downloaded from on june 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche...

34
1 Poisoning with Soman, an Organophosphorus Nerve Agent, Alters Fecal Bacterial Biota and 1 Urine Metabolites: a case for Novel Signatures for Asymptomatic Nerve Agent Exposure 2 3 Derese Getnet a , Aarti Gautam a , Raina Kumar a,b , Allison Hoke a,c , Amrita K. Cheema d , Franco 4 Rossetti e , Caroline R. Schultz f , Rasha Hammamieh a , Lucille A. Lumley g , and Marti Jett a# 5 6 a Integrative Systems Biology Program, US Army Center for Environmental Health Research, 7 Fort Detrick, Maryland 8 b Advanced Biomedical Computing Center, Frederick National Lab for Cancer Research, Fort 9 Detrick, Maryland 10 c The Geneva Foundation, US Army Center for Environmental Health Research, Fort Detrick, 11 Maryland 12 d Departments of Oncology and Biochemistry, Molecular and Cellular Biology, Georgetown 13 University Medical Center, Washington DC 14 e Clinical Research Management, Silver Spring, Maryland 15 f Edmond Scientific Company, Aberdeen Proving Ground, Maryland 16 g US Army Medical Research Institute of Chemical Defense, Aberdeen Proving Ground, 17 Maryland 18 Running title: Soman induced changes in microbiome and urine metabolome 19 Address correspondence to Marti Jett, Chief Scientist of Systems Biology Enterprise at 20 [email protected] 21 Present address: US Army Center for Environmental Health Research, Fort Detrick, Maryland 22 A.G., D.G., and R.K. contributed equally to this work. 23 AEM Accepted Manuscript Posted Online 14 September 2018 Appl. Environ. Microbiol. doi:10.1128/AEM.00978-18 This is a work of the U.S. Government and is not subject to copyright protection in the United States. Foreign copyrights may apply. on June 23, 2020 by guest http://aem.asm.org/ Downloaded from

Upload: others

Post on 17-Jun-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

1

Poisoning with Soman, an Organophosphorus Nerve Agent, Alters Fecal Bacterial Biota and 1

Urine Metabolites: a case for Novel Signatures for Asymptomatic Nerve Agent Exposure 2

3

Derese Getneta, Aarti Gautam

a, Raina Kumar

a,b, Allison Hoke

a,c, Amrita K. Cheema

d, Franco 4

Rossettie, Caroline R. Schultz

f, Rasha Hammamieh

a, Lucille A. Lumley

g, and Marti Jett

a# 5

6

aIntegrative Systems Biology Program, US Army Center for Environmental Health Research, 7

Fort Detrick, Maryland 8

bAdvanced Biomedical Computing Center, Frederick National Lab for Cancer Research, Fort 9

Detrick, Maryland

10

cThe Geneva Foundation, US Army Center for Environmental Health Research, Fort Detrick, 11

Maryland

12

dDepartments of Oncology and Biochemistry, Molecular and Cellular Biology, Georgetown 13

University Medical Center, Washington DC 14

eClinical Research Management, Silver Spring, Maryland

15

fEdmond Scientific Company, Aberdeen Proving Ground, Maryland

16

gUS Army Medical Research Institute of Chemical Defense, Aberdeen Proving Ground, 17

Maryland

18

Running title: Soman induced changes in microbiome and urine metabolome 19

Address correspondence to Marti Jett, Chief Scientist of Systems Biology Enterprise at 20

[email protected] 21

Present address: US Army Center for Environmental Health Research, Fort Detrick, Maryland 22

A.G., D.G., and R.K. contributed equally to this work. 23

AEM Accepted Manuscript Posted Online 14 September 2018Appl. Environ. Microbiol. doi:10.1128/AEM.00978-18This is a work of the U.S. Government and is not subject to copyright protection in the United States.Foreign copyrights may apply.

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 2: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

2

24

Abstract: The experimental pathophysiology of organophosphorus (OP) chemical exposure has 25

been extensively reported. Here, we describe an altered fecal bacterial biota and urine 26

metabolome that follows intoxication with soman, a lipophilic G class chemical warfare nerve 27

agent. Non-anaesthetized Sprague-Dawley male rats were subcutaneously administered soman at 28

0.8 (sub-seizurogenic) or 1.0 (seizurogenic) of the median lethal dose (LD50) and evaluated for 29

signs of toxicity. Animals were stratified based on seizing activity to evaluate effects of soman 30

exposure on fecal bacterial biota and urine metabolites. Soman exposure reshaped fecal bacterial 31

biota by altering Facklamia, Rhizobium, Bilophila, Enterobacter, and Morganella genera of the 32

Firmicutes and Proteobacteria phyla, some of which are known to hydrolyze OPs. However, 33

analogous changes were not observed in the bacterial biota of the ileum, which remained the 34

same irrespective of dose or seizing status of animals after soman intoxication. However, at 75 35

days post soman exposure, bacterial biota stabilized and no differences were observed between 36

groups. Interestingly, when considering just the seizing status of animals, we found that the urine 37

metabolome was markedly different. Leukotriene C4, kynurenic acid, 5-hydroxyindoleacetic 38

acid, norepinephrine, and aldosterone were excreted at much higher rates at 72 hrs in seizing 39

animals, consistent with early multi-organ involvement during soman poisoning. These findings 40

demonstrate the feasibility of using the dysbiosis of fecal bacterial biota in combination with 41

urine metabolome alterations as forensic evidence for pre-symptomatic OP exposure temporally 42

to enable administration of neuroprotective therapies of the future. 43

Importance: The paucity of assays to determine physiologically relevant OP exposure presents 44

an opportunity to explore the use of fecal bacteria as sentinels in combination with urine to 45

assess changes in the exposed host. Recent advances in sequencing technologies and 46

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 3: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

3

computational approaches have enabled researchers to survey large community level changes of 47

gut bacterial biota and metabolomic changes in various biospecimens. Here, we profiled changes 48

in fecal bacterial biota and urine metabolome following a chemical warfare nerve agent 49

exposure. The significance of this work is a proof-of-concept that fecal bacterial biota and urine 50

metabolites are two separate biospecimens rich in surrogate indicators suitable for monitoring 51

OP exposure. The larger value of such an approach is that assays developed around these 52

observations can be deployed at any setting with a moderate clinical chemistry and microbiology 53

capability. This can enable estimation of the affected radius or to screen, triage, or rule out 54

suspected cases of exposures in mass casualty scenarios, transportation accidents involving 55

hazmats, refugee movements, humanitarian missions, and training settings when coupled to an 56

established and validated decision-tree with clinical features. 57

Keywords: soman, gut microbiome, 16S rRNA gene, and urine metabolome 58

Background 59

Despite the serious health threat posed to communities, organic derivatives of phosphorus 60

containing acids have a wide range of applications in modern society (1-3). OP-containing 61

products are in excessive use world-wide for the control of agricultural or household pests. OP-62

containing pesticides account for almost 38% of all pesticides used across the globe, leading to 63

nearly 3 million poisonings, over 200,000 deaths annually, and the contamination of numerous 64

ecosystems (4). In addition, application of OP derivatives as agents of war and terrorism in the 65

form of nerve agents poses a significant threat to both civilians and the warfighter. Exposure to 66

OP leads to various degrees of neurotoxicity due to cholinergic receptor hyperactivity, mediated 67

primarily by the inhibition of acetylcholinesterase (AChE) (5). The excessive accumulation of 68

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 4: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

4

acetylcholine leads to severe physiological complications that may manifest both as muscarinic 69

symptoms (e.g. lacrimation, salivation, diarrhea, miosis, and bradycardia) as well as nicotinic 70

symptoms (e.g. tachycardia, hypertension, convulsions, and paralysis of skeletal and respiratory 71

muscles) and death (1-3, 6, 7). 72

Soman, pinacolyl methylphosphonofluoridate or GD (German agent D), is one of the G class 73

nerve agents (volatile agents associated with inhalation toxicity) that inhibits AChE much more 74

rapidly but less specifically than V class nerve agents (viscous agents associated with 75

transdermal toxicity) (2, 6, 8). Whole body autoradiography studies in mice revealed that 76

intravenously administered tritiated-soman (3H-soman) spreads through the entire body in less 77

than 5 min (9). High levels of accumulation were noted in the lungs, skin, gallbladder, intestinal 78

lumen, and urine during the first 24 hrs. 3H-pinacolyl methylphosphoric acid (

3H-PMPA), a 79

hydrolyzed acid and primary metabolite of soman, was found to be concentrated in specific 80

organs such as lungs, heart, and kidneys within minutes of 3H-soman administration, which 81

reflected the highly reactive (i.e. rapid aging) nature of soman in vivo (10). Significant amounts 82

of soman were also detected in red blood cells, a major esterase depot, when compared to the 83

plasma. In addition, these studies revealed that the common route of excretion for PMPA, a 84

major soman metabolite, was via urine and the intestinal lumen content (9, 11). Interestingly, 85

only trace amounts of 3H-soman,

3H-PMPA, or

3H-methylphosphonic acid (hydrolyzed PMPA) 86

were observed in the central nervous system. Current clinical nerve agent exposure assessments 87

are primarily based on overt physiological reactions such as convulsions, loss of consciousness, 88

and salivation for high-dose exposures or pupil constriction and respiratory distress for low-dose 89

exposures (12, 13). Recent studies have also demonstrated the feasibility of identifying OP 90

hydrolysis products in hair and nail clippings to verify nerve agent exposure after 30 days (13, 91

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 5: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

5

14). Hence, monitoring asymptomatic exposure or verifying suspected exposure during the pre-92

symptomatic phase using minimally invasive and rapid molecular methods is an ideal approach 93

and capability. Thus, identification of new surrogate biomarkers of toxicity and/or exposure to 94

soman and other OPs is essential both from a clinical and a public health standpoint, especially 95

for triaging population level exposures. 96

Using an omics approach, we have assessed the potential value of correlation between changes in 97

fecal bacterial biota or urine metabolites and OP exposure to determine if these biospecimens are 98

suitable for diagnostic use for exposure surveillance and monitoring in a rat model of soman 99

exposure. More importantly, we filled in a knowledge gap of how OP exposure directly or 100

indirectly impacts bacterial communities of the mammalian gut as well as alter the global urine 101

metabolic profile. For over 20 years, specific species from the Bacteroidetes and Proteobacteria 102

phyla have been implicated in enhanced biodegradation of OP pesticides in the bioremediation 103

field. Therefore, exploring the role of the microbiome in a mammalian host’s response to OP is 104

the next logical step (4, 15). Furthermore, recent advances in sequencing technologies have 105

enabled a detailed analysis of structural changes in the gut microbiome, revealing the dynamic 106

ecosystem of the bacterial biota and its essential role in health and disease. We also identified 107

urine as a suitable specimen type for investigation in this study because urine consists of 108

numerous metabolites as outputs from multiple pathways and provides a snapshot of both local 109

and systemic physiological changes (16). With these concerns in mind, we focused our efforts on 110

exploring and describing soman-induced dysbiosis of the gut microbiota and alterations in the 111

urine metabolome. 112

Results 113

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 6: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

6

Clinical manifestation of soman insult 114

To establish soman-induced toxicity with and without seizure, Sprague-Dawley rats were 115

subcutaneously injected with saline or 0.8 LD50 (sub-seizurogenic and unmitigated by treatment) 116

or 1 LD50 (seizurogenic and mitigated by treatment) exposure equivalent of soman. To reduce 117

mortality, the animals given 1 LD50 were also administered atropine sulfate and HI-6 one minute 118

after exposure. Rats that developed seizures at 1 LD50 were given diazepam to control seizing. 119

Approximately 42% of the animals exposed to soman (0.8 or 1.0 LD50) experienced seizure 120

irrespective of dose or the medical treatment regimen administered. As expected, control rats did 121

not experience seizure from administration of the vehicle. Seizing animals experienced a notable 122

weight loss and displayed increased activity in the days immediately after soman poisoning (Fig. 123

1a and Fig. S1a). Body temperature was not altered between seizing and non-seizing groups (Fig. 124

S1b). The Racine scale score, a quantitative assessment of seizure-related activities such as 125

degrees of tremors, convulsions, and seizures, was significantly higher in seizing subjects as 126

compared to non-seizing subjects, as expected (Fig. 1b)(17-19). Based on EEG activity, body 127

weight, and Racine score, we broadly categorized our analysis groups into a non-seizing group 128

(no seizure, n= 13) or a seizing group (exposure seizure or sustained seizure, n=10). We also 129

further subdivided cohorts based on dose because of seizure differences (Fig. 1c and 1d). Fecal 130

matter, urine, and tissues were harvested from animals to examine the gut microbiota and urine 131

metabolic changes due to the soman insult. 132

Taxonomic changes to soman exposure 133

To assess the effect of soman exposure on the bacterial biota, we sequenced the V3 and V4 134

hypervariable regions of the 16S ribosomal RNA from the feces and ileum of all animals at 72 135

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 7: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

7

hrs. Collecting and sequencing specimens from individual animals enabled us to assess the effect 136

of dose- or seizing status- driven changes in individual bacterial compositions. By measuring 137

multiple alpha diversity estimates within a dose (0.8 LD50 or 1.0 LD50) or seizing status (non-138

seizing or seizing) specimens, we found statistically significant (p<0.05) altered diversity in the 139

fecal bacterial biota of seizing animals at both sub-seizurogenic 0.8 LD50 and seizurogenic 1.0 140

LD50 (with a pronounced effect in seizing animals of the 0.8 LD50 exposure) that was markedly 141

different than control or non-seizing groups (Fig. S2a). When measuring alpha diversity between 142

the ileum specimens of control and soman-exposed animals, however, a higher degree of taxa 143

similarity was observed irrespective of dose or seizing status (Figure S2b) of specimens. The 144

bacterial biota of the ileum was not altered by exposure to soman while the fecal bacterial biota 145

was substantially altered. 146

In order to investigate if organism abundance accounted for the alpha diversity differences 147

observed between the fecal and ileum bacterial biota in response to soman insult, we rank 148

ordered all phyla identified in the study based on relative abundance (data not shown). To our 149

surprise, no differences were observed between the fecal and ileum bacterial phyla abundance 150

distribution. In both fecal and ileum specimens, Firmicutes and Bacteroidetes accounted for the 151

most abundant phyla as expected followed by Verrucomicrobia, Tenericutes, and Actinobacteria. 152

However, within the ileum the relative abundances of all the phyla were evenly distributed 153

around 10% while the phyla within the fecal specimens displayed a wide range of distribution 154

from 6% to 18% (data not shown). To further understand if soman exposure-driven bacterial 155

biota differences were attributable to dose or seizure individually, we measured beta diversity 156

using weighted UniFrac distances between specimens (fecal or ileum) and visualized the output 157

using principal coordinate analysis (PCoA) (Fig. S3). However, due to overlapping clusters, the 158

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 8: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

8

PCoA was unable to distinguish any of the groups into significant clusters based on dose or 159

seizing status in either the ileum or fecal specimens. 160

Effect of soman exposure on fecal bacterial biota composition 161

Next, we examined the microbial communities of the feces and ileum of each subject and 162

enumerated each phylum to decipher where microbial diversity was substantially altered by 163

exposure to soman, based on dose or seizure status. Several structural changes were observed in 164

the fecal microbiota compositions correlating with dose or seizure status of animals (Fig. 2). We 165

found that the relative abundance of fecal TM7 phylum was reduced in response to increasing 166

dose of soman and seizure status of animals, concurrently. Conversely, increasing relative 167

abundance of fecal Proteobacteria and Cyanobacteria trended with increase in soman dose insult 168

and seizing status of animals (Fig S4). However, the relative abundance of Actinobacteria, 169

Firmicutes, Tenericutes, and Verrucomicrobia phyla was unchanged in response to soman dose 170

or seizing status of animals. The Bacteroidetes phylum showed a positive trend of relative 171

abundance increase, although not statistically significant. Similar microbial community 172

composition analyses were also completed for ileum specimens collected from each individual 173

subject to identify if any changes were masked during the initial analyses (Fig. S5). Consistent 174

with previous findings in the ileum, no significant structural bacterial biota changes were 175

observed in any subjects in response to increase in dose or seizing status. 176

Furthermore, to measure the bacterial biota diversity and relative abundance (with respect to 177

quantitative and qualitative analysis), we looked into higher phylogenetic resolution and 178

taxonomic breakdowns. Distinct bacterial populations were observed at the genus level of 179

Firmicutes and Proteobacteria in the fecal specimens of soman-exposed animals (Fig. 3). The 180

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 9: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

9

Facklamia genus of the Aerococcaceae family was only detected in the fecal specimens of 181

soman-exposed rats and not in the ileum tissues content. Rhizobium, Bilophila, Enterobacter, and 182

Morganella genera of the Proteobacteria phylum exhibited expansion in the soman exposed rats 183

similar to that observed for the Facklamia genus. Interestingly, the Rhizobium genus, formerly 184

known as Agrobacterium, was primarily identified in the lower dose (0.8 LD50) soman exposure 185

group while Bilophila, Blautia, Enterobacter, and Morganella genera were detected in all soman 186

doses irrespective of seizing status of animals (Fig S6). Furthermore, several unclassified OTUs 187

were observed for Alcaligenaceae and Comamonadaceae of the Burkholderiales order and some 188

Enterobacteriaceae families, primarily in the seizing high soman dose exposure animals, and 189

these genera clusters were not observed in the control group or the low soman dose exposure 190

group. All difference in bacterial biota composition here were only observed at the 72 hr time 191

point and negligible at 75 days post exposure (data not shown). 192

Soman exposure alters the urine metabolome 193

The untargeted metabolomic profiling of biological substrates within a biospecimen provides 194

direct and simultaneous measurements of catabolic biochemical outputs that make up a given 195

phenotype (20). Comparative metabolomics profiling of urine specimens was performed to 196

assess soman exposure-associated metabolic alterations at 72 hrs post exposure, for seizing and 197

non-seizing animals. XCMS was used to pre-process metabolomics data to generate an analysis 198

matrix of mass-over-charge, retention time, fold change, and p-value for over 500 unique 199

metabolite peaks identified from both positive and negative mode analyses. Volcano plots from 200

Metaboanalyst v3.0. were used to visualize and enrich for metabolites with ≥ 2 fold change at 201

p<0.05 (Fig. S7). Furthermore, PCA was used to visualize if seizing and non-seizing groups 202

clustered separately (Fig S8). At the 72 hr time point, we found >200 metabolite peaks with 203

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 10: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

10

>2.0 fold change, of which 9 analytes had >20 fold change in seizing rats rather than non-seizing 204

rats. Many of these metabolites with >20 fold change appear to be either food-derived 205

metabolites such as methylmaysin and acetylpicropolin, or cresol derivatives. We also found a 206

large fraction of modified amino acids (such as acetylated, methylated, or formylated) common 207

urine solutes along with many phenyl and indole compounds. We used an in-house algorithm to 208

identify bacterial metabolites that are known to be detected in urine. Hence, we identified routine 209

urine metabolites associated with microbial origin such as p-cresol, D-alanine, and phenylacetic 210

acid. Interestingly, these metabolites were detected at a significantly higher rate (>5 fold change) 211

in the urine of seizing rats than non-seizing rats. The tryptophan catabolism by-product 212

kynurenic acid, the inflammatory lipid leukotriene C4, and the neurotransmitter norepinephrine 213

were also secreted in the urine at a significantly higher rate in seizing animals (Table 1) (21). 214

However, urinary citric acids levels remained unaltered between seizing and non-seizing 215

animals, suggesting that the presence of physiological indicators of physiological dysregulation 216

(acidemia and inflammation) in the urine of soman-exposed animals was not primarily the result 217

of kidney injury. 218

To gain insights into biological networks perturbed by soman in seizing rats, we annotated all 219

molecular networks associated with candidate metabolites identified at the 72 hrs time point in 220

an unbiased manner using publicly available databases. Consistent with the neurotoxicity 221

pathology of soman assault, large proportions of the metabolites in the urine of seizing animals 222

were primarily related to nervous system signaling activity representing norepinephrine and 223

serotonergic pathway products (data not shown). Next, we narrowed down our pathway 224

annotation to only consider best peak-matched metabolites (such as highest mass accuracy 225

(lowest dppm) and frequent database hits.) and to manually remove unlikely identifications, such 226

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 11: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

11

as drug action pathways, to restrict off-target hits in our annotation algorithm. This careful focus 227

of the data input revealed key molecular networks (catabolic processes) reflective of products 228

associated with kidney function, canonical central nervous system (CNS) inflammation, amino 229

acid and lipid metabolism, and vitamin absorption (Fig. 4). To increase the confidence in this 230

analysis, we have focused our study only on a select set of metabolites that were validated. 231

Discussion 232

Here, we employed a high throughput systems approach to survey changes in the gut bacterial 233

biota and urine metabolites and to explore new targets with potential diagnostic value for OP 234

exposure and toxicity in noninvasively collected specimens. We initially noted that seizing 235

subjects were significantly different than controls and non-seizing subjects based on multiple 236

parameters, including alpha diversity measures. However, composition analysis indicated that 237

the TM7 and Cyanobacteria phyla were probably the main drivers of the large observed alpha 238

diversity differences while the Proteobacteria phylum was a modest contributor to these 239

observed differences. Unfortunately, Cyanobacteria represent organisms such as chloroplasts, 240

and the TM7 phylum is not well understood. Due to limited knowledge, we concluded that these 241

two phyla are insufficient for drawing meaningful conclusions for the observed biological 242

differences at this moment. Thus, we focused our efforts to closely examine the genera 243

compositions of the various phyla identified in our study especially Proteobacteria. To this end, 244

we noted the presence of Facklamia genus, a Firmicutes, only in the feces and not in the ileum 245

digesta of rats exposed to soman. This genus is hardly associated with invasive disease, but it is 246

known to be occasionally isolated from specimens of urinary tract infections and 247

chorioamnionitis infections (22). We also observed Blautia, another Firmicutes, which was 248

highly enriched only in the feces of soman-exposed animals. Unfortunately, limited data is 249

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 12: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

12

available regarding the metabolic functions specifically hydrolysis, reduction, or esterification of 250

xenobiotics for these organisms within the gut and additional investigations are needed in this 251

area. 252

Many of the organophosphate-degrading genes (opd) implicated in the enhanced degradation of 253

OPs are located on mobile elements known to be transferred between organisms (23). In 254

addition, a majority of these genes exhibit a broad specificity and activity against OPs by either 255

directly hydrolyzing the phosphoester bonds in organophosphorus compounds or by further 256

degradation of methylphosphonate-esters to reduce toxicity or reactivation of OP (4, 23). Such 257

genes have been isolated from select species of Flavobacterium, Pseudomonas, Alteromonas, 258

Burkholderia, Bacillus, Alcaligense, Enterobacter, and Rhizobium genera, to name a few. Since 259

we had identified multiple genera such as Enterobacter and Rhizobium that are known to be 260

carriers of opd genes and involved in enhanced biodegradation of OPs, we investigated opd 261

genes from pooled fecal specimens and detected several targets that indicated the expected gene 262

templates were present in the bacterial biota (Fig. S9). However, we do not have evidence for 263

either a specific species that is carrying the opd gene templates or transcript data supporting 264

active expression of the genes (no RNA or protein). Rhizobium, an Alphaproteobacteria, is an 265

aerobic, oxidase-positive, Gram-negative bacillus isolated from the soil environment and known 266

to cause plant tumors (24, 25). Currently, there is limited information about which species of 267

Rhizobium reside in the gut of mammalians as routine commensals, which will be possible to 268

identify with whole genome sequencing approach, and whether these have functional role (15, 269

24). The Bilophila genus is a Deltaproteobacteria, where B. wadsworthia is the only known 270

species in this genus. B. wadsworthia is known to thrive on taurine and H2 production, 271

especially around highly fermentative sites (26). We found that B. wadsworthia was highly 272

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 13: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

13

enriched and detected in all soman-exposed cohorts irrespective of dose or seizing status but not 273

in control animals suggesting an unusually highly anaerobic environment could have promoted 274

this organism’s detection. Hence, we decided to validate B. wadsworthia using a PCR assay 275

targeting the 16S rRNA in fecal DNA extracts, pooled due to limited original specimens (Fig. 276

S9). We observed that B. wadsworthia levels were altered across pooled groups, especially in 277

seizing animals with high dose exposure, and were faintly expressed in controls. 278

Seizures that accompany soman intoxication lead to profound brain damage through excitotoxic 279

cell death of neurons, accompanied by neuroinflammation (27). Along this line, we detected 280

elevated levels of tryptophan catabolism (>2 fold change) revealed by the detection of kynurenic 281

acid and quinoline, in the urine of seizing animals. Kynurenic acid is a putative neuroprotective 282

metabolite and an antagonist of N-methyl-D-aspartate (NMDA) receptors (21, 28, 29). Elevated 283

levels of quinolone indicate an overall increased systemic tryptophan catabolism during soman 284

intoxication in seizing animals, presumably driven by inflammation in the brain, known to 285

induce tryptophan-degrading indoleamine-2,3-dioxygenase (IDO) (28, 29). This notion of the 286

inflammatory neuropathology of soman exposure was further supported by the detection of 287

increased systemic levels of the eicosanoid inflammatory mediator, leukotriene C4 (LTC4), in 288

seizing animals’ urine. LTC4 is one of the cysteinyl leukotrienes (cys-LTs) generated by the 289

enzymatic oxidation of arachidonic acid, a fatty acid released from neuronal membrane 290

glycerophospholipids during a secondary phase of brain injury (via a cascade of physiological 291

reactions to primary injury) (30, 31). Furthermore, after acute soman intoxication and seizure 292

onset, high levels of acetylcholine accumulate in the CNS. This is followed by decreased levels 293

of norepinephrine, aspartate, and glutamate (GLU) (i.e. excitatory amino acid (EAA) 294

neurotransmitters) in the brain (32, 33). Post neuronal damage, the sustainment of seizures 295

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 14: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

14

involve increases in concentration of EAAs like GLU along with norepinephrine, serotonin (5-296

HT), and its’ metabolite 5-hydroxyindolacetic acid (5-HIAA) (27). In our experimental system, 297

the primary stimulant soman is known to cause severe neuropathology, specifically inflammation 298

of the CNS. Consistent with this notion, we observed and validated increased levels of the 299

neurotransmitters norepinephrine (> 11 fold change), the serotonin metabolite 5-HIAA, and 300

increased levels of the cys-LT LTC4 in the urine metabolite profile of seizing animals, clearly 301

indicating increased inflammation associated with CNS neuropathology. Aldosterone is an 302

essential mineralocorticoid hormone directly involved in the regulation of sodium absorption and 303

potassium excretion in the kidney, salivary glands, sweat glands, and colon (34). Elevated levels 304

of aldosterone are known to alter glomerular structure and function via pro-oxidative and pro-305

fibrotic changes (35). Hence, aldosterone increases glomerular permeability to albumin, leading 306

to increased protein urinary excretion. We found that urinary excretion of aldosterone is more 307

than 5 times higher in seizing rats than non-seizing rats, suggesting that symptomatic soman 308

exposure may cause proteinuria and increased electrolyte excretion. Furthermore, in this context, 309

aldosterone-driven nephropathy and renal damage of modest effect are possible secondary 310

features of soman insult that can be easily exploited for pre-symptomatic monitoring of 311

exposure. However, our experimental design did not include proper specimen preservation or 312

analysis of renal function, specifically protein in urine to use readily available gold standard 313

clinical pathology assays. 314

Xenobiotic metabolites of microbial origin have been previously detected in urine (36, 37). In 315

our study, we identified a few of the most common microbial urinary metabolites, p-cresol and 316

phenylacetic acid (38). Both of these metabolites are a result of tyrosine catabolism by colon 317

microbes then sulfated by the colonic epithelium, resulting in reabsorption back into the host. 318

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 15: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

15

High levels of p-cresol sulfate are correlated with cardiovascular diseases as well as the mortality 319

of chronic kidney disease patients (36). Furthermore, the phenolic features of such compounds 320

are speculated to mimic neurotransmitters and interfere with the blood-brain barrier(36). 321

Consistent with our finding of dysbiosis of the gut bacterial biota, the soman-exposed animals in 322

our study had a significant increase in p-cresol (27x) and phenylacetic acid (7x) in their urine as 323

compared to non-seizing animals. However, we did not obtain enough 16S sequencing resolution 324

for species-level information in our efforts to identify organisms, especially those associated 325

with tyrosine catabolism such as Clostridium difficile (36). 326

Although this was a proof-of-concept (i.e. feasibility study) that applied untargeted technologies, 327

we were able to enrich for candidate diagnostic markers both from the bacterial biota and from 328

urine solutes and to also identify organisms with a potential role as forensic signatures of 329

exposure. With additional resources and better targeted technology, we plan to pursue a more 330

detailed analysis on a large cohort of animals, specifically to temporally populate quantitative 331

ratios of specific bacteria and urine metabolites in male and female rats. Furthermore, we plan to 332

profile multiple chemical nerve agents and pesticides. Based on these findings, one can envision 333

moderately complex assays with a correlation matrix and a decision algorithm to enhance current 334

clinical microbiology and urine chemistry workflows for diagnostics purposes. This would be 335

valuable for pre-symptomatic phase screening to enable disease modifying therapies, 336

determining the radius of impact during mass exposure, and casual/passive screening of training 337

sites and camps, as well as monitoring of refugees from non-invasive samples. 338

Methods and Materials 339

Animals 340

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 16: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

16

Male Sprague-Dawley rats (250-300 g; Charles River Laboratories (Kingston, NY)) were 341

individually housed on a 12:12 hrs light cycle with ad libitum access to food and water. Rats 342

were weighed daily. The experimental protocol was approved by the Institutional Animal Care 343

and Use Committee at the United States Army Medical Research Institute of Chemical 344

Defense(IACUC U-908), and all procedures were conducted in accordance with the principles 345

stated in the Guide for the Care and Use of Laboratory Animals and the Animal Welfare Act of 346

1966 (P.L. 89–544), as amended. 347

348

Surgery and EEG recording 349

Rats were surgically implanted with a subcutaneous transmitter (F-40EET; Data Sciences 350

International, Inc. (DSI; St. Paul, MN)) as described in Schultz et al. (2014), to record bi-351

hemispheric cortical EEG waveform activity as well as body temperature and activity throughout 352

the duration of the experiment. Surgery was conducted under isoflurane (3 - 4% induction; 1.5 -353

3% maintenance) and rats received buprenorphine (0.03 mg/kg, sc; Reckitt Benckiser 354

Pharmaceuticals, Inc, Richmond, VA) immediately after full recovery from anesthesia. Rats 355

recovered from surgery for 7-14 days prior to soman -exposure. RPC-1 physiotel receivers (DSI) 356

were placed under the rats’ home cage for EEG acquisition (24 hrs/day) with baseline recordings 357

made at least 24 hrs prior to exposure. Data were digitized at 250 Hz and recorded using 358

Dataquest ART 4.1 (Acquisition software; DSI). 359

360

Soman Exposures 361

Rats were exposed 0.8 or 1.0 LD50 soman (0.5 mL/kg, using 157.4 and 196.8 µg/ml respectively) 362

or saline (control) and evaluated for seizure activity. Soman LD50 is 98.4 g/ml (39). Soman was 363

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 17: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

17

obtained from the Edgewood Chemical Biological Center (Aberdeen Proving Ground, MD). To 364

promote survival, rats that received 1.0 mg/kg soman were treated with medical countermeasures 365

(MCM; an admix of 2 mg/kg atropine sulfate (ATS, Sigma-Aldrich Chemical Company, St. 366

Louis, MO, USA) and 93.6 mg/kg HI-6 (0.5 ml/kg, i.m., Starkes Associates, Buffalo, NY, USA)) 367

at 1 min after exposure and rats that developed seizures were treated with 10 mg/kg diazepam 368

(DZP; 2 ml/kg, s.c., Hospira Inc., Lake Forest, Illinois, USA) at 30 min after seizure onset (with 369

average seizure onset of 8 min). 370

371

Behavioral seizure 372

Behavioral seizures were scored using a modified Racine scale (17-19): stage 1: mastication, 373

tongue fasciculation, oral tonus; stage 2, head tremors, head bobs; stage 3, limb clonus or tonus, 374

body tremor; stage 4, rearing with forelimb clonus; and stage 5, rearing and falling with 375

generalized convulsions. For analysis, rats received a score corresponding to the maximum stage 376

reached per time interval. Observations were made continuously for up to 5 hrs after exposure to 377

soman. 378

379

EEG analysis 380

Full-power spectral analysis of EEG, identification of epileptiform activity, and other EEG 381

anomalies were analyzed according to the methods described previously (40). EEG recorded 382

seizures were confirmed through visual screening and characterized by sustained frequencies and 383

the value of the most prominent frequencies in Hz (e.g.: the highest power calculated by MatLab 384

program (www.mathworks.com) in µV2/Hz). Some soman -exposed rats developed seizure 385

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 18: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

18

activity (SE) with seizure onset at ~ 8 min for those exposed to 1.0 LD50, while others did not 386

develop seizures. 387

388

Sample collection 389

All surfaces and tools, including the guillotine, collection foils, sample tubes, saline bottle, and 390

syringes, were sprayed with RNase AWAY®. Animals were administered Fatal-Plus® (sodium 391

pentobarbital) and once fully anesthetized the rats were euthanized using a guillotine. Urine was 392

collected directly from the bladder using a syringe with 21 gauge needle, saved in RNase-free 393

microfuge tubes, and flash frozen in liquid nitrogen. The ileum was removed and flash frozen. 394

Digesta was flushed out by rinsing with sterile saline during sample processing. Organs were 395

collected in foils and flash frozen in liquid nitrogen. Fresh fecal pellets were collected during rat 396

handling and stored at -80°C for processing. 397

398

DNA extraction 399

Samples were kept cold at all time before the extraction. The Ileum tissues were weighed and 400

homogenized in 50 mM Tris-HCl (Lonza, Walkersville, MD, USA) with 2 nM EDTA Solution 401

(Lonza) using the BeadBeater (Bio Spec Products, Inc., Bartlesville, OK, USA), the DNA 402

extraction was carried out using the QIAGEN DNeasy Blood and Tissue Kit (QIAGEN Inc., 403

Germantown, MD, USA), and RNA was extracted using TRIzol Reagent (Invitrogen, Life 404

Technologies, Grand Island, NY, USA) in conjunction with the QIAGEN miRNeasy Mini Kit 405

(QIAGEN). The fecal samples were weighed and RNA was extracted using MoBio PowerSoil 406

Total RNA Isolation Kit (MO BIO Laboratories, Inc, Carlsbad, CA, USA) and DNA was 407

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 19: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

19

extracted using the MoBio PowerSoil DNA Elution Accessory Kit. The extracted DNA was used 408

for PCR and sequencing. 409

410

Library preparation and sequencing 411

We used primers that were previously designed to amplify the V3-V4 hyper-variable regions of 412

the 16S rRNA gene (41). A limited cycle PCR generated a single amplicon of ~460 bp, and this 413

was followed by addition of Illumina sequencing adapters and dual‐index barcodes. Using paired 414

300 bp reads, and MiSeq v3 reagents, the ends of each read were overlapped to generate high‐415

quality, full‐length reads of the V3 and V4 region in a single run. 416

417

Data Analysis 418

Sequence read quality assessment, filtering, barcode trimming, and chimera detection were 419

performed on de-multiplexed sequences using the USEARCH method in the Quantitative 420

Insights Into Microbial Ecology (QIIME) package (v.1.9.1) (42). OTUs were defined by 421

clustering with 97% sequence similarity cutoffs (at 3% divergence). The representative sequence 422

for an OTU was chosen as the most abundant sequence showing up in that OTU’s by collapsing 423

identical sequences, and choosing the one that was read the most abundant sequences. Then 424

representative sequences were aligned against Greengenes database core set (v.gg_13_8) using 425

PyNAST alignment method (www.greengenes.secondgenom.com). The minimum sequence 426

length of 150nt and the minimum percent of 75% match were used for the alignment (43, 44). 427

The RDP Classifier program (v.2.2) was used to assign the taxonomy to the representative set of 428

sequences using a pre-built database of assigned sequence of reference set (45). Alpha Diversity 429

was performed using PhyloSeq R package(www.bioconductor.org), Chao1 metric (estimates the 430

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 20: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

20

species richness.), the observed species metric (the count of unique OTUs found in the sample) 431

and Phylogenetic Distance (PD_whole_tree) were calculated (46). Similarly, Beta Diversity were 432

calculated using QIIME and visualized using Principal Coordinate Analysis (PCoA) to visualize 433

distances between samples on an x-y-z plot. The ranked abundance profile was created using 434

BiodiversityR R Bioconductor package (2.7-2) to highlight the most abundant phylum in all 435

samples. To determine the differentially abundant taxonomic groups over different groups in 436

soman against non-soman exposed at different doses and seizing versus non-seizing groups were 437

examined by fitting linear models using moderated standard errors and the empirical Bayes 438

model following TMM normalization on OTU count. The normalized abundance profile was 439

created across different doses (0.8 and 1.0 LD50) of soman exposure compared with control 440

samples, similarly seizing versus non-seizing rats compared against non soman exposed control 441

samples. Sequences can be accessed from the NCBI Sequence Read Archive (SRA) at study 442

accession SRP116704 bioproject PRJNA401162. 443

To predict the metabolites contributed as per microbial composition we used PICRUSt (v.1.0.0), 444

open source tool that uses precomputed gene content inference for 16S rRNA. PICRSUSt uses 445

the OTU abundance count generated using ‘closed-reference’ OTU picking against Greengenes 446

database, for normalization of OTU table, each OTU was first divided by known as well as 447

predicted 16s copy number abundance (47). Final metagenome functional predictions were 448

performed by multiplying normalized OTU abundance by each predicted functional profile. 449

Statistical hypothesis testing analysis of metagenomics profiles was performed using R to 450

compare KEGG Orthologs (v.80.0) between pre- and post-Soman exposed samples and Principal 451

coordinates analysis was also performed. The predicted metagenome functional counts were 452

normalized using TMM normalization method to fit linear models using contrast function to 453

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 21: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

21

compute fold changes by moderating the standard errors using empirical Bayes model. Logodds 454

and moderated t- statistic of differential predicted significant metabolite derived from KEGG 455

Orthologs was computed between pre- and post-soman exposed samples at different doses and 456

seizing vs non seizing groups with p value cutoff ≤ 0.05. Significant metabolites were further 457

annotated using in-house metabolite annotation function with other databases such as HMDB 458

(v.2.5), KEGG (compounds, pathways, orthologs and reactions) (v.80.0), SMPDB (v.2.0) and 459

FOODB (v.1.0). 460

Metabolomic profiling and data analysis 461

Urine samples were processed using the method of Tyburski (48). Briefly, the samples were 462

thawed on ice and vortexed. For metabolite extraction, 20 µL of urine was mixed with 80 µL of 463

50% acetonitrile (in water) containing internal standards (10 µL of debrisoquine (1mg/mL) and 464

50 µL of 4, nitro-benzoic acid (1mg/mL)). The supernatant was transferred to a fresh tube and 465

used for UPLC-ESI-Q-TOF-MS analysis (Xevo G2, Waters Corporation). Each sample (5 μL) 466

was injected onto a reverse-phase 50 × 2.1 mm BEH 1.7 μm C18 column using an Acquity 467

UPLC system (Waters Corporation, USA). The gradient mobile phase was comprised of water 468

containing 0.1% formic acid solution (A) and acetonitrile containing 0.1% formic acid solution 469

(B). Each sample was resolved for 10 min at a flow rate of 0.5 ml/min. This approach has been 470

extensively used for metabolomic profling of biofluids; UPLC gradients conditions and the mass 471

spectrometry parameters and has been described in details (49-51). The UPLC gradient consisted 472

of 100% A for 0.5 min then a ramp of curve 6 to 60% B from 0.5 min to 4.5 min, then a ramp of 473

curve 6 to 100% B from 4.5 to 8.0 min, a hold at 100% B until 9.0 min, then a ramp of curve 6 to 474

100% A from 9.0 min to 9.2 min, followed by a hold at 100% A until 10 mins. The column 475

eluent was introduced directly into the mass spectrometer by electrospray. Mass spectrometry 476

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 22: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

22

was performed on a Quadrupole-time-of-flight mass spectrometer operating in either negative or 477

positive electrospray ionization mode with a capillary voltage of 3.2 KV and a sampling cone 478

voltage of 35 V. The desolvation gas flow was 800 L/h and the temperature was set to 350°C. 479

The cone gas flow was 50 L/h, and the source temperature was 150°C. The data was acquired in 480

the V mode with scan time of 0.3 seconds, and inter-scan delay at 0.08 seconds. Accurate mass 481

was maintained by infusing sulfadimethoxine (311.0814 m/z) in 50% aqueous acetonitrile (250 482

pg/µL) at a rate of 30 µL/min via the lockspray interface every 10 seconds. Data were acquired 483

in centroid mode from 50 to 850 m/z mass range for TOF-MS scanning, in duplicates (technical 484

replicates) for each sample in positive and negative ionization mode and checked for 485

chromatographic reproducibility. For all profiling experiments, the sample queue was staggered 486

by interspersing samples of the two groups to eliminate bias. Pooled sample injections 487

throughout the run (one pool was created by mixing 2 µL aliquot from all 110 samples) were 488

used as quality controls (QCs) to assess inconsistencies that are particularly evident in large 489

batch acquisitions in terms of retention time drifts and variation in ion intensity over time. QCs 490

were projected in the orthogonal partial least squares-discriminant analysis (OPLS-DA) model 491

along with the study samples to ensure that the technical performance did not impact the 492

biological information(52). The raw data were pre-processed using the XCMS (53) software for 493

peak detection and alignment. The resultant three dimensional data matrix consisting of 494

mass/charge ratios with retention times and feature intensities was subjected to multivariate data 495

analysis using Metaboanalyst v 3.0. Quantitative descriptors of model quality for the OPLS-DA 496

models included R2 (explained variation of the binary outcome: Treatment vs. control and Q2 497

(cross-validation based predicted variation of the binary outcome). We used score plots to 498

visualize the discriminating properties of the OPLS-DA models. The features selected via OPLS- 499

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 23: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

23

were used for accurate mass based database search; subsequently the identity of a sub-set of 500

metabolites was confirmed using tandem mass spectrometry. 501

Ethics approval and consent to Participate 502

This research complied with the Animal Welfare Act and implementing Animal Welfare 503

Regulations, the Public Health Service Policy on Humane Care and Use of Laboratory Animals, 504

and adhered to the principles noted in The Guide for the Care and Use of Laboratory Animals 505

(NRC, 2011). 506

The views, opinions, and findings contained in this report are those of the authors and should 507

not be construed as official Department of the Army position, policy, or decision, unless so 508

designated by other official documentation. Citations of commercial organizations or trade 509

names in this report do not constitute an official Department of the Army endorsement or 510

approval of the products or services of these organizations. 511

Consent for publication 512

Not Applicable. Human subjects did not participate in this study. 513

514

Availability of data and materials 515

Data generated and analyzed during this study are included in this published article and 516

supplementary information files with the exception of raw urine mass spectrometry and 75 day 517

microbiome assay, which will be available from the corresponding author on reasonable request. 518

Sequences can be accessed from the NCBI Sequence Read Archive (SRA) at study accession 519

SRP116704 bioproject PRJNA401162. 520

521

Competing Interest 522

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 24: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

24

The authors declare no competing interests. BARDA was not involved in the study design or in 523

the collection, analysis and interpretation of data or the decision to write this manuscript and 524

submit it for publication. The views expressed in this manuscript are those of the authors and do 525

not reflect the official policy of the Department of Army, Department of Defense or the US 526

Government. 527

528

Funding 529

Support was provided by an interagency agreement between the Biomedical Advanced Research 530

and Development Authority (BARDA), the Geneva Foundation, and the US Army Medical 531

Research Institute of Chemical Defense (USAMRICD) as well as a memorandum of agreement 532

between USAMRICD and US Army Center of Environmental Health (USACEHR). The 533

Metabolomics Shared Resource in Georgetown University (Washington DC, USA) partially 534

supported by NIH/NCI/CCSG grant P30-CA051008. 535

536

Authors’ Contributions 537

DG analyzed data and wrote the manuscript. AG coordinated microbiome and metabolomics 538

analyses and obtained the samples, AH processed microbiome specimens and completed 539

validation, RK analyzed the data, and AKC completed metabolomics analysis. FR analyzed the 540

EEG data, CS conducted animal experiments. LL, MJ and RH conceived and designed the study 541

and edited manuscript. All authors read and approved the final manuscript. 542

543

Acknowledgments 544

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 25: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

25

We thank Ms. Kirandeep Gill (Georgetown University) for technical assistance with 545

metabolomics data as well as Dr. Matthew Rice and Dr. Julia Scheerer for editorial input. 546

547

548

References 549

1. Abdollahi M, Karami-Mohajeri S. 2012. A comprehensive review on experimental and 550

clinical findings in intermediate syndrome caused by organophosphate poisoning. 551

Toxicol Appl Pharmacol 258:309-14. 552

2. Marrs TC. 1993. Organophosphate poisoning. Pharmacol Ther 58:51-66. 553

3. Acon-Chen C, Koenig JA, Smith GR, Truitt AR, Thomas TP, Shih TM. 2016. Evaluation 554

of acetylcholine, seizure activity and neuropathology following high-dose nerve agent 555

exposure and delayed neuroprotective treatment drugs in freely moving rats. Toxicol 556

Mech Methods 26:378-88. 557

4. Singh BK, Walker A. 2006. Microbial degradation of organophosphorus compounds. 558

FEMS Microbiology Reviews 30:428-471. 559

5. Shih T-M, McDonough JH. 1997. Neurochemical Mechanisms in Soman-induced 560

Seizures. Journal of Applied Toxicology:255-264. 561

6. Misik J. 2014. Acute toxicity of some nerve agents and pesticides in rats. Drug and 562

Chemical Toxicology 32-36. 563

7. Kassa J, Korabecny J, Sepsova V, Tumova M. 2014. The Evaluation of Prophylactic 564

Efficacy of Newly Developed Reversible Inhibtors of Acetylcholinesterase in Soman-565

Poisoned Mice - A comparison with commonly used Pyidostigmine. Basic and Clinical 566

Pharmacology and Toxicology:571-576. 567

8. Villa AF, Houze P, Monier C, Risede P, Sarhan H, Borron SW, Megarbane B, Garnier R, 568

Baud FJ. 2007. Toxic doses of paraoxon alter the respiratory pattern without causing 569

respiratory failure in rats. Toxicology 232:37-49. 570

9. Kadar T, Raveh L, Cohen G, Oz N, Baranes I, Balan A, Ashani Y, Shapira S. 1985. 571

Distribution of H-3 Soman in Mice. Archives of Toxicology 58:45-49. 572

10. Reynolds ML, Little PJ, Thomas BF, Bagley RB, Martin BR. 1985. Relationship between 573

the Biodisposition of [H-3] Soman and Its Pharmacological Effects in Mice. Toxicology 574

and Applied Pharmacology 80:409-420. 575

11. Shih ML, Mcmonagle JD, Dolzine TW, Gresham VC. 1994. Metabolite 576

Pharmacokinetics of Soman, Sarin and Gf in Rats and Biological Monitoring of Exposure 577

to Toxic Organophosphorus Agents. Journal of Applied Toxicology 14:195-199. 578

12. Brown MA, Brix KA. 1998. Review of health consequences from high-, intermediate- 579

and low-level exposure to organophosphorus nerve agents. Journal of Applied 580

Toxicology 18:393-408. 581

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 26: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

26

13. Appel AS, McDonough JH, McMonagle JD, Logue BA. 2016. Analysis of Nerve Agent 582

Metabolites from Hair for Long-Term Verification of Nerve Agent Exposure. Anal Chem 583

88:6523-30. 584

14. Appel AS, Logue BA. 2016. Analysis of nerve agent metabolites from nail clippings by 585

liquid chromatography tandem mass spectrometry. J Chromatogr B Analyt Technol 586

Biomed Life Sci 1031:116-22. 587

15. Horne I, Sutherland TD, Harcourt RL, Russell RJ, Oakeshott JG. 2002. Identification of 588

an opd (Organophosphate Degradation) Gene in an Agrobacterium Isolate. Applied and 589

Environmental Microbiology 68:3371-3376. 590

16. Countrepois K, Jiang L, Snyder M. 2015. Optimized Analytical Procedures for the 591

Untargeted Metabolomic Profiling of Human Urine and Plasma by Combining 592

Hydrophilic Interaction (HILIC) and Reverse-Phase Liquid Chromatography (RPLC)–593

Mass Spectrometry. Molecular and Cellular Proteomics 14:1684-1695. 594

17. Racine R, Chipashvili S, Okujava V. 1972. Modification of Seizure Activity by Electrical 595

Stimulation .3. Mechanisms. Electroencephalography and Clinical Neurophysiology 596

32:295-+. 597

18. Racine RJ. 1972. Modification of Seizure Activity by Electrical Stimulation .1. After-598

Discharge Threshold. Electroencephalography and Clinical Neurophysiology 32:269-&. 599

19. Racine RJ. 1972. Modification of Seizure Activity by Electrical Stimulation .2. Motor 600

Seizure. Electroencephalography and Clinical Neurophysiology 32:281-&. 601

20. Patti GJ, Yanes O, Siuzdak G. 2012. Innovation: Metabolomics: the apogee of the omics 602

trilogy. Nature Reviews Molecular Cell Biology 13:263-269. 603

21. Dantzer R, O'Connor JC, Lawson MA, Kelley KW. 2011. Inflammation-associated 604

depression: From serotonin to kynurenine. Psychoneuroendocrinology 36:426-436. 605

22. Lawson PA, Collins MD, Falsen E, SJoden B, Facklam R. 1999. Facklamia languida sp. 606

nov., Isolated from Human Clinical Specimens. Journal of Clinical Microbiology 607

37:1161-1164. 608

23. Karpouzas DG, Singh BK. 2006. Microbial Degradation of Organophosphorus 609

Xenobiotics: Metabolic Pathways and Molecular Basis. 51:119-225. 610

24. Pereira LA, Chan DSG, Ng TM, Lin R, Jureen R, Fisher DA, Tambyah PA. 2009. 611

Pseudo-Outbreak of Rhizobium radiobacter Infection Resulting from Laboratory 612

Contamination of Saline Solution. Journal of Clinical Microbiology 47:2256-2259. 613

25. Moreau-Gaudry V, Chiquet C, Boisset S, Croize J, Benito Y, Cornut PL, Bron A, 614

Vandenesch F, Maurin M. 2012. Three Cases of Post-Cataract Surgery Endophthalmitis 615

Due to Rhizobium (Agrobacterium) radiobacter. Journal of Clinical Microbiology 616

50:1487-1490. 617

26. da Silva SM, Venceslau SS, Fernandes CLV, Valente FMA, Pereira IAC. 2007. 618

Hydrogen as an energy source for the human pathogen Bilophila wadsworthia. Antonie 619

van Leeuwenhoek 93:381-390. 620

27. McDonough JH, Jr., Shih TM. 1997. Neuropharmacological mechanisms of nerve agent-621

induced seizure and neuropathology. Neurosci Biobehav Rev 21:559-79. 622

28. Savitz J, Dantzer R, Wurfel BE, Victor TA, Ford BN, Bodurka J, Bellgowan PSF, 623

Teague TK, Drevets WC. 2015. Neuroprotective kynurenine metabolite indices are 624

abnormally reduced and positively associated with hippocampal and amygdalar volume 625

in bipolar disorder. Psychoneuroendocrinology 52:200-211. 626

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 27: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

27

29. Savitz J, Drevets WC, Smith CM, Victor TA, Wurfel BE, Bellgowan PSF, Bodurka J, 627

Teague TK, Dantzer R. 2014. Putative Neuroprotective and Neurotoxic Kynurenine 628

Pathway Metabolites Are Associated with Hippocampal and Amygdalar Volumes in 629

Subjects with Major Depressive Disorder. Neuropsychopharmacology 40:463-471. 630

30. Farias S, Frey LC, Murphy RC, Heidenreich KA. 2009. Injury-Related Prodcution of 631

Cysteinyl Leukotrienes Contributes to brain Damage following experimnetal Traumatic 632

Brain Injury Journal of Neurotrauma 26:1977-1986. 633

31. Raithel M, Zopf Y, Kimpel S, Naegel A, Molderings G, Buchwald F, Schultis H, Kressel 634

J, Hahn E, Konturek P. 2011. The Measurement of Leukotrienes in Urine as Diagnostic 635

Option in Systemic Mastocytosis Journal of Physiology and Pharmacology 62:469-472. 636

32. Shih TM, McDonough JH, Jr. 1997. Neurochemical mechanisms in soman-induced 637

seizures. J Appl Toxicol 17:255-64. 638

33. Hu CY, Robinson CP. 1988. The Effects of Soman on Norepinephrine Uptake, Release, 639

and Metabolism. Toxicology and Applied Pharmacology 96:185-190. 640

34. Bizzarri C, Pedicelli S, Cappa M, Cianfarani S. 2016. Water Balance and ‘Salt Wasting' 641

in the First Year of Life: The Role of Aldosterone-Signaling Defects. Hormone Research 642

in Paediatrics 86:143-153. 643

35. Bernardi S, Toffoli B, Zennaro C, Bossi F, Losurdo P, Michelli A, Carretta R, Mulatero 644

P, Fallo F, Veglio F, Fabris B. 2015. Aldosterone effects on glomerular structure and 645

function. Journal of the Renin-Angiotensin-Aldosterone System 16:730-738. 646

36. Tanaka H, Sirich TL, Meyer TW. 2015. Uremic Solutes Produced by Colon Microbes. 647

Blood Purification 40:306-311. 648

37. Tanaka H, Sirich TL, Plummer NS, Weaver DS, Meyer TW. 2015. An Enlarged Profile 649

of Uremic Solutes. PLoS One 10:e0135657. 650

38. Meyer TW, Hostetter TH. 2012. Uremic solutes from colon microbes. Kidney 651

International 81:949-954. 652

39. Schultz MK, Wright LKM, Furtado MD, Stone MF, Moffett MC, Kelley NR, Bourne 653

AR, Lumeh WZ, Schultz CR, Schwartz JE, Lumley LA. 2014. Caramiphen edisylate as 654

adjunct to standard therapy attenuates soman-induced seizures and cognitive deficits in 655

rats. Neurotoxicology and Teratology 44:89-104. 656

40. Furtado MD, Lumley LA, Robison C, Tong LC, Lichtenstein S, Yourick DL. 2010. 657

Spontaneous recurrent seizures after status epilepticus induced by soman in Sprague-658

Dawley rats. Epilepsia 51:1503-1510. 659

41. Klindworth A, Pruesse E, Schweer T, Peplies J, Quast C, Horn M, Glockner FO. 2013. 660

Evaluation of general 16S ribosomal RNA gene PCR primers for classical and next-661

generation sequencing-based diversity studies. Nucleic Acids Res 41:e1. 662

42. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello EK, Fierer 663

N, Pena AG, Goodrich JK, Gordon JI, Huttley GA, Kelley ST, Knights D, Koenig JE, 664

Ley RE, Lozupone CA, McDonald D, Muegge BD, Pirrung M, Reeder J, Sevinsky JR, 665

Tumbaugh PJ, Walters WA, Widmann J, Yatsunenko T, Zaneveld J, Knight R. 2010. 666

QIIME allows analysis of high-throughput community sequencing data. Nature Methods 667

7:335-336. 668

43. McDonald D, Price MN, Goodrich J, Nawrocki EP, DeSantis TZ, Probst A, Andersen 669

GL, Knight R, Hugenholtz P. 2012. An improved Greengenes taxonomy with explicit 670

ranks for ecological and evolutionary analyses of bacteria and archaea. Isme Journal 671

6:610-618. 672

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 28: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

28

44. Caporaso JG, Bittinger K, Bushman FD, DeSantis TZ, Andersen GL, Knight R. 2010. 673

PyNAST: a flexible tool for aligning sequences to a template alignment. Bioinformatics 674

26:266-267. 675

45. Wang Q, Garrity GM, Tiedje JM, Cole JR. 2007. Naive Bayesian classifier for rapid 676

assignment of rRNA sequences into the new bacterial taxonomy. Applied and 677

Environmental Microbiology 73:5261-5267. 678

46. McMurdie PJ, Holmes S. 2013. phyloseq: An R Package for Reproducible Interactive 679

Analysis and Graphics of Microbiome Census Data. Plos One 8. 680

47. Langille MGI, Zaneveld J, Caporaso JG, McDonald D, Knights D, Reyes JA, Clemente 681

JC, Burkepile DE, Thurber RLV, Knight R, Beiko RG, Huttenhower C. 2013. Predictive 682

functional profiling of microbial communities using 16S rRNA marker gene sequences. 683

Nature Biotechnology 31:814-+. 684

48. Tyburski JB, Patterson AD, Krausz KW, Slavik J, Fornace AJ, Jr., Gonzalez FJ, Idle JR. 685

2009. Radiation metabolomics. 2. Dose- and time-dependent urinary excretion of 686

deaminated purines and pyrimidines after sublethal gamma-radiation exposure in mice. 687

Radiat Res 172:42-57. 688

49. Patterson AD, Maurhofer O, Beyoglu D, Lanz C, Krausz KW, Pabst T, Gonzalez FJ, 689

Dufour JF, Idle JR. 2011. Aberrant lipid metabolism in hepatocellular carcinoma revealed 690

by plasma metabolomics and lipid profiling. Cancer Res 71:6590-600. 691

50. Li F, Patterson AD, Krausz KW, Tanaka N, Gonzalez FJ. 2012. Metabolomics reveals an 692

essential role for peroxisome proliferator-activated receptor alpha (PPARalpha) in bile 693

acid homeostasis. J Lipid Res doi:jlr.M027433 [pii10.1194/jlr.M027433. 694

51. Tyburski JB, Patterson AD, Krausz KW, Slavik J, Fornace AJ, Jr., Gonzalez FJ, Idle JR. 695

2008. Radiation metabolomics. 1. Identification of minimally invasive urine biomarkers 696

for gamma-radiation exposure in mice. Radiat Res 170:1-14. 697

52. Bylesjo M, Rantalainen M, Cloarec O, Nicholson JK, Holmes E, Trygg J. 2006. OPLS 698

discriminant analysis: combining the strengths of PLS-DA and SIMCA classification. 699

Journal of Chemometrics 20:341-351. 700

53. Smith CA, Want EJ, O'Maille G, Abagyan R, Siuzdak G. 2006. XCMS: processing mass 701

spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and 702

identification. Anal Chem 78:779-87. 703

704

Figure Legends 705

Figure 1 Clinical manifestation of toxicity after soman exposure. Panel a represents trends in 706

body weight changes based on seizing status of rats before and after soman exposure. Panel b 707

represents Racine Score differences based on seizing status of rats after soman exposure. Panel c 708

and d represents seizure activities (latency to seize or seizure duration during 72hr monitoring) 709

of rats exposed to 0.8 LD50 of soman without treatment (unmitigated seizure model, n=4) or 1.0 710

LD50 of soman and treated with medical counter measure (mitigated seizure model, n=6) a 711

minute after exposure. 712

713

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 29: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

29

Figure 2 Trends in relative abundance of fecal bacterial biota (phylum level) based on dose 714

(Panel a) or seizing status (Panel b) of rats 72 hrs after soman exposure. TM7 phylum exhibited 715

decreased 716

normalized counts with increasing dose or seizing status while the inverse was observed for the 717

Proteobacteria phylum. 718

719

Figure 3 3 SPADE-like analysis showing taxonomic changes in Proteobacteria and Firmicutes 720

(Bacilli only) phyla. Circle size is assigned by automatically binning normalized OTU counts 721

into any of the three sizes. Distance between circles is assigned arbitrarily. Colors are added as a 722

visual aid for following branch and do not have connotation. Panel a represents average OTU 723

count of fecal bacterial biota detected for the two phyla in soman unexposed control animals. 724

Panel b shows representative average of OTU counts of bacterial biota detected after exposure 725

(solid circles) compared to bacterial biota of unexposed control (faded circles). Unassigned OTU 726

clusters at the order or genus level are represented by blue or red circles. 727

728

Figure 4 Snapshot of validated altered urine metabolites, affected candidate site, and parent 729

pathways. Circos image represents urine metabolites significantly altered (p<0.05) in seizing 730

animals (outer ring labels on white background) and associated catabolic pathways (outer ring 731

ribbons) and sites (ribbon colors). 732

733

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 30: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

Figure 1

Fig. 1 Clinical manifestation of toxicity after soman exposure. Panel a represents trends in body weight

changes based on seizing status of rats before and after soman exposure. Panel b represents Racine Score

differences based on seizing status of rats after soman exposure. Panel c and d represents seizure

activities (latency to seize or seizure duration during 72hr monitoring) of rats exposed to 0.8 LD50 of

soman without treatment (unmitigated seizure model, n=4) or 1.0 LD50 of soman and treated with medical

counter measure (mitigated seizure model, n=6) a minute after exposure.

A B

C D

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 31: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

Fig. 2 Trends in relative abundance of fecal bacterial biota (phylum level) based on dose (Panel a)

or seizing status (Panel b) of rats 72 hrs after soman exposure. TM7 phylum exhibited decreased

normalized counts with increasing dose or seizing status while the inverse was observed for the

Proteobacteria phylum.

Figure 2

B

Bacteroidetes Deferribacteres Firmicutes Proteobacteria TM7

CTRL NS SE CTRL NS SE CTRL NS SE CTRL NS SE CTRL NS SE

CTRL: Control

NS: Non-seizing

SE: Seizing

0.04

0.08

0.12

A Bacteroidetes Deferribacteres Firmicutes Proteobacteria TM7

0 0.8 1.0 0 0.8 1.0 0 0.8 1.0 0 0.8 1.0 0 0.8 1.0

0.04

0.08

0.12

0: No Soman

0.8: 0.8LD50 Soman

1.0: 1.0LD50 Soman

No

rm

ali

ze

d

co

un

t N

orm

ali

ze

d

co

un

t

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 32: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

Fig. 3 SPADE-like analysis showing taxonomic changes in Proteobacteria and Firmicutes (Bacilli only)

phyla. Circle size is assigned by automatically binning normalized OTU counts into any of the three sizes.

Distance between circles is assigned arbitrarily. Colors are added as a visual aid for following branch

and do not have connotation. Panel a represents average OTU count of fecal bacterial biota detected for the

two phyla in soman unexposed control animals. Panel b shows representative average of OTU counts of

bacterial biota detected after exposure (solid circles) compared to bacterial biota of

unexposed control (faded circles). Unassigned OTU clusters at the order or genus level are represented

by blue or red circles.

Figure 3

A

B

Control

Soman exposed

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 33: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

Fig. 4 Snapshot of validated altered urine metabolites, affected candidate site, and parent pathways.

Circos image represents urine metabolites significantly altered (p<0.05) in seizing animals (outer ring

labels on white background) and associated catabolic pathways (outer ring ribbons) and sites

(ribbon colors).

Figure 4

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

Page 34: Downloaded from on June 1, 2020 by guest · 3 47 computational approaches ha ve enabled researche rs to survey large community level changes of 48 gut bacterial biota and metabolomic

Table 1 Summary of soman-exposure altered urine solute

ESI

Mode

Fold

Change

p value

Name

Formula

RT

M/z

dppm

NEG 0.45193 0.00160 Guanosine-3',5'-

Diphosphate

C10H15N5O11P2 4.58 442.0211 9.092775

NEG 0.4864 0.00773 Citric acid C6H8O7 0.42 191.0192 2.781159

NEG 2.3912 0.04993 Uric acid C5H4N4O3 0.34 167.0199 6.901621

POS 2.2575 0.00119 Kynurenic acid C10H7NO3 1.57 190.0503 2.260052

POS 2.7 0.00409 Quinoline C9H7N 3.13 130.0645 4.847519

POS 3.4701 0.00000 Leukotriene C4 C30H47N3O9S 2.08 626.3129 3.806747

POS 11.883 0.00033 Norepinephrine

sulfate

C8H11NO6S 3.17 250.0398 7.291454

POS 3.0192 0.00018 5-

Hydroxyindoleacetic

acid

C10H9NO3 1.62 192.0666 5.670386

NEG 27.481 0.0018 p-Cresol C7H8O 3.11 107.0501 1.270889

POS 6.8123 0.0301 Phenylacetic acid C8H8O2 2.96 137.060 4.37399

POS 6.572 0.00467 N-Butyryl-L-

homoserine lactone

C8H13NO3 2.39 172.0968 0.089186

POS 3.0117 0.00435 L-Carnitine C7H15NO3 0.39 162.1123 1.041845

POS 4.0547 0.00140 Glycine C2H5NO2 2.45 76.0397 5.287456

POS 4.2532 0.00000 Neuraminic acid C9H17NO8 0.39 268.1006 7.883747

POS 6.8278 0.00337 Retinoyl

glucuronide

C26H36O8 5.48 477.2527 9.227536

POS 7.083 0.00276 Glutathione C10H17N3O6S 3.16 308.091 0.198757

POS 7.6032 0.00970 D-Alanine/L-

Alanine

C3H7NO2 0.69 90.0557 8.396408

POS 5.78 0.00861 Aldosterone C21H28O5 5.72 361.2036 7.455842

Urine metabolites whose excretion significantly differed (p<0.05) in seizing animals compared to

non-seizing animals. These metabolites were validated by an additional independent experiment.

on June 23, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from