Microbial Discovery Forge

Adam Arkin

Research Question

Does 16S community similarity across the 9 ENIGMA SSO wells (3x3 grid, ~4 m span at Oak Ridge) recapitulate the spatial arrangement in X, Y, and Z? Where it deviates, can hydrogeological connectivity or environmental gradients explain the pattern? Can we infer functional differences from taxonomy and what they imply about subsurface environmental parameters?

Research Plan

Hypothesis

H0a: Community similarity across SSO wells is spatially random — geographic distance does not predict Bray-Curtis dissimilarity.
H1a: Community dissimilarity increases with geographic distance (distance-decay), but specific well pairs deviate from this trend in ways consistent with subsurface hydrology (e.g., shared flow paths) or environmental discontinuities (e.g., lithological boundaries).
H0b: Inferred functional profiles (from taxonomy → pangenome mapping) show no spatial gradient across the SSO grid.
H1b: Functional composition varies systematically across the grid, with gradients in anaerobic metabolism, sulfur cycling, or metal response that imply spatial variation in redox, sulfate, or metal concentrations.
H0c: Depth (vertical zonation through VZ → VSZ → SZ1 → SZ2) has no effect on community composition independent of horizontal position.
H1c: Vertical stratification (hydrogeological zone) structures communities more strongly than horizontal distance at this spatial scale, reflecting the steep redox and moisture gradients with depth.

Revision History

v1 (2026-04-03): Initial plan — restart from corrupted start; builds on existing extracted data + prior project lessons
v1.1 (2026-04-03): Added nearby well geochemistry (EU/ED from 100WS/27WS) as regional context; added SSO isolate genomes (18 from M6-C2); confirmed SSO geochemistry values not in BERDL despite 221 registered sample tubes
v2 (2026-04-04): Added NB07 (guild interactions) and NB08 (temporal stability) to analysis plan. Renamed NB06 to reflect contamination plume model framing. Updated NB04 description to reflect multi-resolution (class + genus) approach with 65 annotated genera. All 8 notebooks now documented.

Overview

The ENIGMA Subsurface Science Observatory (SSO) at Oak Ridge Reservation Area 3 consists of 9 boreholes in a 3×3 grid (Upper: U1-U3, Middle: M4-M6, Lower: L7-L9) spanning ~6 meters. The SSO sits downhill and southwest of a contamination source delivering high nitrate, low pH, and heavy metals. Each borehole was cored through the vadose zone into saturated rock, yielding depth-resolved 16S amplicon profiles of sediment-associated and groundwater microbial communities.

We find that community similarity tracks the contamination plume rather than hillslope topography. A diagonal corridor of wells (U3-M6-L7) shares community composition along the inferred plume flow path from NE to SW. Depth dominates community structure (PERMANOVA R²=27.5%, p=0.0001), consistent with the plume traveling through the saturated zone. Genus-level functional inference maps the thermodynamic redox ladder (denitrification → iron reduction → fermentation) onto the physical grid, with Rhodanobacter denitrification peaking at M5 — the plume mixing zone. Groundwater carries a distinct plume-associated planktonic community enriched in denitrifiers and iron oxidizers.

Data Collections

enigma_coral — ENIGMA CORAL SSO 16S ASV data, sample metadata, well coordinates

Data Sources

Sediment 16S ASV: 37 samples across all 9 wells, depths 1.7-9.3 m (Feb-Mar 2023)
Groundwater 16S ASV: 40 samples across 5 wells, 2 depths (Sep 2024)
Pump test ASV: 14 communities from 3 wells (Mar 2024) — available but not yet extracted
Well metadata: Lat/lon, depth, lithological zone, collection date

Key Constraints

No geochemistry in BERDL (samples registered, measurements not loaded)
Species-level taxonomy unavailable (genus best at 35-63%)
Groundwater covers only 5/9 wells

Key Findings

1. Community Similarity Tracks Spatial Arrangement at Meter Scale

Bray-Curtis dissimilarity heatmap across the 9 SSO wells

Sediment-associated microbial communities across the 9 SSO wells (3×3 grid, ~6 m span) show significant distance-decay of similarity (Mantel test: Spearman ρ = 0.323, p = 0.029, 9,999 permutations). Mean Bray-Curtis dissimilarity = 0.747, ranging from 0.558 (U3–M6, most similar) to 0.872 (U3–L9, most dissimilar). NMDS ordination achieves low stress (0.067), and Procrustes analysis shows marginal correspondence between community ordination and physical grid (m² = 0.379, p = 0.080).

Critically, community turnover is not aligned with the hillslope (uphill-downhill Mantel ρ = −0.049, p = 0.580) but instead with the east-west axis (column Mantel ρ = 0.227, p = 0.092). This contradicts the naive expectation of topography-driven hydrology and points to a lateral structuring force.

(Notebook: 02_sediment_spatial.ipynb)

2. The Column 3 Corridor: A Plume Flow Path

NMDS ordination vs physical grid showing nearest community neighbor

Three wells — U3, M6, and L7 — are far more similar to each other than their geographic distances predict:
- U3–M6: BC = 0.558, residual = −0.170 (most similar pair in the grid)
- M6–L7: BC = 0.615, residual = −0.154
- U3–L7: BC = 0.646, residual = −0.133

These wells trace a diagonal corridor from the northeast (U3) to the southwest (L7) that aligns with the expected flow path of the contamination plume from the Area 3 source (high nitrate, low pH, heavy metals) located uphill and northeast of the SSO. The corridor's shared community composition reflects shared plume exposure. Conversely, the largest dissimilarity (U3–M4, BC = 0.871, residual = +0.105) spans from the plume entry point to the well farthest from the plume core.

Residual analysis and row-separation effects

(Notebook: 02_sediment_spatial.ipynb)

3. Depth Dominates Over Horizontal Position

Sample-level ordination colored by zone and well

PERMANOVA on sample-level communities (37 sediment core segments) reveals:
- Hydrogeological zone (VZ/VSZ/SZ1/SZ2) explains 27.5% of community variance (F = 4.05, p = 0.0001)
- Well identity explains 19.2% but is not significant (F = 0.80, p = 0.979)

Samples from the same well but different depths are nearly maximally dissimilar (median BC = 0.977), while samples from the same depth zone in different wells are substantially more similar (median BC = 0.835). This is consistent with the contamination plume traveling through the saturated zone: depth controls whether a sample intersects the plume, while horizontal position is secondary.

Phylum-depth correlations

Ten of 12 dominant phyla show significant depth associations (p < 0.05), splitting into biologically coherent groups:
- Shallow-enriched (vadose, oxic): Chloroflexi (ρ = −0.73), Patescibacteria (−0.70), Myxococcota (−0.54), Spirochaetota (−0.53)
- Deep-enriched (saturated, anoxic): Firmicutes (ρ = +0.76), WPS-2 (+0.52), Bacteroidota (+0.50), Proteobacteria (+0.49)

(Notebook: 03_depth_zonation.ipynb)

4. Genus-Level Biogeochemical Processes Map the Redox Ladder

Genus-inferred biogeochemical processes across the SSO grid

Multi-resolution functional inference (class-level: 22 classes, 78% coverage; genus-level: 65 annotated genera, 21% coverage) maps specific biogeochemical processes onto the 3×3 grid. The spatial distribution of processes recapitulates the thermodynamic redox ladder expected along a contamination plume:

Process	Hotspot	Key Genus	Inferred Environment
Denitrification	M5 (7.7%)	Rhodanobacter	Plume mixing zone: NO₃⁻ meets organic C
Iron oxidation	U3 (2.8%)	Sideroxydans	Plume entry: Fe²⁺ meets O₂
Nitrification	U3 (2.3%)	Ca. Nitrosotalea	Plume entry: NH₄⁺ meets O₂
Iron reduction	U1 (2.3%)	Anaeromyxobacter	Available Fe(III), suboxic
Sulfur oxidation	M4 (1.9%)	Arcobacter, Thiobacillus	Sulfide meets oxygen
Methanotrophy	M4 (1.9%)	Ca. Methanoperedens	Methane flux from deep zones
Fermentation	L9 (5.3%)	Spirochaeta, Paenisporosarcina	Terminal reduction, all electron acceptors depleted

M5 — the central well — hosts the highest denitrification potential (7.7% Rhodanobacter), consistent with its position at the plume mixing zone where nitrate-rich contaminated groundwater meets native organic carbon. Rhodanobacter is the hallmark denitrifier of the ORR contaminated subsurface (Green et al. 2012).

M6 is consistently the lowest for every oxidative process (iron oxidation, sulfur oxidation, nitrification) — the anaerobic dead zone of the grid, where the plume core has consumed all terminal electron acceptors.

Clustered heatmap of genus-level processes

(Notebook: 04_functional_inference.ipynb)

5. Groundwater Carries a Plume-Associated Planktonic Community

Groundwater and sediment communities at the same well are substantially different (median BC = 0.424). Comparison of plume-indicator genera reveals a distinct planktonic assemblage:

Genus	Function	Sediment %	GW %	Enrichment
Rhodanobacter	Denitrification	1.23	3.62	2.9× ▲GW
Gallionella	Iron oxidation	0.01	0.14	8.9× ▲GW
Sideroxydans	Iron oxidation	0.01	0.06	7.0× ▲GW
Geobacter	Iron reduction	0.00	0.01	5.5× ▲GW
Anaeromyxobacter	Iron/U reduction	1.24	0.03	0.02× ▼GW
Arcobacter	Sulfur oxidation	0.54	0.00	0× ▼GW
Ca. Methanoperedens	Methane oxidation	0.42	0.00	0× ▼GW

The groundwater is enriched in denitrifiers (Rhodanobacter) and iron oxidizers (Gallionella, Sideroxydans) — organisms that thrive in the flowing, metal-rich contaminated plume water. Sediment-attached anaerobes (Anaeromyxobacter, Arcobacter, Ca. Methanoperedens) are depleted in groundwater, indicating distinct attached vs planktonic communities rather than simple detachment.

Phylum composition: groundwater vs sediment

(Notebook: 05_gw_vs_sediment.ipynb)

6. Metabolic Guild Structure and Inferred Interactions

Metabolic guild composition across wells

Assigning 65 annotated genera to 11 metabolic guilds reveals distinct functional assemblages at each well. Guild co-occurrence analysis across the 9 wells identifies tightly coupled functional partnerships:

Nitrifier × iron oxidizer: ρ = +0.95 — these co-occur strongly, both concentrated at U3 (plume entry). Both are chemolithotrophs exploiting reduced compounds (NH₄⁺, Fe²⁺) arriving in the plume.
Syntroph × fermenter: ρ = +0.55 — classic anaerobic food web coupling. Fermenters produce organic acids and H₂; syntrophs consume these in obligate partnerships with sulfate reducers or methanogens.
Fermenter × predator (Bdellovibrio): ρ = +0.85 — predators track bacterial prey biomass, which is highest where fermentation supports dense populations.
Denitrifier × syntroph: ρ = −0.67 — mutual exclusion across the redox gradient. Denitrifiers thrive where nitrate is available (oxidizing); syntrophs require strictly anaerobic conditions.
Sulfate reducer × aerobic heterotroph: ρ = −0.75 — textbook redox separation. These guilds occupy opposite ends of the electron acceptor hierarchy.

Guild co-occurrence matrix

(Notebook: 07_hotspot_interactions.ipynb)

7. Groundwater Community Stability Over 9 Days

Temporal vs spatial variation

Groundwater communities at 5 wells sampled 9 days apart (Sep 9 vs Sep 18, 2024) show remarkable short-term stability:

Factor	R²	p	Interpretation
Well	49.9%	0.001	Spatial identity dominates
Filter size	10.1%	0.001	Free-living vs particle-associated are distinct
Depth (SZ)	2.5%	0.430	SZ1 vs SZ2 minor effect in GW
Date	0.8%	0.998	No detectable temporal change

The variation hierarchy — temporal (median BC = 0.351) < filter (0.750) < spatial (0.917) — confirms that spatial patterns are temporally stable at the 9-day scale. The Mantel correlation between date-1 and date-2 distance matrices is ρ = 0.867 (p = 0.001): which wells are most similar to each other is nearly identical across the two sampling dates.

This stability strengthens the plume model: the spatial community patterns we observe are not transient fluctuations but reflect persistent environmental structure maintained by the contamination plume.

Note on sediment temporal resolution: Sediment cores were collected once per well (Feb-Mar 2023). No within-well temporal replication exists for sediment. The 18-month offset between sediment (2023) and groundwater (2024) confounds material type with time, and cross-material temporal comparisons should be interpreted with this caveat.

Spatial pattern stability across dates

(Notebook: 08_temporal_stability.ipynb)

Results

Spatial Structure (NB02)

9 wells, 23,458 ASVs, 37 sediment core samples aggregated per well
Mean Bray-Curtis = 0.747; significant distance-decay (Mantel ρ = 0.323, p = 0.029)
East-west axis dominates (ρ = 0.227) over uphill-downhill (ρ = −0.049)
U3-M6-L7 corridor: 3 most negative residuals (more similar than expected)
U3-M4 and U3-L9: most positive residuals (more different than expected)

Distance-decay scatter plot with pair annotations

Procrustes superimposition of community ordination onto grid

Depth Zonation (NB03)

37 samples classified into VZ (33), VSZ (25), SZ1 (54), SZ2 (45) by depth + description
PERMANOVA: zone R² = 27.5% (p = 0.0001); well R² = 19.2% (p = 0.979, NS)
10/12 top phyla significant for depth (Spearman p < 0.05)

Well grid geometry with inter-well distances

Depth-zone profile across wells

Within-zone vs within-well dissimilarity

Phylum composition by hydrogeological zone

Zone enrichment patterns for top phyla

Functional Inference (NB04)

Class-level: 22 classes, 78% coverage; redox index range 0.047 (M6) to 0.227 (U3)
Genus-level: 65 genera annotated, 12 biogeochemical process categories
Denitrification range: 1.9–7.7% (M5 peak); fermentation: 2.0–5.3% (L9 peak)

All trait profiles mapped onto grid

Key functional gradients: redox, fermentation, nitrogen

Class-level trait clustermap

Groundwater vs Sediment (NB05)

5 wells with both materials; within-well BC = 0.364–0.450
GW enriched in Rhodanobacter (2.9×), Gallionella (8.9×), Sideroxydans (7.0×)

Interpretation

The Contamination Plume Model

Synthesis: contamination plume model overlaid on the SSO grid

All findings converge on a single explanatory model: the spatial structure of SSO microbial communities is governed by a contamination plume entering from the northeast and flowing southwest through the saturated zone.

The SSO sits downhill and southwest of a contamination source at Oak Ridge Reservation Area 3, which delivers high nitrate, low pH water laden with heavy metals (uranium, chromium, nickel). The plume enters the grid near U3 (upper-east) and flows diagonally toward L7 (lower-west), creating the Column 3 corridor identified in NB02.

This model explains every major observation:

Why east-west turnover exceeds uphill-downhill: The plume edge runs laterally across the grid, not along the hillslope. Communities track contamination, not surface topography.
Why U3-M6-L7 share community composition: These wells intercept the plume flow path. Shared geochemical exposure homogenizes their communities despite being in different rows.
Why depth dominates over well identity: The plume travels through the saturated zone (SZ1/SZ2). Depth controls whether a sample is within the plume (deep = affected) or above it (shallow = background).
Why M5 is the denitrification hotspot: M5 sits at the plume mixing zone where nitrate-rich plume water meets native organic carbon — the thermodynamic sweet spot for denitrification. Rhodanobacter denitrificans, the dominant denitrifier at ORR contaminated sites (Green et al. 2012), reaches 7.7% relative abundance here.
Why M6 is the anaerobic dead zone: M6 lies in the plume core downstream of M5, where denitrification has consumed the nitrate and further electron acceptors are depleted. Only fermentative metabolism persists.
Why the redox sequence maps onto the grid: The spatial distribution of biogeochemical processes (O₂ → NO₃⁻ → Fe(III) → SO₄²⁻ → fermentation) follows the thermodynamic electron acceptor sequence, running from the plume entry (oxidative processes at U3) through the mixing zone (denitrification at M5) to the plume terminus (fermentation at L9).
Why groundwater is enriched in plume taxa: The flowing water carries Rhodanobacter (denitrifier), Gallionella and Sideroxydans (iron oxidizers) — organisms adapted to the contaminated, metal-rich plume water — while sediment-attached anaerobes remain in biofilms.

Literature Context

The dominance of Rhodanobacter in contaminated ORR groundwater is well-established. Green et al. (2012) showed that Rhodanobacter species dominate bacterial communities in the most contaminated zones of the ORR subsurface, accounting for up to 45% of 16S sequences in low-pH, high-nitrate wells. Our finding of 7.7% Rhodanobacter at M5 is consistent with moderate plume influence — lower than the most contaminated Area 2 wells but clearly elevated above background.

Anaeromyxobacter dehalogenans, found throughout SSO sediments (1.3% mean), is a model ENIGMA organism for iron and uranium reduction in the ORR subsurface (Thomas et al. 2010). Its enrichment in sediment over groundwater (41× higher in sediment) is consistent with its known biofilm-forming, surface-attached lifestyle.

Recent work generating 77 sediment and 33 groundwater metagenomes from the ORR (MRA 2025) examines attached vs planktonic communities — our 16S-based findings provide complementary amplicon-level evidence supporting the same pattern of distinct sediment and groundwater assemblages.

The observation of significant community turnover at meter scale is noteworthy. Most subsurface distance-decay studies operate at kilometer scales (Fierer & Jackson 2006). The SSO demonstrates that contamination plumes can structure communities at sub-decameter resolution, consistent with the known sharp geochemical gradients at plume fringes.

The guild co-occurrence patterns (NB07) are consistent with established syntrophic theory. The tight coupling of syntrophs and fermenters (ρ = +0.55) reflects the thermodynamic interdependence described by McInerney et al. (2009): syntrophic fatty acid degradation requires fermentation products (H₂, acetate, formate) to be maintained at low concentrations, creating obligate partnerships. The mutual exclusion of denitrifiers and syntrophs (ρ = −0.67) maps onto the canonical redox zonation of contaminated aquifers (Chapelle 2001), where electron acceptor availability segregates metabolic guilds spatially.

The 9-day temporal stability of groundwater communities (NB08: well R² = 49.9%, date R² = 0.8%) is consistent with long-term monitoring at contaminated aquifers showing that "composition of native microbial communities varied temporally, yet remained distinctive from well to well" (Shi et al. 2021). The spatial signal overwhelms temporal noise at this timescale, supporting the interpretation that community structure reflects persistent geochemical conditions rather than stochastic assembly.

Novel Contribution

This analysis provides the first spatially explicit mapping of biogeochemical process distributions across the SSO 3×3 grid, inferred from 16S community composition at three taxonomic resolutions (phylum, class, genus). The key novelties are:

The plume flow path is visible in community similarity: The U3-M6-L7 corridor, identified purely from Bray-Curtis dissimilarity patterns, aligns with the expected NE→SW plume trajectory — demonstrating that 16S community similarity can map subsurface hydrology at meter scale.
The redox ladder is spatially resolved: Genus-level functional inference maps the classic thermodynamic electron acceptor sequence (O₂ → NO₃⁻ → Fe(III) → SO₄²⁻ → fermentation) onto the physical grid, with specific process hotspots at specific wells.
M5 as the mixing zone: The central well's denitrification peak (Rhodanobacter at 7.7%) identifies the precise location where plume nitrate meets background organic carbon — a prediction that can be validated when SSO geochemistry data is loaded into CORAL.
Depth × plume interaction: The vertical zonation (PERMANOVA R² = 27.5%) reflects the plume's confinement to the saturated zone, not just generic depth gradients. Above the water table, communities are unaffected by contamination.
Guild co-occurrence maps the metabolic network: The strong nitrifier-iron oxidizer coupling (ρ = +0.95) and denitrifier-syntroph exclusion (ρ = −0.67) provide a functional network topology that overlays the spatial redox gradient — connecting community composition to biogeochemical mechanism.
Temporal stability validates spatial inference: With date explaining only 0.8% of GW variance (vs 49.9% for well), the spatial patterns are not stochastic snapshots but persistent features of the subsurface environment, consistent with stable plume geochemistry.

Limitations

No direct geochemistry: SSO geochemistry (221 sample tubes registered in CORAL) has not been loaded into BERDL. All environmental inferences are from community composition alone. The contamination plume model generates testable predictions that await geochemical confirmation.
Temporal offset: Sediment cores (Feb-Mar 2023) and groundwater (Sep 2024) were collected 18 months apart. Seasonal or plume dynamics could confound the comparison.
Groundwater coverage: Only 5 of 9 wells have groundwater ASV data (L7, L9, M4, M6, U2), missing the critical wells M5 (denitrification hotspot) and U3 (plume entry). The denitrification and iron oxidation hotspot interpretations are based on sediment data; GW validation at these wells would substantially strengthen the plume model. Pump test ASV data (Brick 460-462, wells L8/M5/U2) remains available for future extraction.
Genus coverage uncertainty: Genus-level functional annotation covers only 21% of total reads (65 of 1,038 genera). Process abundance estimates are lower bounds; the unclassified 56% of reads could harbor additional functional capacity. Sensitivity to this coverage gap should be assessed by comparing genus-level patterns with the higher-coverage class-level traits (78%), which show consistent redox patterns.
Taxonomy resolution: Species-level classification is ~0%; genus-level covers only 44% of sediment reads. Functional inference relies on literature-based trait assignments rather than direct genomic evidence.
Single timepoint for sediment: Temporal dynamics of the plume and community response cannot be assessed from one core sampling event.
Trait dictionary subjectivity: Phylum/class-level trait scores are consensus estimates, not empirical measurements for these specific populations.

Future Directions

Load SSO geochemistry into CORAL: The 221 registered geochemistry samples (metals, IC/TOC, isotopes, NH₃/NO₂) would allow direct correlation of community composition with measured environmental parameters, validating or refuting the plume model.
Extract pump test ASV data (Brick 460-462): A third temporal snapshot (Mar 2024) from L8, M5, U2 would test the M5 denitrification hotspot prediction and add temporal resolution.
UniFrac analysis with ASV sequences: The actual 16S sequences (Bricks 457/460/477) could enable phylogenetic distance-based beta-diversity (weighted UniFrac), which may be more sensitive to plume effects than Bray-Curtis on ASV counts.
Metagenomics: Shotgun metagenomics at the same spatial resolution would confirm functional inferences that are currently based on taxonomy-to-trait mapping, and would capture the 56% of reads without genus-level classification.
Temporal monitoring: Repeat 16S profiling across seasons and plume dynamics would reveal whether community structure tracks plume fluctuations in real time.

Data

Sources

Collection	Tables Used	Purpose
`enigma_coral`	`sdt_sample`, `sdt_location`, `sdt_community`, `ddt_ndarray`, `ddt_brick0000457-459`, `ddt_brick0000477-479`	SSO 16S ASV data (sediment + groundwater), sample metadata, well coordinates

Generated Data

File	Rows	Description
`data/well_distances.csv`	9×9	Inter-well geographic distances (meters)
`data/bc_dissimilarity_sediment.csv`	9×9	Bray-Curtis dissimilarity between wells
`data/spatial_stats.csv`	36	All pairwise comparisons with residuals
`data/community_matrix_sediment_asv.csv`	9×23,458	Well-aggregated ASV abundance matrix
`data/community_matrix_sediment_phylum.csv`	9×85	Well-aggregated phylum abundance matrix
`data/sediment_sample_zones.csv`	159	Sample-to-hydrogeological-zone assignments
`data/permanova_results.csv`	2	PERMANOVA results for zone and well effects
`data/zone_indicators.csv`	12	Phylum zone-association statistics
`data/trait_profiles_class.csv`	9×9	Class-level trait profiles per well
`data/genus_function_grid.csv`	9×12	Genus-inferred process abundances per well
`data/trait_spatial_gradients.csv`	9	Spatial gradient tests for each trait
`data/hotspot_profiles.csv`	135	Top 15 genera per well with guild assignments
`data/guild_composition.csv`	9×11	Guild abundance per well

References

Green SJ, Prakash O, Jasrotia P, Overholt WA, Cardenas E, Huber D, Schadt CW, Dalton D, Kaber K, Brooks SC, Watson DB, Tiedje JM. (2012). "Denitrifying bacteria from the genus Rhodanobacter dominate bacterial communities in the highly contaminated subsurface of a nuclear legacy waste site." Applied and Environmental Microbiology 78(4):1039-47. PMID: 22179244
Prakash O, Green SJ, Jasrotia P, Overholt WA, Canez A, Palumbo AV, Tiedje JM, Kostka JE. (2012). "Rhodanobacter denitrificans sp. nov., isolated from nitrate-rich zones of a contaminated aquifer." International Journal of Systematic and Evolutionary Microbiology 62:2457-62. PMID: 22140175
Green SJ, Prakash O, Gihring TM, Akob DM, Jasrotia P, Jardine PM, Watson DB, Brown SD, Palumbo AV, Kostka JE. (2010). "Denitrifying bacteria isolated from terrestrial subsurface sediments exposed to mixed-waste contamination." Applied and Environmental Microbiology 76(10):3244-54. PMID: 20305024
Thomas SH, Wagner RD, Arakaki AK, Skolnick J, Kirby JR, Shimkets LJ, Sanford RA, Löffler FE. (2008). "The mosaic genome of Anaeromyxobacter dehalogenans strain 2CP-C suggests an aerobic common ancestor to the delta-proteobacteria." PLoS ONE 3(5):e2103. PMID: 18461131
Fierer N, Jackson RB. (2006). "The diversity and biogeography of soil bacterial communities." Proceedings of the National Academy of Sciences 103(3):626-31. PMID: 16407148
Watson DB, Wu WM, Mehlhorn T, Tang G, Earles J, Lowe K, Gihring TM, Zhang G, Phillips J, Boyanov MI, Spalding BP, Schadt C, Kemner KM, Criddle CS, Jardine PM, Brooks SC. (2013). "In situ bioremediation of uranium with emulsified vegetable oil as the electron donor." Environmental Science & Technology 47(12):6440-48. PMID: 23631611
McInerney MJ, Sieber JR, Gunsalus RP. (2009). "Syntrophy in anaerobic global carbon cycles." Current Opinion in Biotechnology 20(6):623-32. PMID: 19897353
Chapelle FH. (2001). "Ground-Water Microbiology and Geochemistry." 2nd Edition. John Wiley & Sons.
Shi Z, Xu M, Gong H, Ye Q, Yan C, et al. (2021). "Long-term dynamic changes in attached and planktonic microbial communities in a contaminated aquifer." Environmental Pollution 286:117219. PMID: 34052673
Arkin AP, Cottingham RW, Henry CS, et al. (2018). "KBase: The United States Department of Energy Systems Biology Knowledgebase." Nature Biotechnology 36:566-569. PMID: 29979655

Testable Predictions

SSO geochemistry (when loaded into CORAL): NO₃⁻ concentration, pH, and metal levels should show a NE→SW gradient, with highest contamination at U3/M6 and lowest at M4/U1.
Nearby EU/ED well metals (100WS/27WS bricks, 90–120 m NE of SSO): Should confirm that the contamination plume approaches from the northeast, with metal concentrations decreasing toward the SSO.
Pump test groundwater (Brick 460-462, wells L8/M5/U2): Rhodanobacter should be highest at M5 (mixing zone) and lower at L8 and U2.
SSO isolate genomes (18 genomes from M6-C2): Should encode anaerobic metabolisms (fermentation, sulfate reduction) consistent with M6's position in the plume core.

Discoveries

16S community similarity maps contamination plume at meter scale

April 2026

The SSO 3×3 well grid (~6 m span) shows significant distance-decay (Mantel ρ=0.323, p=0.029) driven by the east-west axis rather than the hillslope gradient. A diagonal corridor of wells (U3-M6-L7) shares community composition along the NE→SW plume flow path. Genus-level functional inference maps th

SSO geochemistry not in BERDL despite 221 registered samples

April 2026

221 METALS/ICTOC/ISOTOPES/NH3NO2 sample tubes from the SSO Subsurface Observatory campaign are registered in ENIGMA CORAL (sdt_sample) but zero Assay Geochemistry processes are linked — the analytical measurement values were never ingested. This is the key validation dataset for the plume model.

Nitrifier × iron oxidizer coupling (ρ=+0.95) at plume entry

April 2026

Candidatus Nitrosotalea (archaeal ammonia oxidizer) and Sideroxydans (iron oxidizer) co-occur almost perfectly across the 9 SSO wells (ρ=+0.95), concentrated at U3. Both are chemolithotrophs exploiting reduced compounds arriving in the contamination plume. This tight coupling suggests shared env

Data Collections

🧬

Pangenome Collection

kbase_ke_pangenome

KBase, DOE

🪰

ENIGMA CORAL

enigma_coral

ENIGMA SFA, LBNL

Review

Summary

This project presents an exceptional analysis of microbial community structure across the ENIGMA Subsurface Science Observatory (SSO) at Oak Ridge. The work successfully addresses a clear research question with hypothesis-driven methodology, producing compelling evidence that contamination plume flow paths—rather than surface topography—govern subsurface community structure at meter scale. The contamination plume model elegantly explains all major observations: the U3-M6-L7 diagonal corridor of similar communities, the dominance of depth over horizontal position (PERMANOVA R²=27.5%), and the spatial mapping of redox processes onto the physical grid. The analysis is comprehensive, well-documented, and generates actionable predictions for validation when SSO geochemistry data becomes available. This represents high-quality subsurface microbiology research that advances our understanding of contaminated aquifer ecology.

Methodology

The research design is exemplary with clearly stated, testable hypotheses comparing spatial random null models against distance-decay and hydrogeological connectivity alternatives. The multi-resolution approach—combining well-aggregated spatial analysis with sample-level depth analysis—is methodologically sound and addresses different scales of environmental variation appropriately. Data sources are clearly identified and limitations openly acknowledged (no direct geochemistry, temporal offsets, incomplete groundwater coverage). The statistical approach is rigorous, employing appropriate methods (Mantel tests, PERMANOVA, NMDS ordination, Procrustes analysis) with proper multiple testing corrections. The functional inference strategy acknowledges genus-level limitations and employs conservative phylum/class-level trait mapping alongside exploratory pangenome approaches. The analysis pipeline is reproducible with clear dependencies and a logical progression from data integration through synthesis.

Code Quality

The SQL approaches are appropriate for ENIGMA CORAL data extraction, and the analysis properly handles string-typed numeric columns (a known BERDL pitfall) through explicit casting. Statistical methods are correctly implemented with proper handling of distance matrices, community dissimilarity calculations, and ordination techniques. The notebooks are well-organized with clear markdown documentation, logical flow from data integration through synthesis, and appropriate separation of Spark-dependent and local analysis components. The functional inference approach appropriately acknowledges the limitations of genus-level taxonomy resolution (21% coverage) and employs multiple complementary strategies. Known pitfalls from docs/pitfalls.md are properly addressed, including Spark DataFrame handling and DECIMAL column casting.

Findings Assessment

The conclusions are strongly supported by converging evidence across multiple analytical approaches. The contamination plume model provides a parsimonious explanation for community patterns, redox gradients, and biogeochemical process distributions that would be difficult to achieve through data fabrication or over-interpretation. The spatial correlation patterns (U3-M6-L7 corridor, east-west dominance over uphill-downhill) align with known Oak Ridge hydrogeology and contamination source locations. The taxonomic findings are consistent with established literature (Rhodanobacter as an ORR contamination indicator, Anaeromyxobacter in reducing zones), providing external validation. Limitations are comprehensively acknowledged, including the lack of direct geochemistry validation, temporal offsets between sediment and groundwater sampling, and the constraints of genus-level functional inference. The analysis avoids over-interpretation while extracting maximum insight from available data.

Suggestions

Validate pump test data extraction: The research plan mentions available pump test ASV data (Brick 0000460-462) for wells L8/M5/U2 that was not extracted. Given that M5 is identified as the critical denitrification hotspot, extracting and analyzing this data could provide crucial validation of the plume mixing zone model.
Strengthen temporal analysis robustness: The 18-month offset between sediment (Feb-Mar 2023) and groundwater (Sep 2024) sampling potentially confounds material type with temporal variation. While the 9-day groundwater stability analysis is excellent, adding sensitivity analysis for the sediment-groundwater comparisons would strengthen interpretations.
Quantify functional inference uncertainty: The 21% genus-level coverage for functional annotation represents a significant limitation. Consider adding bootstrapping or sensitivity analysis to assess how robust the spatial functional gradient patterns are to the unmapped 79% of reads.
Expand geochemical context: While SSO-specific geochemistry is unavailable, the nearby EU/ED well metals data (90-120m northeast) could provide stronger regional gradient context. The current analysis mentions this data but appears incomplete in the implementation.
Consider phylogenetic approaches: The analysis relies on taxonomy-based (Bray-Curtis) dissimilarity. UniFrac analysis using the available 16S sequences could reveal phylogenetically-informed patterns that might be more sensitive to environmental gradients than abundance-based measures.
Document potential confounders: While core disturbance is mentioned as a potential confounder for shallow samples, this could be explored more systematically by examining depth-abundance patterns for potential surface contamination signatures.