Does perennial forage diversity influence microbially-mediated weed seed mortality in the soil seedbank?

January 2026 Update

Since January 2025, we analyzed 16s and ITS microbiome data for the soil and A. powellii seed DNA using Dada2 for bioinformatic pipeline and conducted data analysis with Phyloseq, PCOA and PERMANOVA in adonis2. Kara's Master's Thesis was successfully defended based on these results! In the Fall of 2025, we began writing our first paper to be submitted to weed science mostly focused on the effect of perennial forages on weed seed mortality, soil abiotic effects on mortality and perennial forage effects on microbiomes. The paper was submitted in December 2025. 

January 2025 Update 

By the summer of 2024, we completed all field measurements, including soil abiotic factors and the final seed bag retrieval. In Fall 2024, we concluded seed viability testing from the Spring 2024 seed bag retrieval and we extracted DNA from the Spring 2024 seeds and soil samples. In November 2024, we submitted all 16s and ITS PCR products from seed and soil samples for amplicon sequencing on a Next Seq2000 at the Huck Institute's Genomics Core Facility at Penn State. Currently, we are analyzing microbiome sequencing data and processing soil N samples.  We also have preliminary results for analyses of weed seed mortality, some soil abiotic factors (temperature and moisture) and biomass.  

In 2021, the perennial forage diversity gradient was established by the Lowry Weed Ecology Lab. Through a small seed grant from the Weed Science Society of America (WSSA Innovation Grant), we obtained funding for microbial sequencing of the weed seed microbiome for only a single timepoint. Additional funds from NE SARE were used to analyze the seed and soil microbiomes and soil abiotic factors at additional timepoints over the study duration (2.5 years) which improved our understanding of how crop species and soil abiotic factors influence the soil microbial communities and colonization of weed seeds over time.   

Experimental Design and Study Site 

All objectives  

The field site was established in September 2021 as a randomized complete block design with four blocks (4x7) at the Russell E. Larson Agricultural Research Center in Rock Springs, PA (40° 43′N, 77°56′W, 350m elevation). Soils are predominately comprised of Hagerstown silt loam (fine, mixed, mesic Typic Hapludalfs). Average air temperature at the site is 10.4°C with average winter temperatures of -1.4°C and average summer temperatures of 21.3°C; the site receives on average 1050.3mm of annual precipitation (NCEI 2024).   

To evaluate the effect of perennial forage composition and diversity on microbial or abiotic mediated weed seed mortality in the soil seedbank, we examined 7 perennial treatments including monocultures and all possible mixtures of alfalfa (A)(Medicago sativa - legume), orchard grass (O)(Dactylis glomerata - grass), and forage chicory (C)(Cichorium intybus - forb) and are hereby represented as A, C, O, AC, AO, CO, ACO.   

Each block consisted of 7 plots, each plot measuring 12.2 x 12.2m. See Appendix Figure 1.  

Agronomic Management: Prior to planting, the field was chisel plowed, disked, and then a soil finisher and cultimulcher were used to prepare the seedbed. Alfalfa, chicory, and orchard grass were drill seeded on 19-cm rows (see Appendix Table 1). Due to high winter kill, the chicory was re-seeded in April 2022.  We applied an equal low-level amount of N fertilizer across all treatments, despite chicory and orchardgrass not being legumes, to ensure any effects on seed mortality were due solely to the perennial forage treatments.  In Spring 2022 we applied 30 kg Nha^-1 equally across the entire field, and in Spring 2023 we increased the N amount we applied to 56 kg ^ha-1. Prior to planting and again in Spring 2023, we added P and K according to soil test recommendations. We applied Lambda cyhalothrin (Warrior II, Syngenta Crop Protection,Greensboro, NC) once in July 2022 at 0.034 kg ai ha^-1 for control of leaf hoppers (Empoasca fabae).No herbicides were used throughout the study duration.The field was mowed four times per year when the alfalfa was at bud to early flowering stage, dried and baled for hay. 

Abiotic Monitoring 

Objective 3  

We recorded soil moisture (VWC) weekly during the 2023 growing season (early May – late Oct) using a HydroSense II soil probe; only three measurement time points were taken in 2024 because the last seed extraction time point was in May 2024. Soil temperature readings were collected via HOBO sensors (1 per plot) which collect data every hour and were buried 5cm (2in) below the soil surface, which corresponds with the depth of the seed bags.  The HOBO sensors were buried in plots in 2023 and removed in May 2024. During the winter between 2023 and 2024, five HOBO sensors malfunctioned, and therefore temperature data for these plots were removed from the dataset. We then calculated the daily average, maximum, and minimum temperature for each plot across the duration in which the HOBO sensors were deployed in the field. 

Soil samples were collected twice per year in Spring 2023, Fall 2023, and Spring 2024 for inorganic nitrogen(N) analysis, but because other nutrients are less dynamic than N, we sampled soil once per year (Spring 2023 and 2024) for a full spectrum Mehlich 3 (ICP) analysis. For soil sampling we used a 2.5 cm diameter corer and collected 6 soil samples per plot at a depth of 20cm. Soil samples were then homogenized and a portion sent to the Penn State Ag Analytical Services Lab for the Mehlich 3 analysis.  To quantify soil inorganic N concentration, we incubated 20 g of fresh soil with 100 ml 2M KCl for 1 h on a shaker, then filtered extracts were frozen at-20°C until both NH4-N and NO3-N could be quantified based on the Berthelot and Greiss reaction, respectively. 

Biomass Collection 

All objectives  

Forage biomass was collected 4x per year (immediately prior to the field being mowed for harvest) within a 0.25m² quadrat (2 samples per plot), when alfalfa was at bud stage. Biomass was cut using electric hand sheers 5cm (2in) above the soil surface, sorted to species, dried at 60°C and weighed.  

Weed Seed Mortality Assessment, Soil Microbial Sampling, and Processing 

All objectives  

Weed seeds of Powell amaranth (a pigweed; Amaranthus powellii) and velvetleaf (Abutilon theophrasti) were collected in October 2021. A. powellii is a problematic and increasingly herbicide-resistant species (Aicklen et.al. 2022) and A. theophrasti is a persistent species in the weed seedbank due to its hard seed coat (Davis et.al. 2016). We used an air column to remove chaff and poor-quality seed and then treated A. theophrasti seed with chloroform for 24hrs to control for velvetleaf seed beetle (Althaeus folkertsi). Seeds of each species(100 A. theophrasti and 150 A. powellii, only one species per bag) were combined with 10g of sieved soil (2mm sieve) into fine mesh organza bags. The initial mortality of A. theophrasti was approximately 30%, while starting mortality for A.powellii was approximately 3%. Bags were then sewn shut and a hex nut was attached so bags could be located with a metal detector. In October 2021, we buried weed seed bags (10 bags of each species per plot) at a 5 cm depth adjacent to the crop row.

We first retrieved weed seed bags from the field at year 1 (Fall 2022) and then subsequently at 6-month intervals (Spring 2023, Fall 2023, and Spring 2024). At each timepoint, we retrieved 2 seed bags of each species per plot and immediately placed them on ice to limit disruption of the weed seed microbiome. The seeds were processed within 24 hours. Additionally, soil samples were collected from soil immediately surrounding the seed bags (to examine the soil impact on possible seed microbial shifts) and frozen for microbial analysis (Fall 2023 & Spring 2024 only).  

After retrieval, we pooled the two bags of each species collected from each plot,washed the contents of the seed and soil mixture in EtOH-sanitized sieves to remove soil (500μM sieves). We counted all intact seeds per bag to obtain the proportion of the original seeds we recovered at each retrieval timepoint (% Recovery)  

Equation 1:  

% Recovery = (# of seeds recovered) / (# of seeds originally placed in bag)  

We randomly selected 100 A. powellii and 11 A. theophrasti seeds (equivalent to ~ 100mg) and immediately placed them on ice for microbial processing. Within 24 hours of washing the seed bags, we surface sterilized (10% bleach for 10 min with 3 rinses in sterile water)(Ercoli et.al. 2007; Chakrabarti 1977) and freeze dried the seeds designated for microbial analysis. Seeds were freeze-dried in sterilized microcentrifuge tube covered with breathable sealing film in a freeze dryer for 10 hours at-26°C and <500m Torr. With SARE funding, only the A. powellii seeds were used for future microbial processing. The remaining seeds were stored at 4°C for viability/mortality testing. See Appendix Figure 2.  

To evaluate weed seed mortality, we conducted tetrazolium (TZ) testing according to the AOSA handbook (Miller 2010) using 0.1% TZ with minor alterations; A. powellii seeds were pierced instead of bisected for TZ staining and then later dissected and A. theophrasti seeds were nicked (scarified) to quicken imbibition. Initial mortality testing protocols for A. powellii seeds in Fall 2022 (the first retrieval) varied from the above method and were not as reliable as the later developed method; therefore, A. powellii mortality data from this timepoint were discarded.   To account for the microbial subsample of unknown mortality, total percent weed seed mortality (M_total ) was calculated as:  

Equation 2:  

M_total= 1 - ((% Recovery) x (% TZ viable)) 

Seed mortality data analysis:  

To determine whether perennial forage community influences weed seed mortality, we used generalized linear mixed effects models with a beta distribution (recommended for proportional data) with forage treatment as a fixed effect and block as a random effect. All data analyses will occur in R (version 2023.09.1).  

Microbial DNA Extraction and Sequencing 

All objectives  

We used the Qiagen DNeasy Plant Mini Kit to extract DNA of seed endophytic microbes. The Qiagen kit's 20mg restriction of biological tissue greatly limits the number of seeds we could extract from (ex. For velvetleaf, 3 seeds may reach 20mg). To obtain a more representative sample, we pooled subsamples to reach an initial 40mg (2x) weed seeds, ground in a bead mill, and doubled the extraction solutions for the initial steps. Then only 1x of lysate (350uL) was put into the Qiagen Shredder Column and normal protocols were followed thereafter. Additionally, we extracted DNA from the original, unburied seeds collected in Fall 2021 (used to fill the seed bags) to determine a baseline of microbial endophytes. We extracted DNA from soil, frozen from field collection, using the Qiagen Power Soil Kit. Negative controls to account for contaminants in the processing steps included (1) blank runs on each extraction kit as well as (2) extractions of sterilized metal beads that were subjected to the same protocols as the seeds (surface sterilization, freeze drying, weighing, etc.) and would indicate contaminants in processing steps. Extracted DNA was quantified using a Qubit Fluorometer (broad spectrum kit) and all the negative controls had undetectable amounts of DNA.

Then, we performed the first of a two-step amplicon PCR process using universal primers with Illumina adapters to amplify the ITS (1-2) marker region for fungal taxonomic diversity and the 16S rRNA gene (V4) region for bacterial taxonomic diversity. For the initial PCR reaction, we used ITS 1-2 primers (Gardes and Bruns 1993; Illumina 2019) or 16S rRNA gene V4 515F and 806R (Parada et al. 2016; Apprill et al. 2015), with Illumina adapters. For both ITS and 16S rRNA gene amplification, we used Invitrogen^TM 2X Platinum^TM Superfi^TM II Green PCR Master Mix to run 20 μL PCR reactions containing: 10 μL mastermix (1x), 0.5 μM each forward and reverse primer, ~15 ng DNA, and sterile nuclease free water to reach 20 μL (concentrations are listed as final concentrations). pPNA and mPNA blockers from PNABio were added to 16S rRNA gene reactions for a final concentration of 1.25 μM each to reduce amplification of plastids and mitochondria, respectively. PCRs were conducted the same for both 16S rRNA and ITS and run on a thermocycler according to the Superfi II Mastermix protocol: initial denaturing (98°C for 30sec), denature-anneal-extend repeating 30 times (98°C for 10 sec, 60°C for 10 sec, 72°C for 15 sec), final extension (72°C for five min) and a hold at 4°C (Invitrogen 2022). PCR blanks (PCR reagents without DNA) were also run to identify any contamination in the PCR reagents. PCR products were checked using a 2% agarose gel, and all controls had no visible amplification. Samples were then submitted to the Penn State University Huck Institute’s Genomics Core Facility where a second round of PCR was completed to attach index barcodes for sample identification and Illumina sequence adapters, and then amplicons were sequenced (175k pair-end reads) with the Illumina NextSeq 2000 XLEAP P1 600 cycle kit and performing 300 x 300 paired-end sequencing.

Microbiome Data Processing and Analysis

The sequences were demultiplexed by the sequencing facility. We used the DADA2 pipeline (version 3.18; Callahan et al. 2016) to filter, remove primers, trim reads (according to phred score), pseudo-pool samples during sample inference, pair reads and create ASV (Amplicon Sequence Variant) abundance tables. For 16S rRNA gene data, after construction of the sequence table in dada2, only paired ASV’s with length of 254 bp (the length of the v4 segment) were selected for downstream processing; selecting ASV’s of the expected length (based on primers) is intended to remove noise from longer segments likely created from non-specific priming (Callahan et al. 2016). The majority of the ASV’s were exactly 254 bp. We used the GreenGenes2 database (version 2024.09 genus level) (Callahan et al. 2016; McDonald et al. 2024) for 16S rRNA gene taxonomic assignment and the UNITE All-Eukaryotes database (Abarenkov et al. 2024) for ITS taxonomic assignment. Then in “phyloseq” (version 1.46), we removed all 16S rRNA gene ASVs not classified as bacteria, ASVs unclassified at the phylum level as well as 16S rRNA gene ASV's associated with mitochondria or chloroplasts. For ITS, we removed all non-fungal ASVs and all ASVs unclassified at the phylum level. Additionally, we removed poor quality ITS ASVs containing 20 base pairs repetitions or longer homopolymers, which are common to ITS (Lindahl et al. 2013). To eliminate the influence of rare ASVs in both 16S rRNA gene and ITS, we removed all ASVs that consisted of less than 20 total reads or were found in less than 5% of samples. Both the 16S rRNA gene and ITS samples were independently rarefied to the minimum number of reads per sample using “phyloseq” (function rarefy_even_depth), which was 8,284 reads for 16S rRNA gene and 38,980 reads for ITS. We did this because rarefying helps obtain reliable and comparable estimates of alpha diversity across variation in sequencing depth (Weiss et al. 2017). All downstream analyses occurred on rarefied datasets.

To examine whether the perennial forage treatment affected the A.powellii seed microbiome,we analyzed 16S rRNA gene and ITS data separately and performed a PERMANOVA on the Bray-Curtis dissimilarity matrix using the adonis2 function in the vegan package in R (Oksanen et al.2025), and pairwise comparisons between treatments were performed with the pairwise.adonis function in the pairwiseAdonis package. To visualize A.powellii seed communities, we performed Principal Coordinates Analysis (PCoA) on both the Bray-Curtis dissimilarity matrix and a Jaccard dissimilarity matrix (after converting the community matrix to presence-absence). Finally, we calculated the Observed and Shannon Diversity using the estimate_richness function in Phyloseq, and we examined whether both forage richness and treatment affected both diversity metrics for both weed species using linear mixed effect models with richness or treatment,retrieval timepoint, and their interaction as fixed effects and block as the random effect.