The Belly Buttons Will be Revealed, Slowly

We are finally revealing our last batch of results about Belly Button Biodiversity, results of samples of 583 people. In doing so, we are going to try something new. I am going to write a series of short articles, paragraphs really, about belly button life, out of which I will build a paper. Every few days, I will add a few pieces until I have added the data, results, discussion, and, ultimately, produced a whole paper. Once I get to the results, the data will be linked below. If you have participated in the Belly Button Biodiversity project, you will be able to see there a list of your species. As you read, pardon the dust.

Authors: Robert R. Dunn (me), Julie Horvath-Roth, Sarah P. Council, Holly Menninger, Matthew Fitzpatrick

25 November 2013

The layer of life living on every piece of human skin resembles a desert, grassland or forest. Species living on the skin compete for resources provided by the skin much as trees compete for sunlight or soil nutrients. In terrestrial biomes, the composition of species in any particular patch of habitat (for example, an urban forest lot) depends on a mix of the outcome of competition, cooperation, and dispersal (and which species arrive first). It seems likely that similar dynamics govern how many individuals of which species can be found living on skin habitats such as the belly button. In every patch of human skin, microbes variously wage chemical war against each other, build partnerships, and interact with new arrivals.

Relative to the size of microbes, the skin microbiome is immense, more akin to the entire Amazon than it is to a forest fragment. By most estimates, the average adult human is covered by 1.8 m² of skin surface (Grice et al., 2009), an area large enough to contain several trillion bacteria if those bacteria were just one layer thick (which they are not). But if one really unfolded the skin, stretching out its microtopography of glands, bumps and divots, this area would be many times more expansive. In this light, it is not surprising to find that the skin microbiome can be highly diverse. Nearly two thousand microbial taxa have been found in the belly button alone (Hulcr et al. 2011) and perhaps tens of thousands, though no one keeps a cumulative count, have been found among the skin habitats more generally. Yet, while the skin of humans is collectively diverse, the skin of any particular human need not necessarily be. The skin of some individual humans can host thousands of species, but other humans seem to play host to just tens.

29 November 2013

In contrast to popular conceptions of germs in which any bacteria are bad, recent research suggests having a higher diversity of microbes living on your body may be beneficial. Lower microbial diversity has been associated with health problems, particularly chronic inflammatory disorders such as Crohn’s disease, allergies and rheumatism (e.g., Hanski et al 2012). The nature of the link between microbial diversity and health and well-being deserves better study, but so does the question of why the biodiversity of microbes on humans arise in the first place.

30 November 2013

The use and overuse of antibiotics (potentially including antimicrobials such as wipes, sprays, socks and other products containing Triclosan) can lead to decreases in the diversity of gut microbes (e.g., Flores et al. in review). Similar effects on skin microbes seem likely.  Analogy to grasslands and forests, however, suggests that other environmental factors might also affect the biodiversity of skin microbes, particularly those aspects of the environment associated with climate (It is worth noting here than in grasslands and forests, such effects are not always linear but we will ignore this for now only to return to it in the discussion). On individual bodies, wetter skin habitats tend to be less diverse than drier ones (e.g., Costello et al. 2011), much as rain forests are more diverse than deserts. The same might be true of wetter relative to drier people. The ability of the body to provide resources might also influence diversity, with the body’s aging being a key factor that influences condition and hence the resources available to microbes from any individual skin cell or gland. Previous studies suggest a weak effect of age on gut microbes, with humans over the age of sixty experiencing declines in gut microbial diversity (Knight et al…) whereas those under the age of seven are predicted to have higher than normal diversity (references). It is also possible that the biodiversity near bodies (whether inside or outside of homes and other living spaces) in the form of animals and plants, might influence the microbes on skin in as much as  a greater mammal and plant biodiversity leads to a greater number of microbe species able to colonize human skin (von Hertzen et al. 2011). In Finland, Hanski et al. (2012), found that the microbes on the forearms of adolescents differed as a function of the diversity of native plants outdoors. Testing the relative influence of these factors microbial diversity on skin has been difficult because until recently studies have rarely been large enough to include variation in each of these variables. Here we worked with 542 collaborating citizen scientists who swabbed their own belly buttons to test whether any of the variables thought to be associated with local climate (innie/outie), body condition (human age), gender, self-reported ethnicity, and outdoor climate are associated with the diversity of microbial OTUs (a taxonomic level akin to the subgenus).

In addition to varying in how many kinds of species live on their skin, humans also differ in terms of which microbes live on their skin. Like its biodiversity, the composition of microbes  on human skin might also influence human health (Certainly microbiomes dominated by pathogens are worse for us than those that are not). The composition of the skin microbiome might be dominated by the same sorts of deterministic factors thought to influence diversity (microclimate, age, climate, ethnicity, etc…). However, the study of grasslands and forests has revealed that the composition of ecological communities is not always governed exclusively by environmental conditions. Chance and dispersal also play prominent roles. Here, we use Generalized Dissimilarity Modelling (GDM) a nonlinear statistical approach to compare patterns in the composition of belly button bacteria as a function of those same factors considered with regard to belly button diversity.

Methods

December 4th

Samples–Over two years we engaged 542 members of the public who sampled their own belly buttons. Elsewhere, we report on the first 60 individuals sampled. Here we consider the entirety of participants. All participants were provided a written Informed Consent form approved by the North Carolina State University’s Human Research Committee (Approval No. 1987). The University’s Human Research Committee has approved this study. Participants swabbed their own belly buttons with sterile cotton tips that were then immersed in 0.5 ml 10% phosphate saline buffer. Swabbing is an effective means to take a subsample of the bacteria living on the skin surface and is similar to other methods in terms of its effectiveness. Notably, however, swabbing does not sample those microbes living in hair follicles and glands such that it characterizes one part of a broader skin ecosystem. Most samples were collected at events that the yourwildlife.org program staged at North Carolina State University, the North Carolina Museum of Natural Sciences or elsewhere in the region. However, a subset of 105 samples came from kits mailed to participants. Once kits arrived in our lab they were kept at−20°C. Genomic DNA was extracted from 50 µL of the sediment of centrifuged samples using the PowerSoil DNA extraction kit (MoBio, Inc.), which we modified according to Lauber et al. (2009).

Because samples were collected over two phases, sequencing also took place in two phases. In phase one of the project… In phase two of the project…

General Linear Models–We identified those variables associated with the diversity of microbial assemblages in belly buttons using general linear models. To do this, we used a model simplification procedure in which we removed non-significant variables (α = 0.05) in a stepwise fashion. Best models were those in which AIC was minimized. We used this stepwise approach to explore the relative contributions of the various terms included in the initial model. Model simplification approaches have been criticized  (Whittingham et al. 2006). However, we know so little about the potential drivers of diversity of skin microbes that it represents a useful first step (in line with Anderson 2008).

Generalized Dissimilarity Models–

Explanatory variables–

Results

Samples were obtained for a total of 582 participants. Of those participants, we were able to successfully isolated and sequence microbial DNA in 153 participants in the first phase of the project and 255 participants in the second phase of the project. Hereafter, we focus on the larger second phase of the project to derive inferences about the factors structuring the biodiversity and composition of belly button inhabitants (who is there and why). We then use the first half of the data to test the robustness of associations observed in the second half. In the second half of the data we found 805 phylotypes of bacteria and 7 phylotypes of archaea based on  273,000 reads after rarefaction (728,858 before). This excludes reads of poor quality.We refer to these phylotypes (also called OTUs, Operational Taxonomic Units) as species, but in truth there are likely to be more species here than the number of phylotypes suggests.  Phylotypes were regarded as distinct if they were different at greater than 3% of their nucleotides, which is tandard in microbial studies, but very different species of bacteria may differ by far less than 3% divergence.

Figure 1. The relative abundance of different microbial “species” (technically OTUs). Each pie wedge represents a single species, with the size of the wedge a function of the total abundance of that species (total number of reads). Species can have many occurrences either by being present on many people or being represented by many individuals when present. In practice, the most abundant species in belly buttons are all both found on many people and abundant on those individuals. A key take home from this figure, other than the identity of the most abundant organisms is that while many hundreds of species can be found in belly buttons, as in other bodily habitats, that a much smaller number, species we call oligarchs (Hulcr et al. 2012), dominates. As part of the public science efforts associated with this projected, we have written about many of these species in our free online Book of Invisible Life. For example see the stories of Corynebacterium, Staphylococcus (almost certainly the genus of Staphylococcaceae in which the most common species below belongs), Prevotella, and Streptococcus.

The vast majority of phylotypes were both infrequent (encountered on few people) and rare (represented by few reads when present; Table 1). Of 2368 total phylotypes, 2188 were present on less than 10% of individuals sampled (Table 1), and most of those were present on just one individual. Conversely, no phylotypes were present on all individuals sampled and just eight phylotypes were present on more than seventy percent of individuals. These eight phylotypes accounted for nearly half (45%) of the total reads of bacteria in our study.

here I’m putting online our first model. This may be supplanted as we use remote sensing to measure more variables about where people live.

GLM– The model that best explained the diversity of bacteria in belly buttons included the four variables, whether a person lived in a cool temperate (persistent winter snows) or warmer part of the world, participant age, and ethnicity. Individuals from warmer regions tended to have more species of bacteria in their belly buttons than those from colder regions. Our interpretation here is that while in cold regions everyone seems to have a relatively low diversity of skin microbes, in warmer regions a subset of individuals have a much higher diversity of belly button microbes. Biodiversity also decreased with age. Inclusion of ethnicity in models improved overall model performance but the effects of ethnicity were complex and not, on their own, significant.

Discussion

To be written

Box 1. Natural History of the Belly Button

Generally speaking, the skin of a human is a dry, hot, acidic environment. To understand this desert, we need to first understand its floor, the skin itself. The layer of the skin with which bacteria directly interact is called the stratum corneum and is composed of cells known as squames. Squames begin their lives interior to the skin surface and over several weeks go through the transformations that leads them out to the boundary between our bodies and those of our microbes. There, squames live a short life, layered, protectively, one atop another, before being shed into the environment. When it comes to its surface cells, those squames, the belly button is no different than any other part of the skin. In the belly button, bacteria live on top of, but also between and even beneath squame cells, much like moss might grow on and then also between the shingles on a roof.

To some extent, the habitat formed by the squame cells themselves is three dimensional, but the third dimension of the skin microbiome is most pronounced in glands. The skin’s wall of squame cells is pocked by several types of glands. Eccrine glands are the most common type of skin gland.  Eccrine glands secrete sweat, a substance composed of water and salt but also urea and amino acids (which some microbes metabolize) and small peptides that kill some microbes but appear to favor others. Eccrine glands are abundant in the belly button. Apocrine glands are more unusual; instead of sweat they secrete a milky substance that appears to serve primarily as a food source for bacteria. Apocrine glands are primarily found in the armpits, anus, genitals, nipples and belly button. Finally, sebaceous glands are associated with hair follicles. Sebaceous glands secrete the fatty sebum that coats hair. Like sweat, sebum deters the growth of some microbes. Sebaceous glands are found in the belly button in direct proportion to its hairiness (pre-waxing).

The skin’s surface and its glands form a habitat on which thousands of species have been discovered, species of many ancient microbial lineages. No fewer than 36 phyla of bacteria have been named and more phyla await description. Yet, the human body in general and the skin in particular is dominated by just four of these phyla, Actinobacteria, Firmcutes, Bacteroidetes and Proteobacteria. Whereas Firmicutes and Bacteroides, the skin is dominated by species of Firmicutes and Actinobacteria (Figure 1, Figure S1). The commonness of species of Actinobacteria in particular almost certainly reflects a long coevolutionary relationship with animal hosts, a coevolutionary relationship that has endowed species of many Actinobacteria with adaptations that favor their growth under the hostile conditions presented by the skin environment. Similarly, the body itself may bear adaptations for dealing with specific species taxa of Actinobacteria, as has been suggested in the case of Corynebacterium.

Box Figure 1. The number of reads of species of the four major phyla found in on the belly button skin relative to that of all other phyla. One can also consider the relative diversity of phyla in the belly button. The diversity of different taxa appears to be more equitable than is their relative frequency and abundance (Box Figure S1).

Within the skin habitat, sub-habitats differ from each other in predictable ways in terms of their composition of species and higher taxa. An armpit, for example, rarely resembles a forearm in terms of its composition of bacteria. The relative concentrations of the three types of glands and their exudates in different parts of the skin appear to play a key role in this distribution. The hair follicles and their sebaceous glands, for example, appear to favor the growth of Propionibacterium species, including P. acne, the species associated with acne. These species are able to metabolize the lipids. By the same token, Corynebacterium species thrive where apocrine glands are dense. The belly button possesses some of each gland type and so, not surprisingly, both Corynebacterium and Propionbacterium can be found in its interstices. The belly button (like other moist parts of the skin, including the armpits and butt crack, AKA gluteal crease) also hosts abundant populations of Staphylococcus species. Staphylococcus species on the skin are typically aerobic (dependent on oxygen) and have been hypothesized to rely on the urea present in the sweat exuded by the eccrine glands.

The microbes on the skin, including that of the belly button, appear to play a primary role in immune regulation, but also in defending the body against pathogens. More than 3000 skin diseases are known in humans (Lewin Group 2004) and most of these are associated with pathogens that colonize the skin, which first requires them to take over habitat from symbiont bacteria. Microbes on skin have also been hypothesized to play a role in mate choice (e.g., in lemurs), and individual recognition. The extent to which they also play such a role in humans is poorly understood.

Recently, research has begun to focus more on the colonization of bodily habitats by microbes. It is clear that many of the microbes of the gut pass from mother to child during vaginal birth and that others, particularly Bifidobacteria, are spread to infants in breast milk. In both cases, the body has evolved specific adaptations for ensuring this the passage of beneficial microbes to the next. By the same token, a key function of the human appendix now appears to be as a storage organ for beneficial microbes in the event that they are lost from the gut during infection. In this light, the idea that the body also facilitates the spread of certain skin microbes from one generation to the next seems plausible. This may be particularly important in the belly button. After birth, the belly button is prone to infection as what remains of the umbilical cord rots off. During these early moments of life, having a skin microbial fauna that protects the newborn from infection seems incredibly important.

Supplement 1. A Letter from Students
Dear Rob Dunn et al.,

We are budding biologists from Utah State University. We have spent the last few weeks studying the human microbiome. We started by listening to Science Friday podcasts about the human microbiome, including the one in which you were interviewed. We then sampled bacteria from our own microbiomes, grew bacteria from our samples in petri dishes, and further sequenced the DNA from one bacteria species per person. We then looked at the diversity of species found among human microbiomes by constructing phylogenetic trees. Importantly, we also came up hypotheses and analyzed data regarding your belly button biodiversity citizen science project. Below are 1) all of our hypotheses regarding factors that might predict a person’s belly button biodiversity, and 2) the four questions we addressed with your dataset (one question per section of this course) and the answers.

1) Potential belly button biodiversity explanatory variables (these are variables we did not find in your dataset):

a. Work environment. For example, if someone is constantly working outside we think they might have different kinds of bacteria or more of them than someone that works in an office all day.

b. Pregnant vs. non-pregnant females.

c. Types of shampoo / body wash / soap used when washing. For example, some soaps are made just for their scent, whereas others are actually for killing bacteria or reducing them.

d. Similarly, washing with hot vs. cold water. For example, if individuals are just washing with warm water, that might not be as effective as washing with soap and hot water.

e. Whether a person has dry or oily skin.

f. Physical activity (i.e., amount of time they exercise per day).

g. The type of clothing they wear / the kind of fabric (e.g., cotton, polyester). There might be different fabrics that accumulate in the belly button differently and affect the living conditions of the belly button environment.

h. A person’s weight. Larger people might have a harder time keeping their belly button clean because their belly buttons are deeper.

i. Particular sports played often. For example, if someone is on a swim team or goes to the pool often, the chlorine might affect bacteria. Also, wrestlers roll around on mats that are absolutely disgusting.

j. How sanitized their living conditions are.

k. The strength of a person’s immune system (i.e., how many times a person has been exposed to threats to the body).

l. How often a person’s belly button is in contact with other belly buttons.

m. Type of laundry detergent used to wash clothes.

n. How often a person wears clothes.

o. How much hair a person has around their belly button.

p. Average humidity where a person lives / spends most of their time. It is possible that the more moisture present in a person’s belly button, the greater the diversity of bacteria can survive there.

q. How often a person washes their clothes.

r. The number of clothing layers usually worn, which is usually associated with the climate of the place the person is living.

s. A person’s specific genome.

t. Living on the coast vs. inland. Living by water may cause there to be different bacteria in the atmosphere versus living inland. Also, going into the ocean frequently may have an effect.

u. Amount of daily sweat.

v. Living in a developed vs. developing country. In general, developing countries have dirtier living conditions than developed countries.

w. Number of people a person lives with. The roommates can also be slobs or be very clean.

x. Amount of annual sunlight exposure.

y. Presence or absence of belly button rings. We also wonder whether the type of metal affect the types of bacteria present.

z. Use of lotion after taking a shower. Maybe more bacteria would be drawn to lotion instead of having no lotion on. The bacteria might like the lotion environment more and thrive better.

aa. The source of water used during showering.

bb. Medications used.

cc. Amount of stress a person experiences. If a person is under a lot of stress their body doesn’t function as well as it should. Therefore, there might be more bacteria in a stressed person’s belly button because their body can’t fight it off.

dd. Diet. The food we eat causes our bodies to do different things (i.e., affects pH and temperature of the body).

ee. Sleeping conditions. For example, if a person sleeps with a dog or cat and sleeps without a shirt on, the pet may rub against the belly button causing a diversity of bacteria to get in there.

2) Data analyses conducted with a subset of the data in your dataset (each description below is written by a student):

a. From Anthony Y.: Are there differences between male and female belly button biodiversity? First, Lauren went through each individual in the belly button biodiversity data set, and calculated the number of different species in each person’s belly button. Then, we sorted the data between male and female. We then performed a T-test to calculate a p-value to determine if the differences between male and female biodiversity was significant. The p-value came out to .27. The p-value was much greater than .05. This means that the differences in values between male and female had a high probability of being by chance. This means there is no significant difference between male and female belly button biodiversity.

b. From CeAnn C.: Rob, as a class we have been interested in bacteria species and their association with certain areas of the body. We took a look at a part of your data set and came up with this question, “Is the number of bacteria species found in the belly button correlated with the number of times people wash their belly buttons each week?” Do determine this, we calculated a correlation coefficient with the given data set. We sorted the data before we ran the correlation coefficient. When looking at the raw data when the number of species is sorted from smallest to largest, you can see that the amount of times each belly button got washed doesn’t necessarily have a pattern with how many species were found, so initially I concluded that there was no correlation. After computing the coefficient, .154, that verified that my hypothesis was correct, that there was little (weak) to no correlation between the number of species in a belly button and the amount of times that belly button got washed per week.

c. From Megan J.: The question we were addressing as a class was: Is the amount of forest in the places the participants have lived during the ages 13-19 correlated with belly button biodiviersity? To answer this, I first sorted the data in descending order so I could omit the boxes that did not include any data. Then I used the CORREL function in excel which allowed me to compare the two data, the percent forest cover in the center of the zip code where the participants lived during ages 13-19, to the number of species that was found in their belly button. The result I got by using the CORREL function was 0.126572. Correlation coefficients range from -1 to 1, and if it is between the numbers -.02 and .2 it means there is no relationship between the two data sets. This means that this is not a correlation in the amount of forest in the places participants have lived and during the ages 13-19 and belly button biodiversity.

d. From Kent M.:  I have analyzed your belly button biodiversity data to examine whether the temperature of where your participants have lived correlates with their belly button biodiversity. To answer this question I used a correlation function to calculate the correlation coefficient for the number of species an individual’s belly button contained and the mean max monthly temperatures from 0-13 and 13-19 respectively. For ages 0-13 this coefficient was 0.2678, and for ages 13-19 the coefficient was -0.1442. Both of these coefficients indicate that there is not a correlation between the number of bacterial species inhabiting an individual’s belly button and the mean max monthly temperature in which an individual has lived from ages 0-13 and/or 13-19. This result is, perhaps, not surprising as bacteria live and die quite rapidly. A species that was able to survive in an individual’s belly button in a certain mean max temperature in the past may be totally extinct in that same belly button in the present because it is not suited to survive in the present mean max temperature. It is my thought that more useful information for examining the correlation between temperature and belly button biodiversity may be obtained by using the mean max temperature in which the participant currently lives instead of where they have lived. Exciting stuff, Robert Dunn, Thank you.

We hope this proves helpful for your project. We would love to know if you end up using any of our ideas or results.

Very best wishes,
94 students & Lauren Lucas (Instructor)