Fossil genes and microbes in the oldest ice on earth

Abstract

Although the vast majority of ice that formed on the Antarctic continent over the past 34 million years has been lost to the oceans, pockets of ancient ice persist in the Dry Valleys of the Transantarctic Mountains. Here we report on the potential metabolic activity of microbes and the state of community DNA in ice derived from Mullins and upper Beacon Valleys. The minimum age of the former is 100 ka, whereas that of the latter is approximately 8 Ma, making it the oldest known ice on Earth. In both samples, radiolabeled substrates were incorporated into macromolecules, and microbes grew in nutrient-enriched meltwaters, but metabolic activity and cell viability were critically compromised with age. Although a 16S rDNA-based community reconstruction suggested relatively low bacterial sequence diversity in both ice samples, metagenomic analyses of community DNA revealed many diverse orthologs to extant metabolic genes. Analyses of five ice samples, spanning the last 8 million years in this region, demonstrated an exponential decline in the average community DNA size with a half-life of approximately 1.1 million years, thereby constraining the geological preservation of microbes in icy environments and the possible exchange of genetic material to the oceans.

Fig. 1.

Geologic setting of buried ice from Mullins and Beacon Valleys and evidence of encased microbes. (A) Digital elevation model showing the distribution of ice samples collected at various locations in Mullins (white box) and Beacon Valleys. Ice between locations marked one and seven spans an ≈8 million-year transect (1, DLE-98-12, ≈100 ka; 2, DLE-98-11, ≈200 ka; 3, DLE-98-CS-1, ≈300 ka; 4, MCI-04-013, ≈2 Ma; 5, EME-98-08, ≈5–6 Ma; 6, EME-98-01, ≈6–7 Ma; 7, EME-98-03, ≈8 Ma). (Inset Left) Location of the Dry Valleys in relation to the Antarctic continent; (Inset Right) Overview of Beacon Valley and nearby Taylor Glacier. (B) Photograph of Beacon Valley with view to the northeast toward Taylor Glacier. (C and D) Scanning electron micrographs and (E and F) epifluorescence micrographs of ice samples from DLE-98-12 (C and E) and EME-98-03 (D and F), illustrating DNA-containing bacteria cells and their morphology compared with glacial till. DLE-98-12 possessed cocci-like cells scattered among mineral debris (arrows), whereby EME-98-03 had filamentous sheath-like structures.

Fig. 2.

PCR amplification of encased microbial community DNA. Lane 1, environmental DNA from DLE-98-12 ice meltwater; lane 2, Bacteria-specific 16S rDNA PCR amplicons for DLE-98-12 (2a, raw; 2b, gel purified) and E. coli (2c, +control); lane 3, Archaea-specific 16S rDNA PCR amplicons for DLE-98-12 DNA (3a) and Haloferax volcanii (3b, +control); lanes 4 and 6 correspond to PCR reagent negative controls. Lane 5, Bacteria-specific 16S rDNA PCR amplicons for EME-98-03 (5a) and E. coli (5b, +control). Lane 7, Bacteria-specific 16S rDNA PCR amplicons for DLE-98-12 with DNA template serially diluted ×1 (7a), ×5 (7b), ×25 (7c), and ×125 (7d). Lanes 8a–8d, same as lane 7 but for the MilliQ water contamination control. Ice DNA template was routinely diluted (>×10) for PCR amplification. M1 and M2, molecular weight ladders with 0.1- and 1-kb increments, respectively.

Fig. 3.

Assessment of the metabolic activity and viability of microorganisms encased in DLE-98-12 and EME-98-03. (A and B) Incorporation and respiration of ³H-thymidine, ³H-leucine, and ¹⁴C-glucose in ice-meltwater samples during incubation at 4°C. Data are corrected for formalin-killed controls. Note different scales for ³H-leucine incubation in A and ¹⁴C-respiration in A and B. Error bars represent standard deviation among replicate incubations. (C and D) Time course of viable cell growth in nutrient supplemented ice-water cultures. Note the differences in scales, reflecting dramatic growth discrepancies between DLE-98-12 and EME-98-03. Symbols refer to the ice (DLE or EME) and nutrient type (T1–T4). Arrows indicate time points when subsamples were removed and used for PCR amplification of Bacteria-specific 16S rDNA and plating onto solid media (E). Amplification and colony formation were successful only for DLE cultures with recovered phylotypes given in SI Table 4. (F) Nested PCR amplification verified the presence of cultured isolates in the original ice microbial community. DLE011i/A. roseus CMS90 and Arthrobacter sp. primer sets produced their expected ≈770- and ≈440-bp fragment sizes, respectively, using both community Bacterial 16S rDNA amplicons (lanes 1a and 2a) and DLE011i genomic DNA (lanes 1b and 2b, +control) as template. E. coli genomic DNA (lanes 3a and 3b) was used as a negative control for both primer sets.

Fig. 4.

Time-dependent DNA degradation curve for ice samples collected in Beacon and Mullins Valleys. The weighted mean DNA size was calculated from densitometry profiles of total extracted community DNA after gel electrophoresis. The weighted mean distance of the DNA peak was converted to size by using a distance-size calibration from molecular weight standards. Representative images of community DNA from DLE-98-12 and EME-98-03 are provided above their respective data points, illustrating the dramatic differences in DNA size. Error bars represent standard deviation of DNA size and ice age. Curve fit is based on an exponential regression (y = 16804e^−0.0007x, r² = 0.96).

Fig. 5.

The genetic content of community ice DNA from DLE-98-12 and EME-98-03. (A) The distribution of gene categories for a random sampling of BAC (DLE-98-12; n = 132 sequences) and end-repaired (EME-98-03; n = 427 sequences) clone libraries, as determined by BLASTx analysis. A majority (≈55–60%) of the sequences from both ice sources had BLASTx matches to a wide variety of metabolic genes (key), whereas others (≈40–45%) had no statistical similarity to sequences in GenBank below the 10⁻⁵ e-value cutoff. The contribution (%) for the top five categories is given. A complete representation for all categories is provided in SI Table 5. (B) An analysis of the impact of sequence query size on the statistical similarity to sequences in GenBank (e-score). The log (e-score) per unit query length as a function of query length was plotted for each recovered ice sequence. Sequences with no BLASTx hit are distributed along the x axis.