019) on the diversity-disease relationship (DDR), and further 155

explore the relationship between the two relationships (HDR & DDR) and their ecological

interpretations. Similar studies on the spatial distribution (heterogeneity) of host, pathogen and their

implications to the disease of plants and animals have been investigating routinely for decades, and

their significance to plant pathology and husbandry diseases have been well recognized (Noe &

Campbell 1985, Real 1996, Emry et al. 2011). Therefore, the study we conduct in this article should 160

have rather important significance to the investigation, diagnosis and treatment of the human

microbiome associated diseases, similar to the experience in the studies on the diseases of plants and

animals.


https://doi.org/10.1101/730747

6

Materials and Methods 165

The 16s-rRNA datasets of the human microbiome associated diseases

A brief description on the 16s-rRNA datasets from the 25 human MAD (microbiome associated

disease) studies is provided in Table S1 of the online supplementary information (OSI). These

datasets covered five major human microbiome habitats (gut, oral, skin, lung, and vaginal) as well

as milk and semen fluids. They also included the majority of the most common human MADs such 170

as IBD, obesity, diabetes, periodontitis, cystic fibrosis, mastitis, bacterial vaginosis, Parkinson’s

disease, schizophrenia. Therefore, the datasets are rather representative to the field of MADs.

Conceptual comparisons between population and community heterogeneities Methodologically, the study of heterogeneity is dependent on statistical variance and its relationship

with statistical mean of species abundance, or the so-termed Taylor’s power law (TPL) (Taylor 1961) 175

and its extensions (Ma 2015, Oh 2016) as explained below. The variance-mean power law

relationship can reveal certain important community characteristics and insights, which the

entropy-based diversity measures fail to offer (Ma 2015, Ma & Ellison 2018, 2019, Li & Ma 2019).

To explain why TPL is one of the most effective tools to investigate the population spatial distribution 180

(heterogeneity), it is necessary to recognize that there are three fundamental spatial distribution

patterns of any biological populations from microbes to humans; those are: uniform, random and

aggregated. The uniform distribution (sensu biologically) means that individuals of population in

space are distributed with equal space and the distribution (pattern) can be described with the uniform

distribution (sensu statistically). The random distribution means that individuals of population in 185

space are distributed randomly (not necessary with equal space distance) and can be described with

Poisson statistical distribution. The aggregated distribution (also known as patchy, clumped,

contagious, distribution) means that individuals in space are distributed far from random and often

in clusters of different sizes. The statistical distribution for the aggregated or heterogeneous spatial

distribution are usually rather special, and often follow highly skewed long-tail distributions such as 190

negative binomial, Ades distribution, power law distribution, or log-normal distribution (Perry &

Taylor 1985). Indeed, fitting statistical distributions to population abundance frequency is one of the

three major approaches for investigating the spatial distribution (aggregation, heterogeneity, or

pattern) of biological populations. The second approach to investigate the spatial heterogeneity is the

so-termed aggregation (heterogeneity) index approach. In plant ecology, the spatial information is 195


https://doi.org/10.1101/730747

7

explicit in the form of neighbor distances between plant individuals in a plot (Greg-Smith 1964). In

animal and insect ecology, neighbor-distance information is not often convenient to process, but the

metrics and models for measuring aggregation/dispersion/heterogeneity still implicitly deal with

spatial information. Various aggregation (heterogeneity) indexes were proposed to measure the

aggregation (heterogeneity) levels of the spatial distribution of biological populations. The third 200

approach is the so-termed regression modeling, including two regression models. One is Iwao’s

(1968, 1977) m*-m linear model (m* is the population mean crowding, itself is a measure of

aggregation, and m is the population mean). Another is Taylor’s (1961) V-m power-law (TPL) model

(V and m are the variance and mean of population abundance respectively). The advantages of the

third type approaches include the following: (i) The regression models (TPL & Iwao models) are 205

essentially the synthesis of the heterogeneity index (V/m or m*/m) across even larger scales (multiple

populations), and therefore can be utilized to characterize the spatial heterogeneity (distribution) of

many populations of a species. In other words, it synthesizes the heterogeneities of multiple

populations of a species, while the second approach (aggregation) can only characterize the

heterogeneity of a single population. (ii) The first approach—distribution fitting—can only describe 210

the heterogeneity of a single population, just as the second approach. Although it contains more

detailed information on the spatial heterogeneity than the second approach, it also suffers from the

same disadvantage as the aggregation index approach.

At the community level, community spatial heterogeneity (CSH) can be similarly defined with the 215

population spatial heterogeneity (aggregation) (PSH) and investigated with similar approaches.

Fitting the species abundance distribution (SAD) such as lognormal and log-series distributions can

be considered as the counterpart of the first approach (distribution-fitting to population abundances)

in population ecology. The community dominance defined by Ma & Ellison (2018, 2019), which

extended Lloyd (1967) mean crowding concept, can be considered as the counterpart of the second 220

approach (aggregation index approach) in population ecology. The four TPL extensions (TPLEs)

offer the third approach for investigating the community spatial heterogeneity (Ma 2015, Oh 2016),

the counterpart of the third approach (TPL) in population ecology. In the present study, we apply the

TPLEs for investigating the spatial heterogeneity of the human microbiomes as well as the influences

of the microbiome-associated diseases on the heterogeneity. 225

Taylor’s power law extensions (TPLEs)

Taylor (1961, 1984) revealed that the population mean (m) and variance (V) of a biological species

in space follows power function,

€

V = amb , (1)


https://doi.org/10.1101/730747

8

where m is the mean population abundance of the species at a specific spatial site, V is the 230

corresponding variance, b is a species-specific parameter that measures the population aggregation

(heterogeneity) degree, and a is a parameter primarily influenced by sampling scheme and is of

relatively little biological significance. Eqn. (1), known as Taylor’s power law (TPL) in literature,

has been verified by numerous field investigations and theoretical analyses and it is considered one

of few ecological laws in population ecology (Taylor 2019). 235

Ma (2015) extended the original TPL from single species population level to community level and

introduced four power law extensions (TPLEs). Type-I TPLE was defined to quantify the community

spatial heterogeneity, and it has the same mathematical form with the original TPL, but with different

ecological interpretations, i.e., 240

€

Vi = amib i=1, 2, …, N (2)

where mi is the mean species abundance of all species in the i-th community (individual host subject)

sampled, i.e., the mean population size (abundance) per species in the community, Vi is the

corresponding variance, N is the number of total communities (individual subjects) sampled, a is

similarly interpreted as in the original TPL. 245

The parameter b of Type-I TPLE is a measure for the degree (level) of the community spatial

heterogeneity (CSH). The CSH measures the dispersion (difference) in species abundances within

a community as well as the scaling of the dispersion (difference) across space (i.e., across

communities or sampled individual subjects). The larger the b-value is, the higher the community 250

heterogeneity is (Ma 2015, Li & Ma 2019). In the present study, we test whether or not the

heterogeneity scaling parameter (b) can be significantly influenced by MADs.

Type-III TPLE was defined to assess the spatial heterogeneity of the mixed-species population, with

the following formula, which is exactly the same with the original TPL and Type-I TPLE, just with 255

different ecological interpretations:

€

V j = am jb j=1, 2 ,…, S (3)

where mj is the mean population abundance of a specific species across all communities (individual

subjects) sampled, Vm is the corresponding variance, S is the number of species in the communities,

a is sampling scheme related, and b measures the spatial heterogeneity of mixed-species. 260


https://doi.org/10.1101/730747

9

The concept of mixed species was first discussed by Taylor et al. (1983), and it synthesizes

species-level spatial heterogeneity across all species. Type-III TPLE for mixed-species spatial

heterogeneity synthesizes the dispersion (difference) in population abundance of a single species

across space as well as the scaling of the dispersion (difference) across all species. The resultant 265

parameter b measures the heterogeneity (aggregation) characteristic of a virtually mixed species,

assuming that species identity is irrelevant and can be ignored (Ma 2015, Li & Ma 2019).

We use the log-linear transformations of TPLEs (2-3) to fit them statistically, i.e.,

€

ln(V ) = ln(a) + bln(m) . (4) 270

As a side note, Type-II and Type-IV TPLEs were defined to quantify the temporal stability

(heterogeneity), but they are not implicated in this study.

Ma (2015) proposed the concept of community heterogeneity critical abundance (CHCA), which

was defined with the following formula: 275

€

CHCA = exp ln(a)1− b⎛

⎝ ⎜

⎞

⎠ ⎟ (a > 0, b ≠1),

(5)

where a and b are the parameters of TPLE [eqns. (2-3)]. CHCA is a threshold or transition point

between the heterogeneous community and regular (completely even) community. At CHCA, the

distribution of community heterogeneity should be random. The CHCA is an extension of Ma (1991)

population aggregation critical density (PACD) to community scale. 280

Randomization tests for the influence of MADs on the spatial heterogeneity

The randomization (permutation) test is performed to determine whether or not the human

microbiome associated disease (MADs) possesses statistically significant influences on the scaling

(changes) of the spatial heterogeneity. The general procedure for randomization test is referred to

Collingridge (2013), and we adapted the general procedure for our testing the TPLE parameters as 285

following five steps. The number of re-samplings for randomization test is set to 1000 times to

calculate the p-value for determining the significance of differences. The randomization test

algorithm used to determine the differences in TPLE scaling parameters is exactly the same as that

used in Li & Ma (2019).

290


https://doi.org/10.1101/730747

10

Results We fitted Type-I & Type-III TPLE models to the healthy (H) and diseased (D) treatments for each

of the 25 MAD studies, respectively. Detailed fitting results were listed in Table S2 (for Type-I TPLE

or TPLE-I model) and Table S3 (for Type-III TPLE or TPLE-III model). Fig 1 and Fig 2 illustrated

the fittings of Type-I and Type-III TPLE models, respectively. The green lines and points 295

represent for the H treatments and the red (and blue) for the D treatments.

Suggested Location for Fig 1


We further performed randomization (permutation) tests to detect the differences in the community 300

spatial heterogeneity (CSH) parameter (b) and ln(a) between the H and D treatments (Table S4). The

community spatial heterogeneity parameter (b) (of TPLE-I) and the mixed-species population

aggregation parameter (b) (of TPLE-III) were plotted in Fig 3. Fig 3 also marked the cases with

significant differences in b between the H & D treatments, which were also displayed in Table

S4 (marked with grey shading). 305


From Table S2-S4, Figs 1-3, we summarize the following four findings:

(i) All but one treatment in the 25 MAD studies were fitted to Type-I & III TPLE models successfully 310

(p<0.05). The only exception (the H treatment of case #1) exhibited a p-value of 0.067, which is not

too far from the significance level (p=0.05) for determining the model fitting.

(ii) The observed average community spatial heterogeneity parameter (b) of Type-I TPLE model in

H treatment is 2.304 and in D treatment is 2.205. The observed average b of Type-III TPLE model 315

in H treatment is 1.743 and in D treatment is 1.751. These parameter values were computed from the

observed H & D treatment data without performing permutations, and they all fall within the normal

range of the power law and its extensions (Ma 2015, Taylor 2019).

(iii) Through the randomization tests for the differences in the heterogeneity parameter (b) between 320

the H & D treatments, it was found that: Type-I TPLE (for measuring the community spatial

heterogeneity) parameter b is different in 10% of the MAD studies; Type-III TPLE parameter b (for


https://doi.org/10.1101/730747

11

measuring mix-species population aggregation) is different in approximately 12.5% of the MAD

studies.

325

(iv) Through the randomization tests for the differences in the TPLE parameter (a) between the H

& D treatments, it was found that: Type-I TPLE parameter a is different in 12.5% of the MAD studies;

Type-III TPLE parameter a is different in approximately 17.5% of the MAD studies. This and the

previous findings suggest that MADs appear to have a more far-reaching influence at the

mixed-species level than at the community level. 330

Discussion

In the below, we further discuss the implications of the four findings, illustrated above, from the

perspective of the heterogeneity-disease relationship (HDR). Our discussion is approached from

comparison with our previous diversity-disease relationship (DDR) study (Ma et al. 2019). 335

First, the findings from the above HDR analysis echoed the conclusions from our previous DDR

study (Ma et al. 2019). In the previous study, the diversity in the Hill numbers (Chao et al. 2014)

exhibited significant differences between the H & D in only approximately 1/3 of the MAD cases,

and here the percentage from HDR analysis ranged between 10%-12.5%, which is less than ½ of the 340

percentage with difference from previous DDR analysis. This suggests that the MADs as

disturbances on the human microbiomes can only exert limited influences on the community

heterogeneity and/or diversity scaling (changes), and the influences on the heterogeneity appear to

be less than on the diversity. In other words, the human microbiome ecosystem possesses significant

stability (resilience). 345

However, in terms of the individual MAD case, the congruency between previous DDR analysis (Ma

et al. 2019) and HDR analysis in this study may or may not be aligned with each other. In the case

of TPLE-I, there are two case studies (#14 & #18), which show significant disease effects in HDR

analysis and also exhibited significant diseases effects in previous DDR analysis (Ma et al. 2019). 350

However, there are two case studies (#12 & #17) that show significant disease effects in HDR

analysis but exhibited no significant disease effects in previous DDR analysis (Ma et al. 2019). In

the case of TPLE-III, the congruency between previous DDR analysis and HDR analysis in this study

is aligned fully with each other indeed. All four MAD studies (#2, #4, #15 & #16) that exhibit

significant disease effects in the HDR also showed significant disease effects in previous DDR 355

analysis (Ma et al. 2019).


https://doi.org/10.1101/730747

12

Second, as to the reason why HDR analysis reveals a less significant percentage of differences

(disease effects) than previous DDR analysis, the answer can be found by distinguishing the nature

of the diversity metrics (Hill numbers) and the TPLE model (heterogeneity parameter b). What DDR 360

analysis reflects are primarily the ecological characteristics. In contrast, what HDR analysis,

particularly Type-I TPLE, reflects are primarily the evolutionary characteristics. That is, the

parameter (b) of TPLE is an evolutionary characteristic, which is well documented in existing

literature (Taylor 1961, 1980, 1981, 1984, 1986a, 1986b, 2019, Taylor & Taylor 1977, Taylor et al.

1983, 1988). Although, especially for microbes, the evolutionary and ecological characteristics are 365

often interwoven with each other and a clear division is usually difficult, the evolutionary properties

should be more robust (stable) in general. This explains the relatively lower disease effects on the

heterogeneity than on the diversity because the heterogeneity as evolutionary property should be

more robust (stable).

370

Third, as to the reason why TPLE heterogeneity parameter (b) showed a less significant percentage

of differences (disease effects) than the other TPLE parameter (a) in detecting the MAD effects, the

answer also lies in the fact that parameter (a) is not an evolutionary characteristic, instead, a is

primarily influenced by sampling schemes (Taylor 1961, 1984, 1986a, 1986b, 2019). For this reason,

parameter (a) is often attached less importance or even ignored in literature. 375

Finally, as to the reason why Type-III TPLE or the TPLE for mixed-species population aggregation

showed a more significant percentage of disease effects than Type-I TPLE, the answer lies in the fact

that Type-III TPLE is essentially still a population level model, although it uses community level

data. This finding also mirrored the results from shared species analysis in our previous study (Ma 380

et al. 2019). The shared species analysis in that study revealed that in approximately 50% of the MAD

cases, the shared species between the H & D samples were reduced due to the disease effects. In other

words, the disease effects are more significant at species scale than at community scale.

Acknowledgements 385 This work was supported “Industry Technology Talent Grant”, and “A China-US International Collaborative Project” from Yunnan Science and Technology Bureau, of China. Author Contributions ZS Ma designed the study, interpreted the results and wrote the paper. LW Li performed the 390 computation and participated in the interpretation of the results.


https://doi.org/10.1101/730747

13

Data accessibility All datasets analyzed in this study are available in public domain and the data sources for downloading are provided in the online supplementary Table S1. 395 Compliance with ethical standards N/A Conflict of interest: The authors declare that they have no conflict of interest.

400

References

Chao, A., Chiu, CH, & Jost, L (2014) Unifying species diversity, phylogenetic diversity, functional diversity and related similarity and differentiation measures through Hill numbers. Annual Reviews of Ecology, Evolution, and Systematics, 45:297–324. 405 Collingridge DS (2013) A Primer on Quantitized Data Analysis and Permutation Testing. Journal of Mixed Methods Research. 7, 79–95. Cohen JE, Xu M (2015) Random sampling of skewed distributions implies Taylor's power law of fluctuation scaling. PNAS 112(25):7749. 410 Cohen JE, Saitoh T (2016) Population dynamics, synchrony, and environmental quality of Hokkaido voles lead to temporal and spatial Taylor's laws. Ecology, 97(12):3402. Emry, DJ, HM Alexander & MK Tourtellot (2011) Modelling the local spread of invasive plants: importance 415 of including spatial distribution and detectability in management plans. Journal of Applied Ecology 2011, 48, 1391–1400. Falony, G. et al (2016) Population-level analysis of gut microbiome variation. Science 352, 560–564 420 Greg-Smith, P (1964) Quantitative Plant Ecology. 2nd ed. Butterworth, London, England HMP Consortium (2012) Structure, function and diversity of the healthy human microbiome. Nature, vol. 486:207–214 425 Integrative HMP (iHMP) Research Network Consortium (2019) The Integrative Human Microbiome Project. Nature, 569: 641–648. https://www.nature.com/articles/s41586-019-1238-8. Iwao, S (1968) A new regression method for analyzing the aggregation pattern of animal populations. Researches on Population Ecology 10:1–20. 430


https://doi.org/10.1101/730747

14

Iwao, S (1977) The m*-m statistics as a comprehensive method for analyzing spatial pattern of biological population and its importance to sampling problems. JIBP Synth., 21-46. Li LW, Ma ZS (2019) Comparative Power Law Analysis for the Spatial Heterogeneity Scaling of the 435 Hot-Spring and Human Microbiomes. Molecular Ecology. DOI: http://dx.doi.org/10.1111/mec.15124

Ma ZS (1991) Further interpreted Taylor’s Power Law and Population Aggregation Critical Density. Trans Ecol. Soc. China, vol. 1, 284–288. 440 Ma ZS (2015) Power law analysis of the human microbiome. Molecular Ecology, 24, 5428-5445. Ma ZS, Ellison AM (2018) A unified concept of dominance applicable at both community and species scale. Ecosphere, https://doi.org/10.1002/ecs2.2477. 445 Ma ZS, Ellison AM (2019) Dominance network analysis provides a new framework for studying the diversity-stability relationship. Ecological Monographs. https://doi.org/10.1002/ecm.1358 Ma ZS, Li L, Gotelli NJ (2019) Diversity-disease relationships and shared species analyses for human microbiome-associated diseases. The ISME Journal. DOI: 10.1038/s41396-019-0395-y. 450 Miller ET, R Svanback, BJM Bohannan (2018) Microbiomes as Metacommunities: Understanding Host-Associated Microbes through Metacommunity Ecology. Trends in Ecology and Evolution, Vol. 33(12):926-935 455 Noe, JP & CL Cambell (1985) Spatial pattern analysis of plant-parasitic nematodes. Journal of Nematology, vol. 17(2):86-93 Oh J, Byrd AL, Park M, et al (2016) Temporal Stability of the Human Skin Microbiome. Cell, 165(4):854-866. 460 Real LA & P McElhany (1996) Spatial pattern and process in pathogen interactions. Ecology, vol. 77(4): 1011-1025. Reuman DC, Gislason H, Barnes C, et al. (2014) The marine diversity spectrum. Journal of Animal Ecology, 83(4): 963-979. 465 Reuman DC, Zhao L, Sheppard LW et al. (2017) Synchrony affects Taylor's law in theory and data. Proceedings of the National Academy of Sciences of the United States of America, 114, 6788. Rothschild, D. et al. (2018) Environment dominates over host genetics in shaping human gut microbiota. 470 Nature 555, 210–215 Stumpf MPH, Porter MA (2012) Critical truths about power laws. Science, 335, 665–666.


https://doi.org/10.1101/730747

15

Perry, JN & Taylor LR (1985) Ades: new ecological families of species-specific frequency distributions that 475 describe repeated spatial samples with an intrinsic power-law variance-mean property. Journal of Animal Ecology, 54, 931-953. Taylor LR (1961) Aggregation, variance and the mean. Nature, 189, 732-735. 480 Taylor, LR & Taylor, RAJ (1977) Aggregation, migration and population mechanics. Nature, London, 265, 415-421. Taylor LR, Taylor RAJ, Woiwod IP et al. (1983) Behavioural dynamics. Nature, 303, 801–804. 485 Taylor LR (1984) Assessing and interpreting the spatial distributions of insect populations. Annual Review of Entomology, 29, 321–357. Taylor LR (1986a) Synoptic dynamics, migration and the Rothamsted insect survey. J. Anim. Ecol. 55:1-38. 490 Taylor LR (1986b) The four kinds of migration. In “Insect Flight: Dispersal and Migration” edited by W. Danthanarayana. Springer-Verlag Berlin, Germany. Taylor LR, Perry JN, Woiwod IP et al. (1988) Specificity of the spatial power-law exponent in ecology and agriculture. Nature, 332, 721–722. 495 Taylor, RAJ (1980) A family of regression equations describing the density distribution of dispersing organisms. Nature, vol. 286,53-55. Taylor, RAJ (1981). The behavioral basis of redistribution. I. The delta-model concept. Journal of Animal 500 Ecology, 50, 573-586. Taylor, RAJ (2019) Taylor’s Power Law: Order and Pattern in Nature, Academic Press. 657pp. Turnbaugh, PJ and Gordon, JI (2009) The core gut microbiome, energy balance and obesity. 505 J. Physiol. 587, 4153–4158

510


https://doi.org/10.1101/730747

16

Figure Legends 510

5

6

7

8

9

0.5 1.0 1.5 2.0ln(m)

ln(V)

Healthy

CD

UC

IBD1

−10

0

10

20

−5 0 5ln(m)

ln(V)

Healthy

CD

UC

IBD2

−10

−5

0

5

10

−7.5 −5.0 −2.5 0.0 2.5ln(m)

ln(V)

Lean

Overweight

Obesity

Obesity

−2

0

2

4

−3 −2 −1ln(m)

ln(V)

Healthy

Cancer

Cancer

0

4

8

12

−2 0 2ln(m)

ln(V)

Healthy

Treatment

Non−Treatment

HIV1

−5

0

5

10

−5.0 −2.5 0.0 2.5ln(m)

ln(V)

Healthy

Treatment

Non−Treatment

HIV2

−5

0

5

10

15

−6 −4 −2 0 2 4ln(m)

ln(V)

Healthy

T1D

T1D(Lean)

−5

0

5

10

15

−6 −4 −2 0 2 4ln(m)

ln(V)

Healthy

T1D

T1D(Obesity)

−5

0

5

10

−6 −4 −2 0 2ln(m)

ln(V)

Healthy

Control

MHE

MHE

−5

0

5

10

−7.5 −5.0 −2.5 0.0ln(m)

ln(V)

Healthy

Parkinson

Parkinson

8

10

12

1.5 2.0 2.5 3.0ln(m)

ln(V)

Healthy

Schizonphrenia

Schizonphrenia

6

8

10

12

1 2 3ln(m)

ln(V)

Healthy

Autism

Neurotypical

Autism

4

6

8

0.00 0.25 0.50ln(m)

ln(V)

Healthy

Atherosclerosis

Atherosclerosis1

5

6

7

8

9

10

1.6 1.8 2.0 2.2ln(m)

ln(V)

Healthy

Atherosclerosis

Atherosclerosis2

2.5

5.0

7.5

10.0

12.5

1 2 3ln(m)

ln(V)

Healthy

PB

PnB

Oral1

3

4

5

6

7

−0.5 0.0 0.5 1.0ln(m)

ln(V)

Healthy

Control

Diseased

Oral2

3

4

5

6

7

−0.5 0.0 0.5 1.0 1.5ln(m)

ln(V)

Healthy

Smoking

Smoking1

−10

−5

0

5

−6 −4 −2 0ln(m)

ln(V)

Healthy

Smoking

Smoking2

−5

0

5

−4 −2 0ln(m)

ln(V)

Healthy

Smoking

Smoking3

0

4

8

−3 −2 −1 0 1 2ln(m)

ln(V)

Healthy

Normal

Lesion

Skin

9

10

11

12

2.75 3.00 3.25 3.50 3.75ln(m)

ln(V)

Treated

Exacerbation

CF1

5

10

0 1 2 3 4ln(m)

ln(V)

Healthy

CF

CF2

−5

0

5

10

−5.0 −2.5 0.0 2.5ln(m)

ln(V)

Negative

Positive

Lung−HIV

7.5

10.0

12.5

2 3 4 5ln(m)

ln(V)

Healthy

BV

Vaginal

5.0

7.5

10.0

12.5

15.0

1 2 3 4 5ln(m)

ln(V)

Healthy

Subnormal

Abnormal

Infertile1

2.5

5.0

7.5

10.0

12.5

−1 0 1 2ln(m)

ln(V)

Healthy

Subnormal

Abnormal

Infertile2

12

14

16

18

5 6 7ln(m)

ln(V)

Healthy

Mastitis

Milk

Fig 1. Plots from fitting Type-I Taylor’s power law extension (TPLE-I) for each of the 25 human microbiome-associated diseases (MADs): the heterogeneity parameter (b) for measuring community spatial heterogeneity

515


https://doi.org/10.1101/730747

17

0

5

10

−4 −2 0 2 4 6ln(m)

ln(V)

Healthy

CD

UC

IBD1

0

10

20

−5 0 5 10ln(m)

ln(V)

Healthy

CD

UC

IBD2

−5

0

5

10

−3 0 3 6ln(m)

ln(V)

Lean

Overweight

Obesity

Obesity

0

5

−4 −2 0 2ln(m)

ln(V)

Healthy

Cancer

Cancer

−5

0

5

10

15

0 5ln(m)

ln(V)

Healthy

Treatment

Non−Treatment

HIV1

0

5

10

15

−3 0 3 6ln(m)

ln(V)

Healthy

Treatment

Non−Treatment

HIV2

−5

0

5

10

15

20

0 5ln(m)

ln(V)

Healthy

T1D

T1D(Lean)

0

5

10

15

20

0 5 10ln(m)

ln(V)

Healthy

T1D

T1D(Obesity)

0

5

10

−2 0 2 4 6ln(m)

ln(V)

Healthy

Control

MHE

MHE

−5

0

5

10

15

−5.0 −2.5 0.0 2.5 5.0 7.5ln(m)

ln(V)

Healthy

Parkinson

Parkinson

0

5

10

15

−3 0 3 6 9ln(m)

ln(V)

Healthy

Schizonphrenia

Schizonphrenia

−5

0

5

10

15

−5 0 5ln(m)

ln(V)

Healthy

Autism

Neurotypical

Autism

0

5

10

−2.5 0.0 2.5 5.0ln(m)

ln(V)

Healthy

Atherosclerosis

Atherosclerosis1

0

5

10

15

−2.5 0.0 2.5 5.0ln(m)

ln(V)

Healthy

Atherosclerosis

Atherosclerosis2

0

5

10

15

−2.5 0.0 2.5 5.0 7.5ln(m)

ln(V)

Healthy

PB

PnB

Oral1

−4

0

4

8

−2 0 2 4ln(m)

ln(V)

Healthy

Control

Diseased

Oral2

0

4

8

12

−2 0 2 4ln(m)

ln(V)

Healthy

Smoking

Smoking1

−5

0

5

10

−2.5 0.0 2.5 5.0ln(m)

ln(V)

Healthy

Smoking

Smoking2

−5

0

5

10

−2.5 0.0 2.5 5.0ln(m)

ln(V)

Healthy

Smoking

Smoking3

−5

0

5

10

15

−2.5 0.0 2.5 5.0ln(m)

ln(V)

Healthy

Normal

Lesion

Skin

0

5

10

15

−2 0 2 4 6ln(m)

ln(V)

Treated

Exacerbation

CF1

0

5

10

15

−2.5 0.0 2.5 5.0ln(m)

ln(V)

Healthy

CF

CF2

−5

0

5

10

15

0 5ln(m)

ln(V)

Negative

Positive

Lung−HIV

−5

0

5

10

15

−5.0 −2.5 0.0 2.5 5.0ln(m)

ln(V)

Healthy

BV

Vaginal

−5

0

5

10

15

−4 0 4 8ln(m)

ln(V)

Healthy

Subnormal

Abnormal

Infertile1

0

5

10

15

−4 −2 0 2 4 6ln(m)

ln(V)

Healthy

Subnormal

Abnormal

Infertile2

0

5

10

15

20

0.0 2.5 5.0 7.5 10.0ln(m)

ln(V)

Healthy

Mastitis

Milk

515 Fig 2. Plots from fitting Type-III Taylor’s power law extension (TPLE-I) for each of the 25 human microbiome-associated diseases (MADs): the heterogeneity (aggregation) parameter (b) for measuring mixed-species population heterogeneity (aggregation)


https://doi.org/10.1101/730747

18

*

*

**

Milk

Infertile2

Infertile1

Vaginal

Lung−HIV

CF2

CF1

Skin

Smoking3

Smoking2

Smoking1

Oral2

Oral1

Atherosclerosis2

Atherosclerosis1

Autism

Schizonphrenia

Parkinson

MHE

T1D(Obesity)

T1D(Lean)

HIV2

HIV1

Cancer

Obesity

IBD2

IBD1

0 2 4 6b value of Type−I PLE

Healthy Diseased−L1 Diseased−L2

*

*

*

**

*

Milk

Infertile2

Infertile1

Vaginal

Lung−HIV

CF2

CF1

Skin

Smoking3

Smoking2

Smoking1

Oral2

Oral1

Atherosclerosis2

Atherosclerosis1

Autism

Schizonphrenia

Parkinson

MHE

T1D(Obesity)

T1D(Lean)

HIV2

HIV1

Cancer

Obesity

IBD2

IBD1

0.0 0.5 1.0 1.5 2.0b value of Type−III PLE

Healthy Diseased−L1 Diseased−L2

520 Fig 3. Bar graphs displaying the results from randomization tests for the difference in the heterogeneity scaling parameter (b) between the H (healthy) and D (diseased) treatments, the MAD (microbiome associated disease) cases with significant differences are marked with star (*).

525


https://doi.org/10.1101/730747

19

525 Online Supplementary Tables Table S1. Twenty-six 16S-rRNA metagenomic datasets utilized to analyze the effects of microbiome-associated diseases (MADs) on the microbiome heterogeneity 530 Table S2. The parameters of Type-I power law extension (TPLE-I) for the community spatial heterogeneity for each H (healthy) or D (diseased) treatment of the 25 MAD studies Table S3. The parameters of Type-III power law extension (TPLE-III) for mixed-species population spatial aggregation for each H (healthy) or D (diseased) treatment of the 25 MAD studies 535 Table S4. The p-values from the permutation tests (with 1000 times of permutation re-sampling) for the differences in the TPLE-I and TPLE-III parameter (b) between the H and D treatments


https://doi.org/10.1101/730747

Does microbiome-associated disease affect the inter ... · 8/9/2019 · 2 Abstract Space is a...

Documents

Transcript of Does microbiome-associated disease affect the inter ... · 8/9/2019 · 2 Abstract Space is a...