.png)
The short version
When we sequence the DNA in a soil sample, we read the community that is alive right now together with a faint echo of what was alive and active in the recent past. Picture a living city that also happens to contain a graveyard. Almost everything on the map is the bustle of the living, the graveyard is a small corner of it, and most of what is buried there is recent rather than centuries old.
That echo of recently living cells is not a flaw in the measurement. Because it reflects the preceding days and weeks rather than only the day you sampled, it averages out the noise of exactly when the corer went in, which is what you want when you are comparing trials or benchmarking sites. And for bacteria and fungi in particular, sequencing is the practical route to seeing the community at all.
The rest of this piece explains what that buried fraction really is, how much of it there is, how long it lasts, and why we think it makes our trial comparisons more robust rather than less.
A fair question, answered properly
It’s a fair question, and we get it often enough that it deserves a proper answer rather than a reassuring shrug. When we run 16S, ITS, or 18S/trnL metabarcoding on a soil sample, we are not sorting living cells from dead ones before we sequence. We amplify and sequence the DNA that’s in the extraction, and some fraction of that DNA comes from cells that are no longer alive, or no longer intact.
That’s true, and it’s worth explaining properly, because the honest answer is more useful than the reassuring one: relic DNA isn’t just a contaminant to apologize for. Once you understand what it is and how it behaves, it turns out to carry information that a living-cells-only snapshot would miss.
Soil microbes live fast and die constantly
This is the living-city part of the picture. A gram of soil holds billions of microbial cells, and that population is never static. Under favorable conditions, many soil bacteria can double in a matter of hours to a couple of days. But growth is only half the picture. Death is the other, and it’s just as constant. Cells die from predation (protozoa, nematodes, and other grazers), viral lysis, desiccation and rewetting cycles, freeze-thaw, nutrient exhaustion, pH shifts, and simple old age. In an actively cycling soil community, growth and death are running in parallel all the time, not as occasional events but as the steady-state condition of the system.
When a cell dies and lyses, its DNA doesn’t vanish. Some is rapidly degraded by nucleases and consumed as a nutrient source by other microbes. But a portion escapes into the soil matrix as extracellular DNA (eDNA), also called relic DNA, where it can adsorb onto clay minerals and humic substances. That adsorption is protective: bound DNA resists enzymatic degradation far better than DNA in solution. This is why relic DNA is a regular, expected feature of soil metabarcoding data. It isn’t sequencing error or sloppy extraction, it’s a real feature of how soil biogeochemistry works.
How much relic DNA are we actually talking about, and how long does it last?
This is where we want to be precise rather than hand-wavy, because “relic DNA” gets thrown around as if it were a single fixed quantity with a single decay curve. It isn’t. To stay with the graveyard picture: most of it is recent and clears within days to weeks, and only a smaller, mineral-protected fraction lingers for longer.
The foundational estimate here comes from Carini et al. (2016, Nature Microbiology), who used a viability-selective PCR approach across a wide range of soils and found that, on average, roughly 40% of both prokaryotic and fungal DNA recovered from soil was extracellular or from non-intact cells, with substantial variation between soils, and the proportion tracking soil chemistry (notably, lower in soils with higher exchangeable base cations). At the high end, some individual samples showed relic DNA fractions well above that average.
On persistence, the honest picture is that there is no single “half-life”, because decay is biphasic. Studies using synthetic or labeled eDNA in soil microcosms (Morrissey et al., 2015; Sirois & Buckley, 2019) report first-order degradation constants on the order of 0.05 to 0.16 per day for the more labile fraction, which corresponds to a matter of days to a few weeks before that fraction becomes undetectable by sequencing, even while a portion remains detectable by more sensitive qPCR for longer. But decomposition curves are consistently better described by an asymptotic exponential model, not a clean single-rate decay: a mineral-stabilized fraction persists far longer than the labile pool, with independent lines of evidence (Gebhard & Smalla, 2006; radiocarbon dating of soil exDNA by Agnelli et al., 2007, albeit with caveats the authors themselves flag) pointing to persistence on the order of months to years for that protected fraction, and rare, soil-specific outliers extending much further. Worth being clear on the scale of it: that long-lived fraction sits inside the roughly 40% relic pool rather than on top of it, and the fast-turnover pool accounts for most of that, so it is a minority of a minority.
So the accurate statement is: soil relic DNA is a mixture of a fast-turnover pool (days to weeks) and a slow, mineral-protected pool (months to years), and the two are hard to fully separate with standard extraction. We’re not going to pretend we can hand you one clean number, because the literature doesn’t support one, and overstating precision here would be exactly the kind of convenient overclaim we’d rather avoid.
What this means for interpreting a single snapshot, and why it can help
If you take one soil sample and sequence total DNA, you’re looking at a composite: the community that’s alive and active right now, layered with an echo of what was alive and active over roughly the preceding days to weeks (and, at lower resolution, longer). Carini and colleagues showed that this composite can inflate apparent richness by up to 55% relative to a viability-filtered signal, and it can shift the estimated relative abundance of specific taxa. That is exactly the caveat we want you to have in view when you’re reading any single-sample diversity metric, from us or from anyone else.
But soil microbiomes are also genuinely volatile on short timescales. Moisture pulses, tillage, a fertilizer or biostimulant application, or a heat spell can all shift the active community meaningfully within days. A measurement of living cells only, taken at one point in time, captures whatever happened to be true on that day, which makes it more sensitive to sampling-time noise than people often assume. The relic DNA fraction, because it integrates over the preceding stretch of time, has a smoothing effect. It’s a partial record of what has recently been present, not only what happens to be present on the day the corer went into the ground.
That has two concrete, decision-relevant advantages:
Reduced sensitivity to sampling-time noise. A benchmark or trial comparison built on total DNA is less likely to be skewed by a transient bloom or die-off that happened to coincide with sampling, because the recent history is folded into the signal rather than being invisible outside that day.
A short-memory record, not just a snapshot. Total DNA metabarcoding can carry a faint but real signature of taxa that were present and active in the recent past, useful for detecting the aftermath of a management event or disturbance even if the acutely responding population has already died back by the time you sample.
Neither of these is a claim that relic DNA tells you precisely when something died, or lets you reconstruct a clean timeline. The decay kinetics aren’t uniform or fast enough across taxa and conditions to support that kind of precision, and we won’t claim otherwise. It’s a directional, qualitative advantage: more temporal context in the aggregate signal, not a timestamped ledger.
Where this leaves the comparison and the decision
Viability-selective methods exist: PMA (propidium monoazide) treatment, and DNase or Benzonase digestion of extracellular DNA before extraction. They do produce a signal closer to “living cells only.” They also add cost, extraction complexity, and their own biases (efficiency varies by soil type and mineralogy), and they discard the temporal-integration property described above. Which approach is right depends on the question being asked. If the deliverable is a precise, instant-in-time viable-biomass estimate, viability selection is the more defensible tool. If the deliverable is a benchmark comparison, a trial effect, or a placement recommendation meant to be robust to the exact day of sampling, then standard total-DNA metabarcoding, read with appropriate caveats about compositional and relic-DNA effects, is doing useful rather than misleading work.
The measurement is total DNA. The evidence it produces is a composite of present activity and recent history. The interpretation we build on top of it accounts for that composite, rather than pretending it isn’t there. That distinction, between what was measured, what can be derived from it, and what we’re willing to conclude, is the standard we hold every output to, and it’s the same standard we’d want applied to this question.
So when someone says we’re reading a graveyard, the honest answer is yes. A small, mostly recent one. And even that fraction is worth knowing, because it is part of the system we are measuring, not noise laid on top of it.
References: Carini, P., Marsden, P.J., Leff, J.W., Morgan, E.E., Strickland, M.S. & Fierer, N. (2016). Relic DNA is abundant in soil and obscures estimates of soil microbial diversity. Nature Microbiology 2, 16242. Morrissey, E.M. et al. (2015). Dynamics of extracellular DNA decomposition and bacterial community composition in soil. Soil Biology and Biochemistry 86, 42-49. Sirois, S.H. & Buckley, D.H. (2019). Factors governing extracellular DNA degradation dynamics in soil. Environmental Microbiology Reports.
.svg)
.svg)
