The V. nonalfalfae and V. albo-atrum strains were stored in our laboratory. Eight fungal isolates (Exserohilum turcicum, V. nigrescens, Magnaporthe oryzae, Bipolaris maydis, Rhizoctonia cerealis, Fusarium oxysporum f. sp. conglutinans, Ustilaginoidea virens, and Fusarium pseudograminearum) and five bacterial isolates (Acidovorax citrulli, Xanthomonas oryzae pv. oryzae, Pseudomonas syringae, Ralstonia solanacearum, and Xanthomonas campestris pv. campestris) were obtained from the Chinese Academy of Agricultural Sciences and China Agricultural University. All samples were stored in 25% glycerol at -80°C.

Samples of alfalfa stems and roots were collected in an experimental field (Langfang, Hebei Province, China, 39°51’N, 116°60’E) and quickly frozen in liquid nitrogen. The plot had been sampled and tested numerous times previously for infection by V. nonalfalfae and V. albo-atrum; therefore, the infection status of each alfalfa plant was known.

We excised 200 mg fresh alfalfa stems and roots from each sample, respectively. They were each homogenized separately in 4 mL lysis buffer. Nucleic acid quality and quantity were determined using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, USA).

The internal transcribed spacer regions (ITS) sequences of V. albo-atrum (MH856937.1), V. nonalfalfae (KT362917.1), V. alfalfae (NR 126129.1), V. longisporum (MH864843.1), V. nubilum (MH856939.1), Ilyonectria radicicola (KM503139.1), V. dahliae (MH864842.1), Volutella ciliata (JQ647447.1), V. zaregamsianum (JN188008.1), V. tricorpus (MH857388.1), V. klebahnii (NR 126128.1), V. tricorpus (GO336726.1), Acremonium nepalense (LR590100.1), Plectosphaerella cucumerina (KM357306.1), and Brunneochlamydosporium nepalense (MH860634.1) were downloaded from the National Center for Biological Information (https://www.ncbi.nlm.nih.gov/). To construct a phylogenetic tree, the ITS sequences were compared using the ClustalW program (http://www.clustal.org/clustal2/), and the neighbor-joining approach was applied using the MEGA software (version 7.0.26, Auckland, New Zealand).

The biological samples were re-cultured in complete medium and Luria-Bertani liquid medium at temperatures of 25°C and 37°C, respectively. Samples were harvested by centrifugation at 3000 ×g for 10 min. Subsequently, DNA was extracted following the manufacturer’s instructions (KG203; Tiangen, Beijing, China). DNA was extracted from alfalfa stems following the instructions of the Hi-DNAsecure Plant Kit (DP350; Tiangen, Beijing, China). Genomic DNA was isolated from the soil samples using a TIANamp Soil DNA Kit according to the manufacturer’s protocol (DP336; Tiangen, Beijing, China). The quality and concentration of genomic DNA were detected via 1% agarose gel electrophoresis and a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, USA). Subsequently, samples were aliquoted and stored in refrigerator at -80°C.

The ITS sequences were amplified with the primers 5’-TCCGTAGGTGAACCTGCGG-3’ and 5’-TCCTCCGCTTATTGATATGC-3’ in V. nonalfalfae and V. albo-atrum strains and sequenced by Sangon Biotech Co., Ltd. (Shanghai, China). The results were aligned using the Basic Local Alignment Search Tool (https://blast.ncbi.nlm.nih.gov/Blast.cgi).

The primer/probe sets used in the qPCR and ddPCR experiments ( Table 1 ) were preliminarily screened and evaluated to confirm the complete identity using only the target sequences.

Probes were synthesized and labeled with 6-carboxy-fluorescein reporter dye and black hole quencher 1 on the 5’- and 3’-terminal nucleotides, respectively (Sangon Biotech). The concentration was diluted to 100 μM as stock solution, stored at -20°C, and diluted to 10 μM for the assay.

The qPCR amplification was accomplished using the ABI 7500 Fast Real-Time PCR system (Applied Biosystems, Foster City, CA, USA). The final reaction was in a 20 μL volume containing 10 μL of 2×AceQ Universal U+ Probe Master Mix V2 (Vazyme Biotech Co., Ltd, Nanjing, China), 0.4 μL each of the forward and reverse primers, 0.2 μL of the probe solution, 2 μL of genomic DNA, and 7 μL of nuclease-free H₂O. The amplification was performed as follows: denaturation at 95°C for 5 min, followed by 40 cycles of denaturation at 95°C for 10 s and annealing and elongation at 60°C for 50 s. The cycle threshold (Ct) values were calculated using 7500 software v2.3 (Applied Biosystems, USA).

A standard curve for qPCR was constructed using the genomic DNA of V. nonalfalfae and V. albo-atrum. DNA concentration was measured using a NanoDrop 2000 (Thermo Scientific). The standard curve was plotted according to the Ct values from qPCR for various sample amounts, which were obtained by serial dilution of the genomic DNA of V. nonalfalfae and V. albo-atrum.

The ddPCR assay was optimized with respect to the annealing temperature, primer concentration, and probe concentration. Temperatures of 54°C, 56°C, 58°C, 60°C, and 62°C were examined using a Veriti™ 96-Well Thermal Cycler (Applied Biosystems, USA) with a primer/probe concentration of 500 nM/250 nM. Primer/probe concentrations of 600 nM/400 nM, 500 nM/250 nM, and 400 nM/100 nM were examined using a thermal cycler at 56°C.

The ddPCR reactions were performed using 500 nM solutions of each forward and reverse primer, a 250 nM solution of the 2 × ddPCR supermix for probes (Bio-Rad, Pleasanton, CA, USA) (after optimization), and 2 μL of genomic DNA. The total reaction volume was 20 μL. The reaction mixture was added to an 8-well cartridge using a droplet generator (Bio-Rad), and the wells were injected with 70 μL of droplet generation oil (Bio-Rad). The emulsion (70 μL) was transferred to a 96-well ddPCR plate (Bio-Rad) and amplified as follows with a ramp rate of 2°C s⁻¹: denaturation at 94°C for 10 min, followed by 45 cycles of denaturation at 94°C for 30 s and annealing at 56 °C (optimized temperature) for 50 s, followed by a final denaturation at 98°C for 10 min. Subsequently, the 96-well plate was shifted to an QX200 droplet reader (Bio-Rad), and the data analysis was performed using QuantaSoft Version 1.7.4.0917 (Bio-Rad). The ddPCR analysis was performed in triplicate to accurately quantify the amounts of genomic DNA.

Linear regression curves were constructed using ten-fold serial dilutions of genomic DNA as templates, ranging from 4 to 4 ×10⁴ copies/reaction and 6 to 6 ×10⁴ copies/reaction for V. nonalfalfae and V. albo-atrum, respectively. Three independent experiments were performed with four reactions at each concentration. The linear regression curves were generated by plotting the copy number concentrations measured by ddPCR against the expected dilution values.

The limit of quantification (LoQ) was determined as the lowest concentration where all runs were positive and the coefficient of variation (CV) of the measured copy number was up to 25% (Kralik and Ricchi, 2017).

The limit of detection at 95% probability (LoD95%) was defined as the lowest concentration at which at least 95% of the replicates produced positive results, regardless of the accuracy or precision (Forootan et al., 2017). The LoD95% of each assay was estimated using a probit analysis with the target concentration (dilution level) as an explanatory variable and the positive detection rate (positive/total) as a response variable (Corman et al., 2020). The probit analysis was conducted with 95% confidence intervals (95% CI) using the SPSS software (version 20.0; SPSS, Chicago, IL, USA).

To obtain the LoQ and LoD95% of the ddPCR and qPCR assays, a series of two-fold dilutions of genomic DNA (from 1.1 to 294 copies/reaction for V. nonalfalfae and from 1.8 to 236 copies/reaction for V. albo-atrum) were tested to ascertain the end-point dilutions. Two independent experiments were performed over two consecutive days, with five replicates of each concentration on each day (Pavšič et al., 2016). These experiments were tested in parallel, and the template DNA was only frozen and thaw once. The distilled water was used as a negative control.

A specificity evaluation panel was used to explore the specificity of the ddPCR assays for V. nonalfalfae and V. albo-atrum and included the eight fungal and five bacterial isolates listed in the “Fungal, bacterial, soil, and plant samples” section above.

Receiver operating characteristic (ROC) curves were generated to compare the diagnostic performance of the ddPCR and qPCR assays. True positive alfalfa roots and stems that were infected with V. nonalfalfae or V. albo-atrum were used to generate the ROC curves based on previous data. ROC curves were constructed using the SPSS software (version 20.0; SPSS, Chicago, IL, USA).

Quantitative agreement between ddPCR and qPCR was investigated using the method of Bland and Altman (Bland and Altman, 1986), which utilized the log₁₀-transformed concentration of each sample. The inhibition effect of the residual matrix on both ddPCR and qPCR assays was evaluated by quantifying a constant amount of gDNA in the presence of alfalfa stem and root extracts. We spiked the reactions with the same amount of gDNA from cultured isolates (V. nonalfalfae: 1.6 ×10⁴ ITS molecules; V. albo-atrum: 2.4×10⁴ ITS molecules), and 5 μl of extracts from healthy cotton stems and roots. The exact preparation process described by Maheshwari et al. (Maheshwari et al., 2017).

For each alfalfa field sample, CV values for ddPCR and qPCR were determined by calculating the standard deviation within each set of triplicates (n = 4) and dividing each of these by their respective average concentration values. The CV values for qPCR were determined using linear-scale copy number concentrations rather than the Ct values, which were directly comparable to those of ddPCR.

To analyze the position of Verticillium in a broader fungal phylogenetic context, the phylogenetic relationships among major Verticillium spp. and other fungi were analyzed using ITS sequences. Phylogenetic analysis indicated that the 15 selected species were divided into two distinct clades, and V. nonalfalfae and V. albo-atrum had the closest phylogenetic relationship ( Figure 1A ). This is consistent with the fact that both V. nonalfalfae and V. albo-atrum belong to Verticillium and share a similar pathogenic mechanism that results in symptoms of plant stunting, wilting, and early senescence. Hence, we aimed to develop a ddPCR assay that enables the quantitative detection of V. nonalfalfae and V. albo-atrum simultaneously, to improve the efficiency of fungal diagnoses.

To develop an optimal ddPCR assay, four reported primer/probe sets targeting the different regions of the two fungi (KT362917.1 and MH856937.1 in the Genebank database) were initially validated using qPCR systems. The results showed that Va4 primer/probe set performed better than the other primer/probe sets, obtaining the target amplicons with low Ct values for both fungal samples ( Table S1 ; Figure S1 ). Based on the sequence alignment, the amplification sequence of the Va4 primer/probe set was identical for V. nonalfalfae and V. albo-atrum ( Figure 1B ). Therefore, Va4 primer/probe set was selected for further establishment of the ddPCR assay.

The ddPCR reaction conditions were optimized with respect to annealing temperature and primer/probe concentrations by evaluating five temperatures (ranging from 54°C to 62°C) and three primer/probe concentrations (600/400 nM, 500/250 nM, and 400/100 nM). As shown in Figure 2 , the optimal annealing temperature and primer/probe concentrations were 56°C and 500/250 nM, respectively, which enabled the best separation of the positive and negative droplets, with a low abundance of “rain” (droplets falling between the positive and negative populations).

For the linear regression analysis, serial dilutions of genomic DNA extracted from the two cultured isolates were prepared. The ddPCR assays exhibited good linearity between the measured and anticipated values for each interval for V. nonalfalfae (R² = 0.9996 and 0.94<Slope<1.03) and V. albo-atrum (R² = 0.9997 and 0.96<Slope<1.04) over a dynamic range from approximately 10¹ to 10⁴ copies/reaction ( Figures 3A, B ). The LoQ of the ddPCR assay was 43 and 30 copies for V. nonalfalfae and V. albo-atrum, respectively, which met the criterion for a LoQ with a CV lower than 25% ( Figures 3C, D ).

Specificity evaluation indicated that the runs were specific for V. nonalfalfae and V. albo-atrum, and no cross-reactivity was observed with the other eight similar fungal and five bacterial samples ( Table S2 ).

To determine the LoD95% for ddPCR and qPCR, a series of two-fold dilutions of genomic DNA of the two fungi were prepared as templates at concentrations near the detection end points determined in preliminary experiments. As shown in Figure 4 , the probit analysis of the ddPCR assay indicated a LoD95% of 6.8 copies/reaction (95% CI: 5.1–19.3) for V. nonalfalfae and 7.2 copies/reaction for V. albo-atrum (95% CI: 5.6–14.2) ( Figures 4A, B ). In contrast, the estimated LoD95% values for qPCR were slightly lower, at 5.6 copies/reaction (95% CI: 4.4–12.8) and 5.2 copies/reaction (95% CI: 4.0–15.7) for V. nonalfalfae and V. albo-atrum, respectively ( Figures 4C, D ). The results indicated that the two methods showed similar sensitivity for the pure genomic DNA of both fungi, and the resulting LoD95% values of ddPCR were slightly higher than those of qPCR for the two fungi.

We further investigated the applicability of ddPCR via a comparison with qPCR using field samples. We collected stems and roots from 60 alfalfa plants in the field, each including 10 healthy tissue controls and 50 V. nonalfalfae- or V. albo-atrum-infected samples, which had been previously confirmed.

ROC analysis showed that for ddPCR, the area under curve (AUC) values were 0.94 (standard error 0.03, 95% CI: 0.88–0.99) and 0.90 (standard error 0.04, 95% CI: 0.82–0.97) for roots and stems, respectively ( Figures 5A, B ). In contrast, the AUC values for qPCR were 0.86 (standard error 0.05, 95% CI: 0.78–0.95) and 0.76 (standard error 0.06, 95% CI: 0.63–0.89) for roots and stems, respectively ( Figures 5C, D ). The AUC values of the ddPCR assays were broader than those of the qPCR assays, especially for the stems. Therefore, the ddPCR method provided a more robust diagnostic performance than that of qPCR for distinguishing the V. nonalfalfae- and V. albo-atrum-positive cases from healthy samples.

In addition, the cut-off points of the qPCR assays were determined based on the ROC curves, with values of 36 for the stem and 35 for the root samples. Among the 50 positive cases, ten stem and eight root samples were reported to be negative by qPCR. However, six out of the ten false-negative stems and four out of the eight false-negative roots tested positive by ddPCR. Furthermore, the qPCR-negative but ddPCR-positive roots and stems had an average load of 17 and 30 copies, respectively. Hence, ddPCR showed improved detection sensitivity compared to qPCR, especially regarding stem samples.

The copy number concentrations for ddPCR and qPCR were compared using the samples that tested positive by both methods. Linear regression analysis indicated that the copy numbers determined by ddPCR correlated well with those determined by qPCR in both alfalfa tissues (roots, R² = 0.91; stems, R² = 0.76) ( Figures 6A, B ).

Regarding quantitative agreement, ddPCR produced the higher copy numbers of ITS than qPCR in most tested samples. As shown in Figures 6C, D , Band–Altman plots showed that the bias for the agreement ranged from -0.21 to 0.51 log₁₀ and -0.31 to 0.82 log₁₀ for roots and stems, respectively, with most points falling inside the 95% CI. The average difference in quantification between the two methods was 0.16 log₁₀ (a difference of 1.45 times on a normal scale) for roots and 0.25 log₁₀ (1.78 times) for stem samples. Moreover, the intervals defined by the 95% CI for root samples were narrower than those for stem samples. To determine whether the residual matrix of field samples caused the quantification difference, we quantified the samples that spiked with equal amounts of gDNA from cultured isolates and extracts from healthy alfalfa stems and roots. The extracts inhibited the quantification of the spiked DNA by both methods, but the concentrations determined by the ddPCR were significantly higher than qPCR ( Figure 7 ).In addition, absolute quantification by ddPCR revealed greater precision in V. nonalfalfae and V. albo-atrum abundance compared to those of qPCR, and the CV values were 18–72% and 28–76% lower than those of qPCR with respect to overall variation for alfalfa roots and stems, respectively ( Figure 8 ).