Cutting Antibiotic Resistance Genes in Poultry Litter
Walk into a duck house on a dry-litter system and the floor looks like nothing much. Bedding, feathers, droppings, feed dust. It gets raked, topped up and mostly ignored until the next flock goes in. It is also one of the biggest reservoirs of resistance biology on the whole operation.
A 2026 study that sequenced the virome of duck litter from four commercial farms has now put numbers on how much of that load you can strip out. Antibiotic resistance genes in poultry litter fell in a clean, repeatable order across four management groups, and the group that came out best is not the one most people would put money on.
Why litter ends up on the resistance map
In a scaled dry-litter system, birds live on the same bedding for the whole cycle. Everything they shed goes into it. Bacteria, viruses, resistance genes, and the mobile elements that shuttle those genes between organisms. Litter contamination drives the frequent disease outbreaks this system is known for, which is the reason anyone manages it at all.
What has been missing is a way to tell whether a litter treatment does anything about the resistance reservoir, rather than just making the heap smell better and spread more easily. Metagenomics gives you the measurement, and metaviromics, sequencing the viral fraction on its own, adds a layer most litter work skips. Viruses matter because bacteriophages ferry resistance genes between bacteria, and how many of them are around sets an upper limit on how much gene movement the system can support.
How the study was built
Four commercial duck farms, five samples per group, twenty in total. The four groups were set up as a comparison ladder:
| Group | Management | Fermented | Antibiotic history |
|---|---|---|---|
| DL-D | Deep-layer litter | Yes, 15 to 18 days | No prophylactic antibiotics |
| DL-S-NAF | Shallow-layer litter | Yes, 15 to 18 days | No prophylactic antibiotics |
| DL-S | Shallow-layer litter | No | No prophylactic antibiotics |
| DL-S-ANF | Shallow-layer litter | No | From an antibiotic-using farm |
Piles went under a functional intelligent membrane and fermented for 15 to 18 days, temperature logged daily at three depths. That detail matters more than it sounds. Composting only knocks back pathogens and gene load if the heap actually reaches temperature and holds it. Piling it up and walking away is a different process, and this study measured the managed version.
The sequencing was deep. About 1.01 x 10^9 raw paired-end reads off the twenty DNA virome samples, roughly 151.5 Gb. Cleaning left about 9.76 x 10^8 reads, around 145.6 Gb, with Q20 above 97.75 percent and Q30 above 93.88 percent. From there about 4.31 x 10^8 non-host reads assembled into 5,170,738 contigs of 500 bp or longer. Viral mining pulled 57,032 candidate sequences, and 23,130 survived quality filtering: 80 complete, 78 high quality, 208 medium, 22,764 low. Those clustered into 18,321 vOTUs totalling 95.33 Mb, N50 of 5,970 bp, GC content 52.22 percent.
What the viral community looked like
A small number of groups ran the show. On the DNA side, Uroviricota took 95.05, 94.61, 94.69 and 94.38 percent across DL-D, DL-S-NAF, DL-S and DL-S-ANF, with Nucleocytoviricota at 4.44, 4.89, 4.93 and 4.92 percent. At class level, Caudoviricetes and Megaviricetes tracked those numbers almost exactly.
On the RNA side, Pisuviricota ran 71.14, 85.27, 68.79 and 71.53 percent, and Lenarviricota 25.04, 13.75, 28.07 and 22.19 percent.
The groups separated cleanly. For DNA viruses, PCoA axes PC1 and PC2 covered 63.32 percent of variation, split 40.47 and 22.85. For RNA viruses the same two axes explained 55.39 percent, split 34.17 and 21.22.
Fermentation cut RNA viral abundance significantly, with DNA viruses trending down alongside. The authors concluded that RNA viruses, Pisuviricota especially, work as sensitive bioindicators for ecological health assessment in this material. That is a genuinely practical result. If you want an early signal that your litter management is doing something, RNA viral composition moves sooner than DNA viral composition.
One representative sample from the DL-D group went deeper. It gave 34 vOTUs and 343 viral genes, of which 176 (51.3 percent) annotated against NCBI NR, 107 (31.2 percent) against KEGG, 121 (35.3 percent) against Swiss-Prot, 40 (11.7 percent) against COG and 18 (5.3 percent) against GO. Two hundred and ten genes, 61.2 percent, got functional annotation in at least one database. KEGG abundance by RPKM: 129,783 for genetic information processing, 99,441 for translation, 63,610 for metabolism, 17,785 for cofactor and vitamin metabolism, 17,449 for carbohydrate metabolism, 11,629 for energy metabolism.
The resistance gene results
This is the part that should change how people work.
| Gene class | Genes reduced by fermentation | Direction of abundance |
|---|---|---|
| Antibiotic resistance (ARG) | sul1, sul2, tetB(P), qacEdelta1 | DL-D < DL-S-NAF < DL-S < DL-S-ANF |
| Metal resistance (MRG) | arsC, arsM, merA, copR, corR | DL-D < DL-S-NAF < DL-S-ANF < DL-S |
| Biocide resistance (BRG) | mdeA, actP, smdB, cpxA, galE | DL-D < DL-S-NAF < DL-S about equal to DL-S-ANF |
| Virulence factors (VF) | groEL2, lirB | DL-D < DL-S-NAF < DL-S < DL-S-ANF |
Start with the ARG row. Fermented litter sat lowest, unfermented litter from an antibiotic-using farm sat highest, and both fermented groups came in below both unfermented ones. By class, aminoglycoside, florfenicol, multidrug, MLS, tetracycline and sulfonamide genes were all significantly lower in DL-D, DL-S-NAF and DL-S than in DL-S-ANF at P < 0.05.
Individual genes told the same story. In DL-D, arnA, Bado_rpoB_RIF, sul1, tetB(P), sul2 and qacEdelta1 were lower than in DL-S and DL-S-ANF at P < 0.05, and Bado_rpoB_RIF and tetB(P) came in below DL-S-NAF too at P < 0.01. In DL-S-NAF, sul1, tetB(P), sul2 and qacEdelta1 were lower than in both DL-S and DL-S-ANF at P < 0.01. At the sharp end, floR, aadA, qacEdelta1, tet(X) and cfr(A) were lower across DL-D, DL-S-NAF and DL-S than in DL-S-ANF at P < 0.001.
Note tetB(P) and the tetracycline class result sitting in there. That is the molecular machinery behind tetracycline failure, living in bedding, and fermentation knocked it down.
Two findings that break the obvious story
First: simply dropping antibiotics was not enough. DL-S came from a farm using no prophylactic antibiotics and no fermentation, and it still ranked above both fermented groups for ARG abundance. Taking a drug out removes the selection pressure going forward. It does not clear the reservoir already in the bedding, and it does nothing to stop the genes sitting there from moving.
Second, and stranger: metal resistance genes did not follow the antibiotic story at all. The order came out DL-D < DL-S-NAF < DL-S-ANF < DL-S, which puts antibiotic-free unfermented litter in the worst position, above litter from an antibiotic-using farm. Metal and biocide resistance genes answer to things that have nothing to do with your drug cabinet. Trace minerals in the feed, disinfectants, copper, zinc. A flock raised without antibiotics can still carry a heavy metal resistance load if nobody looked at the rest of the input stack. Fermentation pulled that number down. Withdrawal did not.
What to do with this on a working operation
Treat litter as a managed input rather than waste. If the current process is scrape, stockpile and spread, that is an unmanaged version of a process with measurable outputs. Managed fermentation under a membrane cover, held 15 to 18 days with temperature checked daily, cut the resistance gene load reproducibly in this data.
Audit the whole input stack, not just the drug record. Metal and biocide resistance genes moved independently of antibiotic use here, so trace mineral levels and disinfectant routines deserve the same attention as the medicine book.
Use the ARG classes as a checklist for what you are selecting against. Sulfonamide, tetracycline, aminoglycoside, florfenicol, multidrug and MLS genes were all present and all fell under the same intervention. If your treatment history leans on any one of those classes, the matching gene is probably enriched in your bedding right now.
Stop treating antibiotic-free certification as a statement about resistance genes. The DL-S group was antibiotic-free and still carried more metal resistance genes than litter from a farm that used antibiotics. Certification describes inputs. Only measurement describes the reservoir.
If you are buying, add one question to the supplier form. What happens to the litter, for how long, at what temperature? A supplier who answers in numbers is running a process. One who answers with a shrug is running a stockpile.
Where the analysis is thin
Twenty samples from four farms is small, and the groups were not randomised. Each treatment group came from a different farm, so farm effects and treatment effects are tangled together. The antibiotic-using farm differs from the others in more ways than its drug record.
Almost all the sequencing sat on the DNA virome. The deep single-sample work covered one sample, and 34 vOTUs with 343 genes is a pilot look, not a survey. The RNA comparisons use the full twenty samples, which is more defensible but still modest.
The numbers are relative abundance from read mapping, not absolute copies per gram of litter. A fall in relative abundance is not proof of fewer gene copies overall, even though the direction holds across several unrelated gene families.
Of the 23,130 retained viral sequences, 22,764 were low quality. Normal for environmental viromics, but it means most of the viral signal is inferred rather than directly observed. And no bird-level health or production outcome was recorded, so this cannot tell you that a smaller litter reservoir means a better flock. It tells you the reservoir shrank.
Frequently asked questions
What are antibiotic resistance genes in poultry litter?
These are DNA sequences that let bacteria survive a drug, and in litter they sit inside bacteria, inside viruses, and on mobile elements that move between them. A 2026 metaviromic study of duck litter found sulfonamide genes sul1 and sul2, the tetracycline gene tetB(P) and the multidrug gene qacEdelta1. Abundance ranked by management group, so the reservoir is measurable and manageable.
Does removing antibiotics from feed remove resistance genes from litter?
No, and this is one of the less comfortable findings. Litter from a farm with no prophylactic antibiotic use and no fermentation still carried more resistance genes than either fermented group. Withdrawal stops new selection pressure, but genes already in the bedding persist and keep moving. Managed fermentation, not withdrawal alone, lowered the measured load.
How long does litter fermentation take to reduce resistance genes?
Piles were covered with a functional intelligent membrane and fermented for 15 to 18 days, with temperature monitored daily at three depths. That duration gave the lowest resistance gene abundance of the four groups. The paper reports no shorter or longer comparison, so 15 to 18 days is the only window tested. Shorter runs may work, but nothing here supports that yet.
Which resistance genes dropped the most after fermentation?
Sulfonamide genes sul1 and sul2, the tetracycline gene tetB(P), and qacEdelta1 all fell significantly. At the sharpest level, floR, aadA, qacEdelta1, tet(X) and cfr(A) were lower in all three comparison groups than in unfermented litter from an antibiotic-using farm, at P < 0.001. By class, aminoglycoside, florfenicol, multidrug, MLS, tetracycline and sulfonamide genes were all lower at P < 0.05.
Why did metal resistance genes behave differently?
Different inputs select for them. Metal resistance gene abundance fell in the order DL-D < DL-S-NAF < DL-S-ANF < DL-S, so antibiotic-free but unfermented litter carried more of these genes than litter from an antibiotic-using farm. Trace minerals in feed, disinfectants, copper and zinc all apply pressure. Fermentation lowered this load. Removing antibiotics did not touch it.
Can viruses carry antibiotic resistance genes between bacteria?
Yes, which is why sequencing the viral fraction matters here. Bacteriophages move genetic material between bacterial hosts, and how many are around reflects how much gene movement a system can support. The litter virome in this study was dominated by Caudoviricetes at roughly 94 to 95 percent across all groups. Fermentation cut RNA viral abundance significantly.
What is the best bioindicator for litter health in this research?
RNA viruses, and Pisuviricota specifically. The authors concluded these are sensitive bioindicators for ecological health assessment in duck litter. RNA viral composition shifted more clearly between management groups than DNA viral composition. Pisuviricota abundance ranged from 68.79 to 85.27 percent across groups, wider than DNA-dominant Uroviricota at 94.38 to 95.05 percent.
How much sequencing depth does a litter virome study need?
This one generated about 1.01 x 10^9 raw paired-end reads, roughly 151.5 Gb, from twenty DNA virome samples. After cleaning about 9.76 x 10^8 reads or 145.6 Gb remained, with Q20 above 97.75 percent and Q30 above 93.88 percent. That produced 5,170,738 contigs of at least 500 bp and 18,321 vOTUs. Routine monitoring needs less, but this sets the research bar.
Does fermentation also reduce bacterial virulence factors?
Yes. Virulence factor abundance followed the same order as antibiotic resistance genes: DL-D < DL-S-NAF < DL-S < DL-S-ANF. Specific factors including groEL2 and lirB dropped, and groEL2, lirB and two further factors were significantly lower in the deep fermentation group than in unfermented litter at P < 0.001. The intervention lowered both the resistance load and part of the virulence load.
Is antibiotic-free certification proof of low resistance gene load?
No. The clearest evidence against that assumption is the metal resistance result. Litter from an antibiotic-free farm that was not fermented carried a higher metal resistance gene load than litter from a farm that used antibiotics. Certification tells you what went into the birds. It says nothing about what sits in the bedding, and only measurement tells you that.
What should a poultry buyer ask a supplier about litter?
Ask what happens to the litter, how long it is treated, and at what temperature. In this study the managed process ran 15 to 18 days under a membrane cover with daily temperature checks at three depths, and it produced a measurable drop in resistance genes. A supplier who answers with numbers runs a process. One who answers with a shrug runs a stockpile.
What are the limits of this litter research?
Twenty samples from four farms, with each treatment group drawn from a different farm, so farm effects and treatment effects cannot be fully separated. The deep single-sample analysis covered one sample only. Results are relative abundance, not absolute gene counts per gram. Most of the 23,130 retained viral sequences were low quality. No bird health or production outcome was measured.
