Virulence Differences in Blumeria graminis f. sp. tritici from the Central and Eastern United States

Wheat powdery mildew, caused by Blumeria graminis f. sp. tritici, is a globally damaging disease that occurs to varying degrees in most wheat-growing regions of the United States. Historically, within the United States, it has been most damaging in the mid-Atlantic states; however, weather patterns combined with large acreages of susceptible cultivars have recently resulted in multiyear periods of unusually severe and extensive powdery mildew outbreaks in the southeastern states, especially Georgia, and in Montana (J. Johnson and M. Burrows, personal communications). Thus, the geographic range over which powdery mildew epidemics are sometimes or often severe is expanding in the United States.

B. graminis f. sp. tritici oversummers as chasmothecia or mycelium on wheat stubble. In mildew-prone areas, it typically infects susceptible winter wheat crops as soon as weather allows. Early epidemic initiation allows for additional cycles of asexual reproduction. A trend toward warmer winters in the eastern United States, which is predicted (Melillo et al. 2014), should increase the frequency and severity of B. graminis f. sp. tritici epidemics by encouraging earlier epidemic onset.

There have been two previous virulence surveys of the U.S. B. graminis f. sp. tritici population at roughly one-decade intervals. In 1993 and 1994, 520 isolates were derived from samples originating from 17 states, and virulence frequencies were determined (Niewoehner and Leath 1998). In 2003 and 2005, the population in 10 locations in the southeastern United States was sampled, and a collection of 207 isolates was characterized for virulence (Parks et al. 2008).

Shifts in virulence are common in this pathogen population. Most of the wheat powdery mildew resistance (Pm) genes that have been widely deployed in the United States have been defeated. In recent times, the most consequential shifts in B. graminis f. sp. tritici virulence in the mid-Atlantic United States have been the defeat of Pm3a (cultivar Saluda, late 1980s), Pm4a (cultivar Roane, mid- to late 1990s), and Pm17 (cultivars Tribute and McCormick) (Cowger et al. 2009; Niewoehner and Leath 1998; Parks et al. 2008). Pm17 was originally introgressed along with greenbug (Schizaphis graminum) resistance, the primary trait of interest, via a 1AL.1RS wheat-rye translocation into wheat germplasm line Amigo in Oklahoma (Sebesta and Wood 1978). In the mid-Atlantic United States, cultivars bearing Pm17 were released in 2002 and planted on large acreages starting in 2004 (Griffey et al. 2005a, b); the gene’s breakdown began in 2009 (Cowger et al. 2009) and was complete in that region by 2012 (authors’ personal observations).

Although partially mildew-resistant wheat backgrounds are essential and have provided durable resistance in some settings (Brown 2015), genes that even temporarily confer high levels of resistance have also been necessary to achieve adequate protection of wheat crops in the mid-Atlantic United States. In North Carolina, starting in 1987, 16 new and highly effective resistance sources have been identified in diploid and tetraploid common wheat relatives and introgressed into a soft red winter wheat background adapted to the eastern United States, resulting in five Pm gene and two temporary gene designations (Cowger et al. 2012; Petersen et al. 2015; Worthington et al. 2014). These genes were introgressed from relatives in the Triticum and Aegilops genera.

To update information on B. graminis f. sp. tritici virulence patterns in the United States, we conducted a large, multistate survey in 2013 and 2014. With the help of collaborators in 15 states, we sought to determine frequencies of virulence to both the older Pm genes and some of the genes recently introgressed from wild wheat relatives.

Prior to the present study, an observed feature of the U.S. B. graminis f. sp. tritici population was the difference between the U.S. Plains, where hard wheat is grown, and the eastern soft-wheat area. The gene Pm3a was effective in Texas in 1993 and 1994 (Niewoehner and Leath 1998) and still remained effective in Oklahoma in 2008 (Chen et al. 2009). Meanwhile, in 1993 to 1994, the frequency of virulence to Pm3a was intermediate (50 to 56%) in Kansas, Indiana, and Kentucky in 1993 and generally high (50 to 100%) in the Great Lakes and eastern-seaboard states of North Carolina, Pennsylvania, and Georgia (Niewoehner and Leath 1998). The frequency of virulence to Pm3a reached 100% in the mid-Atlantic and southeastern states by 2003 and 2005 (Parks et al. 2008).

Our research using presumably neutral molecular markers had indicated that the B. graminis f. sp. tritici populations east and west of the Appalachian Mountains were genetically subdivided, with migration generally from west to east (Cowger et al. 2016). B. graminis f. sp. tritici populations were isolated by distance, suggesting relatively local inoculum sources rather than annual continent-scale extinction and recolonization, and the Southeast and Southern Plains populations were genetically distinct from the Great Lakes and Mid-Atlantic populations (Cowger et al. 2016). The detection of migration from west to east but not from east to west (Cowger et al. 2016) helped explain the maintenance of Pm3a effectiveness in the Plains but not in the eastern United States.

The present virulence survey occurred approximately a decade after the last survey and covered a wider geographic area. The aim was to assess the distribution of frequencies of virulence to both old and new Pm genes and the implications for Pm-gene deployment, and we hypothesized that these frequencies would vary among regions. In addition, the sampling was conducted at a hierarchy of spatial scales, including at multiple within-field sites, in order to better understand how local epidemics originate and develop, and the distribution of both selected and nonselected variation. The part of the study described here concerns virulence but other aspects of the sampled population, including fungicide sensitivity and population structure, were also assessed, and those results will be reported separately.

MATERIALS AND METHODS

Mildew sampling and isolate collection.

Whole wheat plants with B. graminis f. sp. tritici pustules were kindly collected by collaborators in a total of 15 states over 2 years (Fig. 1; Table 1). B. graminis f. sp. tritici isolates were recovered from plants received from 20 fields across 11 different states in 2013 and 16 fields across 13 states in 2014. Plants were sampled from fields of susceptible or moderately susceptible wheat cultivars (in five fields, the cultivar was unknown) that were not treated with fungicides prior to sampling. GPS coordinates were provided for fields and, in a few cases, for sites within fields. In most cases, three plants, each separated by approximately 1 m from each other, were sampled at each of five widely separated sites per field; the plants were bagged separately for shipment. In the lab, spore transfers were attempted from five different leaves or stems of each sampled plant. Thus, the protocol resulted in up to 75 isolates per field.

Download as PowerPoint

TABLE 1. Year of collection, nearest town, and cultivar of origin of 1,017 Blumeria graminis f. sp. tritici isolates derived from 36 U.S. wheat fields in 2013 and 2014

Year, field	State	Town	Wheat cultivar	Na
2013
1	Alabama	Prattville	Unknown	15
2	Georgia	Donalsonville	AGS2060	18
3	Georgia	Marshallville	Unknown	41
4	Georgia	Pine Mountain	Magnolia	56
5	Michigan	Deckerville	Hopewell	18
6	Michigan	Rogers City	MSU D 8006	23
7	Missouri	Lamar	Unknown	37
8	Mississippi	Barrontown	Magnolia	37
9	Mississippi	Greenwood	Coker 9553 or Armor Ricochet	28
10	Mississippi	Sunrise	Magnolia	39
11	North Carolina	Calypso	Coker Oakes	51
12	North Carolina	Four Oaks	Featherstone	38
13	New York	Aurora	Pioneer 25R34	12
14	New York	Brockport	Otsego	22
15	Ohio	Jeromesville	Croplan 8309	29
16	Oklahoma	Hinton	Jackpot	20
17	Oklahoma	Stillwater	OK Bullet	13
18	Pennsylvania	Chambersburg	P25R39	35
19	Pennsylvania	Pennsylvania Furnace	Growmark 820	38
20	Virginia	Saluda	Southern States 8404	14
Total	…	…	…	584
2014
21	Alabama	Athens	Unknown	49
22	Arkansas	Fayetteville	Ricochet	14
23	FL	Gretna	AGS2060	26
24	Georgia	Sandersville	AGS2000	32
25	Georgia	Thomasville	AGS2060	26
26	KS	Manhattan	2137	22
27	Michigan	Rogers City	MSU-D6234	10
28	Missouri	Bronaugh	MFA2525	15
29	North Carolina	Clayton	USG 3120	45
30	North Carolina	Mt. Olive	Coker Oakes	23
31	NE	Lincoln	Unknown	11
32	New York	Brockport	Otsego	29
33	New York	Groveland	Otsego	54
34	Ohio	Wooster	Bravo	26
35	Oklahoma	Stillwater	Pete	10
36	Pennsylvania	Pennsylvania Furnace	Chemgro DH75	41
Total	…	…	…	433
Grand total	…	…	…	1,017

^a Number of isolates recovered.

View as image HTML

Isolates were cultured by tapping infected tissue over detached leaf segments of the universally susceptible cultivar Chancellor (CI 12333) placed on 0.5% benzimidazole-amended water agar. Genetically pure isolates were obtained through two rounds of single-spore subculturing, which resulted in a total of 1,017 individual isolates, 584 of which were derived in 2013 and 433 in 2014 (Table 1). Isolates were maintained on detached-leaf plates at 17°C with 12 h of light, and transferred to fresh leaf tissue every 10 days.

Virulence ratings.

The virulence profile of each isolate was determined using detached leaves of a set of 15 differential wheat lines in 2013 and 20 lines in 2014, where each line contained a single Pm gene (Table 2). Each plate contained a block of leaf segments consisting of two end-to-end segments of each differential line. In addition, one leaf segment of Chancellor was placed along each of the four edges of the block as a susceptible control. The plate was the experimental unit, and two such plates were inoculated per isolate, resulting in two replicates (each with two leaf segments) per differential line for each isolate. Differential plates were inoculated using settling towers to evenly distribute spores across the plates and leaf surfaces, and plates were incubated under the conditions described above.

TABLE 2. Genes used in virulence tests of Blumeria graminis f. sp. tritici isolates collected from the eastern United States in 2013 and 2014^a

Gene	Cultivarb	Pedigree	Resistance donor species	Genomes	Chrc	Efficacyd	Reference
Pm1a	C.I. 14114	Axminster/8*Chancellor	Triticum aestivum	AABBDD	7AL	R	Sears and Briggle 1969
Pm1b	MocZlatka	ATRI-1509-SLK(TR.MO)/(TR.DR)ATRI-3310-SLK//*Zlatka[1934]	T. monococcum	AA	7AL	R	Hsam et al. 1998
Pm2	C.I. 14118	Ulka/8*Chancellor	Aegilops tauschii	DD	5DS	MS	McIntosh and Baker 1970
Pm3a	C.I. 14120	Asosan/8*Chancellor	T. aestivum	AABBDD	1A	S	Briggle and Sears 1966
Pm3b	C.I. 14121	Chul/8*Chancellor	T. aestivum	AABBDD	1A	S	Briggle and Sears 1966
Pm3f	C.I.15888	Michigan Amber/8*Chancellor	T. aestivum	AABBDD	1A	S	Zeller et al. 1993
Pm4a	C.I. 14123	Khapli/8*Chancellor	T. dicoccum	AABB	2AL	S/PR	Ma et al. 2004
Pm4b	Ronos	Graf/Kormoran//Kronjuwel	T. carthlicum	AABB	2AL	S/PR	Yi et al. 2008
Pm6	Coker 747	Arthur/Coker-68-15	T. timopheevii	AAGG	2B	S/PR	Jorgensen and Jensen 1973
Pm8	Kavkaz	Lutescens 314H147/Bezostaja 1	S. cereale	RR	T1BL.1RS	S/PR	Hsam and Zeller 1997
Pm10,15	Chancellor	Dietz/Carina//Carina/Mediterranean/3/Kanred“S”/Purplestraw	T. aestivum	AABBDD	1D, 7DS	S	…
Pm16		Norman/T. dicoccoides line	T. turgidum var. dicoccoides	AABB	5BS	R	Reader and Miller 1991
Pm17	Amigo	Teewon“S”/6/Gaucho/4/Tascosa/3/Wichita/Teewon/5/2*Teewon	S. cereale		T1AL.1RS	S	Heun et al. 1990
Pm25	NC96BGTA5	Saluda/PI427662//Saluda/3/Saluda	T. monococcum subsp. aegilopoides	AA	1A	R	Shi et al. 1998
Pm34	NC96BGTD7	Saluda *3/TA2492	A. tauschii subsp. tauschii	DD	5DL	R	Miranda et al. 2006
Pm35	NC96BGTD3	Saluda *3/TA2377	A. tauschii subsp. strangulata	DD	5DL	R	Miranda et al. 2007a
Pm36	5-BIL29	Latino *5/MG29896	T. turgidum var. dicoccoides	AABB	5BL	R	Blanco et al. 2008
Pm37	NC99BGTAG11	Saluda *3/ T. timopheevii accession	T. timopheevii subsp. armeniacum	AAGG	7AL	R	Perugini et al. 2008
Pm53	NC09BGTS16	Saluda *3/TAU829	A. speltoides	SS	5BL	R	Petersen et al. 2015
MlAG12	NC06BGTAG12	Saluda *3/PI 538457	T. timopheevii subsp. armeniacum	AAGG	7AL	R	Maxwell et al. 2009
NCA6	NC96BGTA6	Saluda *3/PI 427772	T. monococcum subsp. aegilopoides	AA	7AL	R	Miranda et al. 2007b
NCAG13	NC06BGTAG13	Saluda *3/PI 427442	T. timopheevii subsp. armeniacum	AAGG	Not known	R	Murphy et al. 2007
MlUM15	NC09BGTUM15	Saluda *3/TTCC 223	A. neglecta	UUMM	7AL	R	Worthington et al. 2014

^a Used only in 2013: Pm16 and Pm36. Used only in 2014: Pm3f, Pm25, Pm34, Pm35, Pm37, NCA6, and NCAG13. Other genes were used both years.

^b 5BIL-29 = durum wheat and other cultivars = T. aestivum.

^c Chromosome.

^d North Carolina field efficacy. Data from United States Department of Agriculture–Agricultural Research Service wheat powdery mildew differential nursery, Raleigh, NC. R = resistant, MS = moderately susceptible, S = susceptible, and S/PR = susceptibility in years with heavy mildew epidemics but partial resistance in other years.

View as image HTML

Ten days postinoculation, each leaf segment was rated for B. graminis f. sp. tritici growth and sporulation using a previously described 0-to-9 rating scale, where 0 = no symptoms and 9 = entire leaf segment covered in large pustules (Parks et al. 2008). Numerical ratings were then converted into three categories: 0 to 3 = resistant, 4 to 6 = intermediate, and 7 to 9 = susceptible.

Data analyses.

Fields were grouped into regions according to geography and previous data on population subdivision (Cowger et al. 2016; Parks et al. 2009), and by observation of trends in the current dataset. For example, it was clear that the Arkansas-Missouri (AR-MO) population derived from soft wheats possessed different virulence characteristics than the adjacent Plains population derived from hard wheats (details in Results section below) and, therefore, these two populations were separated for virulence frequency calculations. The five regions were the Plains, AR-MO, Great Lakes, Mid-Atlantic, and Southeast (Fig. 1).

Virulence frequencies were calculated by combining the intermediate and susceptible categories to create a binary virulence or avirulence designation. For each Pm gene, regions were determined to be the same or different in virulence frequency using a Fisher’s exact test.

Virulence complexity.

To evaluate virulence complexity (VC), or the degree to which isolates could overcome multiple Pm genes, the analysis was restricted to the 12 genes that were tested against both years’ collections: Pm1a, 1b, 2, 3a, 3b, 4a, 4b, 6, 8, 17, MlAG12 and MlUM15. For each isolate, VC was defined as (number of genes to which the isolate was virulent/12 genes tested) × 100.

To visualize the way that VC was distributed geographically, contour plots were created from the predictions of a quadratic regression model of the form VC = Intercept + β₁ ⋅ LAT + β₂ ⋅ LON + β₃ ⋅ LAT² + β₄ ⋅ LON² + β₅ ⋅ LAT ⋅ LON, where LAT = latitude, LON = longitude, and β₁ to β₅ are regression parameter estimates. Isolates derived from the same plant had the same latitude and longitude; therefore, the VC for the location of plant 1 was the average VC for all isolates derived from that plant. The regression equations were fitted using PROC REG in SAS (version 9.3; SAS Institute Inc., Cary, NC). The contours were then plotted on the U.S. map using ArcMap (v. 10.2; Esri, Redlands, CA).

Pathotype richness or evenness and spatial aggregation.

The sampling protocol allowed us to assess the degree of spatial aggregation of pathotypes within fields (i.e., among sites) and among fields in order to make inferences about within-season epidemic origin and spread. We expected availability and abundance of initial inoculum to influence richness (the number of different pathotypes in a sample) and evenness (the degree to which pathotypes are equally present in a sample). In a field, an epidemic started by a relatively small number of pathotypes should yield a sample with lower pathotype richness than one started by a large number of pathotypes. In addition, an epidemic started by a given number of pathotypes could lead to either a well-mixed sample in which pathotypes were at relatively even frequencies across sampling sites in the field or one in which pathotypes were associated with site. Our hypothesis was that fields would differ in the degree of pathotype aggregation by site because some would have an initially less abundant or less well-mixed initial pool of inoculum. Of course, relative fitness of pathotypes would affect richness and evenness at the time of sampling. Pm genes in the host crop were not expected to play a large role in selection of pathotypes, given that mainly cultivars without effective Pm genes were sampled.

For these analyses, two isolates were assumed to have the same pathotype if they had the same profile of virulence or avirulence to all 12 Pm genes. The pathotype of each isolate was represented with concatenated 1s and 0s, where 1 indicated virulence to a Pm gene and 0 indicated avirulence, using the order of Pm genes given in Table 2.

The per-field richness and evenness of pathotypes were evaluated by calculating Shannon’s diversity index (H′) (Shannon and Weaver 1949), Simpson’s index (Simpson 1949), and Hill’s evenness index (Hill 1973). Rarefied species richness was computed using the function rarefy in the package vegan of R (v. 3.3.3).

The degree of per-site similarity of pathotype occurrence was assessed using the R package “cooccur” (Griffith et al. 2016), which provided a probabilistic measure of sites sharing a common pathotype, and may be interpreted as a similarity measurement for sites sharing one or more pathotypes. A low co-occurrence probability would indicate that a pair of sites had a low number of common pathotypes. Arithmetic averages were used to produce a single co-occurrence probability across sites with the field. The maximum level of similarity among five sites in a hypothetical field with five pathotypes would be 1.0 if each site had frequency = 0.2 for each pathotype. Only fields with five sampled sites were considered, which excluded six fields (Table 3).

TABLE 3. Number and diversity of Blumeria graminis f. sp. tritici pathotypes sampled from 34 U.S. wheat fields in 2013 and 2014

				Number of					Pathotype		Mean Sorensona
Field	Abbrevb	Regionc	Year	Sitesd	Isolates	Pathe	DIf	SI (λ)g	Richness (SE)h	Evenness	Distance	CI	Cooccuri
25	GAT	SE	2014	5	26	3	0.32	0.86	1.77 (0.67)	0.85	0.24	0.20–0.25	0.21
21	ALA	SE	2014	5	49	3	0.72	0.57	2.33 (0.51)	0.86	0.15	0.08–0.25	0.27
4	GAP	SE	2013	5	52	7	1.16	0.46	3.49 (0.94)	0.68	0.37	0.28–0.47	0.23
6	MIR	GL	2013	5	22	5	1.39	0.29	4.18 (0.65)	0.85	0.42	0.30–0.58	0.27
5	MID	GL	2013	4	18	5	1.52	0.23	4.48 (0.54)	0.96	0.43	0.23–0.62	0.35
29	NCCL	MA	2014	5	45	11	1.71	0.30	4.93 (1.16)	0.61	0.50	0.41–0.58	0.21
12	NCF	MA	2013	5	36	9	1.74	0.24	5.05 (1.02)	0.73	0.45	0.35–0.58	0.19
23	FLG	SE	2014	4	26	8	1.75	0.21	5.18 (0.96)	0.82	0.46	0.32–0.62	0.21
26	KSM	PL	2014	5	22	9	1.70	0.29	5.42 (1.07)	0.64	0.65	0.59–0.79	0.16
24	GAS	SE	2014	5	32	11	1.97	0.21	5.87 (1.13)	0.66	0.62	0.53–0.74	0.15
32	NYB(14)	GL	2014	5	29	9	1.98	0.16	5.95 (0.99)	0.85	0.51	0.40–0.64	0.20
16	OKH	PL	2013	2	19	10	1.98	0.19	6.39 (1.04)	0.72	0.63	0.54–0.82	0.36
3	GAM	SE	2013	5	39	13	2.23	0.14	6.47 (1.14)	0.75	0.65	0.56–0.74	0.09
20	VAS	MA	2013	5	14	8	1.91	0.17	6.49 (0.82)	0.85	0.75	0.63–0.87	0.06
7	MOB	AR-MO	2013	5	31	13	2.22	0.15	6.58 (1.14)	0.74	0.69	0.58–0.79	0.11
15	OHJ	GL	2013	5	28	14	2.26	0.15	6.70 (1.20)	0.70	0.74	0.61–0.85	0.08
9	MSG	SE	2013	5	27	11	2.21	0.13	6.77 (1.02)	0.86	0.63	0.51–0.75	0.13
8	MSP2	SE	2013	5	34	13	2.30	0.13	6.80 (1.11)	0.80	0.63	0.52–0.73	0.11
28	MOB(14)	AR-MO	2014	5	15	9	1.99	0.17	6.81 (0.88)	0.79	0.75	0.59–0.85	0.10
30	NCM	MA	2014	5	23	11	2.21	0.13	6.90 (1.02)	0.87	0.69	0.58–0.80	0.13
35	OKS(14)	PL	2014	3	10	7	1.83	0.18	7.00 (0.00)	0.89	0.80	0.58–1.00	0.16
18	PAC	GL	2013	5	33	15	2.37	0.13	7.00 (1.16)	0.69	0.68	0.61–0.77	0.11
14	NYB	GL	2013	5	22	11	2.22	0.12	7.02 (1.00)	0.87	0.75	0.65–0.84	0.11
19	PAF	GL	2013	5	37	16	2.44	0.12	7.07 (1.16)	0.72	0.66	0.57–0.75	0.10
36	PAF(14)	GL	2014	5	41	14	2.42	0.10	7.10 (1.07)	0.86	0.57	0.47–0.68	0.19
33	NYG	GL	2014	5	54	19	2.57	0.10	7.17 (1.17)	0.74	0.55	0.45–0.65	0.16
34	OHW	GL	2014	5	26	15	2.41	0.12	7.35 (1.14)	0.72	0.81	0.67–0.91	0.06
10	MSP1	SE	2013	5	37	17	2.53	0.11	7.35 (1.16)	0.73	0.70	0.61–0.79	0.08
17	OKS	PL	2013	3	11	8	2.02	0.14	7.55 (0.50)	0.94	0.81	0.72–1.00	0.11
13	NYA	GL	2013	5	12	9	2.09	0.14	7.82 (0.65)	0.89	0.87	0.82–0.95	0.07
2	GAD	SE	2013	5	14	10	2.21	0.12	7.86 (0.80)	0.90	0.88	0.79–0.95	0.06
11	NCC	MA	2013	5	50	24	2.88	0.08	7.99 (1.12)	0.74	0.66	0.60–0.77	0.12
27	MIR(14)	GL	2014	4	10	9	2.16	0.12	9.00 (0.00)	0.96	0.95	0.89–1.00	0.07
1	PRA	SE	2013	5	11	10	2.27	0.11	9.18 (0.39)	0.96	0.96	0.89–1.00	0.06

^a Uniqueness of pathotype composition at sites within fields; smaller values indicated that sites had more distinct sets of pathotypes. CI = confidence interval.

^b State and city abbreviation.

^c SE = Southeast, GL = Great Lakes, MA = Mid-Atlantic, PL = Plains, and AR-MO = Arkansas and Missouri.

^d Sites in field.

^e Virulence or avirulence pathotype (Path) to 12 resistance (Pm) genes tested against all isolates.

^f Shannon’s diversity index.

^g Simpson index.

^h Calculated by rarefaction; numbers in parentheses are standard errors (SE).

ⁱ Mean probability of cooccurrence of pathotypes at sampling sites.

View as image HTML

As an additional measure of pathotype similarity among sites, the Sorenson distance was calculated. The average Sorenson distance per field and its 95% confidence interval were calculated with package “oecosim” within library vegan (Oksanen et al. 2017). Simulated random samples under the null assumption of independence were generated using method “r2table”, which assures fixed marginal frequencies for rows and columns for each simulated sample. Pairwise Sorensen distances (dissimilarities) were generated from each sample and their mean value was calculated. After n = 599 simulations, the simulated distribution of the mean Sorensen distance was used to calculate a mean value, and its 2.5 and 97.5 percentiles to be used as limits of the 95% confidence interval for pairwise Sorensen distances in a given field.

A larger value for Sorenson distance indicated that sampling sites tended to have distinct sets of pathotypes. A high probability of pathotype co-occurrence [prob(cooccur)] indicated that a common set of pathotypes were present at each site. Smaller Sorenson distances and higher prob(cooccur) would be expected to be associated. In general, it was thought that, if sites within a field had highly similar pathotypes, this would suggest that the epidemic was started by a set of pathotypes that either at the initiation or later became relatively evenly distributed across the field. If sites within a field were very different with regard to pathotype presence, it suggested that the epidemic was started in different parts of the field by different pathotypes that did not become evenly distributed throughout the field.

RESULTS

The frequencies of virulence found in each region in 2013 and 2014 are grouped by category below.

Genes effective everywhere.

Pm genes to which there was little to no virulence (frequencies in all regions were < 0.11) in all regional B. graminis f. sp. tritici populations were Pm1a, 1b, 16, 36, 37, 53, MlAG12, NCAG13, and MlUM15 (Tables 4 and 5). Single-gene differential lines bearing those genes are also resistant in North Carolina field efficacy studies (Table 2). These genes are hypothesized to be broadly effective throughout the U.S. range of B. graminis f. sp. tritici. Virulence frequencies were not significantly different among regions in either year for Pm1a, 1b, MlAG12, or MlUM15 (P ≥ 0.17); in 2013 for Pm16 (P = 0.30) or Pm36 (no virulence anywhere); or in 2014 for Pm37, Pm53, or NCAG13 (P ≥ 0.14) (Tables 4 and 5).

TABLE 4. Proportion of isolates collected from five geographic regions in the eastern United States in 2013 that were virulent to wheat powdery mildew resistance (Pm) genes^a

	Great Lakes	Arkansas-Missouri	Mid-Atlantic	Plains	Southeast
Genes	n = 175–177	n = 34–37	n = 101–103	n = 31–33	n = 222–234	P valueb
Pm1a	0.000	0.000	0.000	0.000	0.000	…
Pm1b	0.000	0.000	0.000	0.000	0.000	…
Pm2	0.631	0.162	0.353	0.182	0.506	0.0000
Pm3a	1.000	1.000	1.000	0.121	0.996	0.0000
Pm3b	0.937	0.514 (0.229)	0.961	0.000	0.802	0.0000
Pm4a	0.791	0.806	0.612	0.303	0.768	0.0000
Pm4b	0.223	0.294 (0.294)	0.257	0.161	0.216	0.6760
Pm6	0.994	1.000	1.000 (0.204)	1.000 (0.219)	1.000	0.5986
Pm8	0.768 (0.232)	0.081	0.709 (0.340)	0.879 (0.242)	0.629 (0.249)	0.0000
Pm16	0.000	0.000	0.010	0.000	0.000	0.2962
Pm17	0.887 (0.667)	0.865 (0.622)	0.903 (0.621)	0.727 (0.606)	0.765 (0.598)	0.0012
Pm36	0.000	0.000	0.000	0.000	0.000	…
North Carolina introgressions
MlAG12	0.000	0.000	0.010	0.000	0.013	0.6065
MlUM15	0.000	0.000	0.000	0.000	0.000	…
Pm53	0.000 (n = 67)	0.108 (n = 37)	0.000 (n = 60)	0.000 (n = 33)	0.000 (n = 213)	0.0001

^a Number of isolates tested (n) from each region in 2013. Numbers for Pm53 are exceptions, as shown. Values are combined susceptible and intermediate proportions; parenthetical values are intermediate proportions ≥ 0.20; absence of parenthetical values signifies intermediate proportion < 0.20.

^b Fisher’s exact test P values; ≤ 0.05 indicates at least one population has significantly different virulence frequencies; missing values indicate identical populations where tests could not be performed.

View as image HTML

TABLE 5. Proportion of isolates collected from five geographic regions in the eastern United States in 2014 that were virulent to wheat powdery mildew resistance (Pm) genes^a

	Great Lakes	Arkansas-Missouri	Mid-Atlantic	Plains	Southeast
Genes	n = 160	n = 29	n = 67–68	n = 43	n = 133	P valueb
Pm1a	0.006	0.000	0.000	0.000	0.000	1.0000
Pm1b	0.000	0.000	0.000	0.000	0.000	…
Pm2	0.581	0.138	0.662	0.209	0.481	0.0000
Pm3a	0.981	0.897	1.000	0.140	0.992	0.0000
Pm3b	0.813	0.690	0.971	0.116	0.955	0.0000
Pm3f	0.987	1.000	1.000	1.000 (0.267)	1.000	1.0000
Pm4a	0.738	0.897	0.603	0.279	0.887	0.0000
Pm4b	0.150	0.241	0.191	0.140	0.075	0.0486
Pm6	0.988	0.966 (0.276)	0.985 (0.235)	0.977 (0.419)	1.000	0.2052
Pm8	0.844 (0.213)	0.759 (0.621)	0.971 (0.206)	0.953 (0.256)	0.985	0.0000
Pm17	0.944 (0.594)	0.897 (0.448)	0.956 (0.603)	0.930 (0.605)	0.977 (0.594)	0.2436
North Carolina introgressions
Pm25	0.750 (0.519)	0.241 (0.241)	0.721 (0.618)	0.070	0.970 (0.519)	0.0000
Pm34	0.763 (0.550)	0.724 (0.448)	0.868 (0.632)	0.116	0.835 (0.586)	0.0000
Pm35	0.925 (0.363)	0.793 (0.586)	0.662 (0.559)	0.535 (0.465)	0.962 (0.729)	0.0000
Pm37	0.019	0.000	0.000	0.000	0.015	0.8899
NCA6	0.894 (0.769)	0.828 (0.621)	0.971 (0.912)	0.093	0.962 (0.722)	0.0000
MlAG12	0.006	0.000	0.000	0.000	0.000	1.0000
NCAG13	0.000	0.000	0.000	0.000	0.008	0.6305
MlUM15	0.000	0.000	0.000	0.023	0.000	0.1667
Pm53	0.000 (n = 136)	0.0036 (n = 28)	0.016 (n = 62)	0.000 (n = 38)	0.008 (n = 123)	0.1378

^a Number of isolates tested (n) from each region in 2014. Numbers for Pm53 are exceptions, as shown. Another exception was Pm3f, with sample sizes of Great Lakes n = 75, AR-MO n = 18, Mid-Atlantic n = 23, Plains n = 30, and Southeast n = 47. Values are combined susceptible and intermediate proportions; parenthetical values are intermediate proportions ≥ 0.20; absence of parenthetical values signifies intermediate proportion < 0.20.

^b Fisher’s exact test P values; ≤ 0.05 indicates at least one population has significantly different virulence frequencies; missing values indicate identical populations where tests could not be performed.

View as image HTML

Genes largely or entirely defeated.

Genes that were defeated everywhere (virulence frequencies in all regions were ≥ 0.96) were Pm3f and Pm6 (Tables 4 and 5). Those genes were also mostly or entirely defeated by North Carolina field populations, although Pm6 retained some efficacy under lighter disease pressure (Table 2). Virulence frequencies were not significantly different among regions in 2014 for Pm3f (P = 1.0) or in either year for Pm6 (P ≥ 0.21). In the case of Pm6, the intermediate (not fully virulent) proportion of isolates in the 2 years was 0.20 to 0.42 of all isolates in the Mid-Atlantic and Plains regions and 0.14 to 0.28 in AR-MO (Tables 4 and 5 and data not shown).

For Pm8, all populations were virulent except the 2013 population from AR-MO, which was avirulent (Tables 4 and 5). Regional frequencies differed for Pm17 virulence in 2013 (P = 0.001), with virulence being higher in the Mid-Atlantic and lower in the Plains, but they did not differ in 2014 (P = 0.24). For both Pm8 and Pm17 in every region, a significant fraction of the virulent isolates produced an intermediate reaction (Tables 4 and 5). For Pm8, the intermediate isolates were 0.21 to 0.34 of the total, except in 2014, when they were 0.62 in AR-MO. For Pm17, the intermediate proportion was 0.45 to 0.67 of the total. Under North Carolina field conditions, Pm8 confers partial resistance to B. graminis f. sp. tritici, collapsing under heavy mildew pressure, whereas Pm17 is completely defeated by those field populations (Table 2).

Pm4b.

Among the Pm genes tested, this gene was unique in that B. graminis f. sp. tritici populations from all regions in both years exhibited low to moderate frequencies (0.08 to 0.29) of virulence to it (Tables 4 and 5). In the field in North Carolina, this gene confers partial resistance that breaks down under high powdery mildew pressure (Table 2).

Genes effective in field but not in lab.

An unusual pattern was observed with four of the genes recently introgressed from wild wheat relatives: Pm25, Pm34, Pm35, and NCA6. They conferred resistance to North Carolina populations of B. graminis f. sp. tritici in the field (Table 2) but failed in the laboratory detached-leaf experiments against most isolates from every U.S. region except the Plains (Table 5). By contrast, as described above, the other genes recently introgressed from wild wheat relatives (Pm37, Pm53, MlAG12, NCAG13, and MlUM15) were effective both against North Carolina B. graminis f. sp. tritici field populations and in detached-leaf tests against isolates from all regions. Pm25, Pm34, Pm35, and NCA6 are all from diploid donors Triticum monococcum and Aegilops tauschii (Table 2), although the connection if any is unclear.

Most AR-MO isolates were avirulent to Pm25, like the Plains population, but displayed intermediate virulence to Pm34, Pm35, and NCA6, like isolates from other soft-wheat regions. With all four of these mildew resistance genes, in the detached-leaf experiments, intermediate responses were produced by substantial percentages (0.24 to 0.91, usually >0.50) of the isolates from the soft-wheat regions of the Mid-Atlantic, Great Lakes, Southeast, and AR-MO (Table 5). With Pm25, AR-MO isolates were 0.76 avirulent and 0.24 intermediate; there were no completely virulent responses. By contrast, Plains isolates only exhibited a large proportion (0.47) of intermediate responses to one of the four genes, Pm35, and were overwhelmingly avirulent to the others.

Older genes with regionally specific effectiveness.

In both years, the majority of isolates from the Plains hard-wheat region were avirulent to several genes that have long been mostly defeated in other regions: Pm2, Pm3a, Pm3b, and Pm4a (Tables 4 and 5). This indicates that those genes are likely to remain effective in the southern portion of the U.S. hard-wheat region, even though they are ineffective elsewhere. It suggests that Pm2, Pm3a, Pm3b, and Pm4a have not been widely deployed in Plains commercial hard wheat. The same regionally specific avirulence was observed for isolates from the AR-MO soft-wheat region with respect to Pm2 in both years. Pm2 appears to retain at least partial effectiveness in Arkansas and Missouri, as well as the Plains.

VC.

B. graminis f. sp. tritici populations from both years had a median VC of six genes per isolate (Fig. 2). In 2013 and 2014, isolates with 5 to 7 virulences out of the 12 tested Pm genes accounted for approximately 73 and 83%, respectively, of all isolates tested. The VC of the 2014 population skewed higher than the 2013 population, with a larger percentage of 2014 isolates virulent to 7 of the 12 genes.

Download as PowerPoint

From regression, the highest predicted VC in both years was east of the Mississippi River in the soft-wheat region, and predicted VC diminished going west (Fig. 3). In 2013, the highest-VC area was in the Great Lakes region, particularly Illinois, Indiana, Wisconsin, and Kentucky (Fig. 3A). In 2014, the longitude of the highest predicted-VC region remained the same but the center of that region had shifted south to encompass the southern parts of Indiana and Ohio and the northern portions of Alabama and Georgia (Fig. 3B). Averaging the 2 years together, mean VC ranged from 30% in the Plains to 50 to 51% in the Great Lakes, Mid-Atlantic, and Southeast regions.

Download as PowerPoint

Pathotype diversity and spatial aggregation.

The number of different pathotypes in the sample was 64 in 2013 and 50 in 2014. The number of pathotypes sampled per field varied from a low of 3 to a high of 24 (Table 3). Measures of pathotype richness and diversity showed a considerable range. For example, Shannon’s diversity index varied from 0.32 to 2.57. Pathotype richness, which takes sample size into account, varied from 1.8 to 9.2. Considering the simple average across fields, mean pathotype richness for each region was 6.3 to 6.7, except for the Southeast, which was 5.7. Three Southeast fields had notably low pathotype diversity (richness 1.8 to 3.5). Leaving aside the two AR-MO fields, the other four regions were all represented in both the top and bottom 10 fields for both Shannon’s diversity and pathotype richness. This suggested that in any region, a field-level epidemic could be started by relatively few or many pathotypes (e.g., some epidemics started from a small pool of genetic individuals and others from a large pool).

Fields also varied significantly in the extent to which sampling sites within them gave evidence of pathotype distribution, as measured by Sorenson’s distance and the prob(cooccur) (Table 3). Sorenson’s distance, a measure of pathotype specificity by site, varied from 0.15 (sites tended to have unique sets of pathotypes) to 0.96 (sites tended to have common sets of pathotypes), with higher values tending to be found in fields with greater pathotype richness. Prob(cooccur) varied from a low of 0.06 to a high of 0.36; thus, the upper limit was not high in absolute terms. When these two measures of pathotype specificity by site within field were plotted against pathotype richness, a common pattern was observed (Fig. 4). When pathotype richness was low, Sorenson distance was low and prob(cooccur) was high, indicating that sets of specific pathotypes tended to be nonrandomly associated with certain sites in the field. When pathotype richness was high, Sorenson distance was high and prob(cooccur) was low, indicating that sites did not share common pathotypes. For intermediate levels of pathotype richness, both of the pathotype aggregation measures had a wider range, meaning that, in some fields, pathotypes were less specifically associated with site and in other fields more so.

Download as PowerPoint

Taking pathotype diversity and spatial aggregation measures into account, it appeared that some epidemics were started with larger, more well-mixed populations of pathotypes and, thus, provided a high degree of pathotype diversity in which pathotypes were not clustered or aggregated within specific parts of the field. Other epidemics began with a smaller number of pathotypes, and there was more pathotype clustering or aggregation within the field. Neither scenario appeared specific to geographic region (Table 3).

DISCUSSION

Virulence surveys of the U.S. B. graminis f. sp. tritici population have now occurred three times at approximately one-decade intervals, and have highlighted evolution of the population in response to certain genes. This survey is the largest to date in number of isolates tested, although the 1993-to-1994 survey (Niewoehner and Leath 1998) covered a similar geographic range. It was significantly larger than the 2003-to 2005 survey (Parks et al. 2008) due to the inclusion of locations in the Great Lakes and Plains region. The present survey is also the first to utilize a hierarchical sampling scheme to illuminate within-field distribution of pathotype diversity.

Virulence of B. graminis f. sp. tritici populations has also been recently characterized in Australia (Golzar et al. 2016), China (Zeng et al. 2014), and Egypt (Abdelrhim et al. in press). Virulence frequencies varied for some genes; for instance, Pm1a was effective in the United States but ineffective in Australia, China, and parts of Egypt. Common to all four countries, however, was that the fungal populations were virulent to many or most older Pm genes tested, suggesting the need to carefully utilize those Pm genes that are commercially deployed and still effective, as well as new mildew resistance genes that are introgressed into wheat.

We hypothesized there would be differences in the virulence frequencies of the B. graminis f. sp. tritici populations from different U.S. regions. What we did not anticipate was the clear differences the survey would illuminate between the B. graminis f. sp. tritici population of the Plains hard-wheat region and those of the soft-wheat regions of the eastern United States. The Plains population distinguished itself by having low virulence or avirulence to Pm2, Pm3a, Pm3b, and Pm4a, all of which were defeated in the Mid-Atlantic and Southeast regions by 2003 (Parks et al. 2008). It also had a greater intermediate percent virulence to Pm3f and Pm6 than isolates from other regions. The Plains B. graminis f. sp. tritici population was also more sensitive to triazole fungicides than the population originating on soft wheat (unpublished data). The maintenance of this distinct hard-wheat pathogen population is possible because of the U.S. B. graminis f. sp. tritici migration pattern, which features west-to-east but not east-to-west migration (Cowger et al. 2016).

Despite their geographic proximity, the Plains hard-wheat B. graminis f. sp. tritici population and the soft-wheat population west of the Mississippi River (AR-MO population) exhibited significant pathotypical differences. The AR-MO population generally behaved like the eastern soft-wheat population but had low virulence to Pm2, like the Plains population. Unique among the regions, AR-MO had low to intermediate virulence to Pm8.

Overall, this pattern suggests that the selective influence of genes deployed in hard- versus soft-wheat germplasm has affected virulence frequencies differentially, even in adjacent geographies. It also supports the model proposed for this pathogen population (Cowger et al. 2016), in which isolation by distance operates over relatively small geographic scales and wheat powdery mildew epidemics are rekindled annually from local sources.

Our results indicated that VC was highest in the east and diminished going west. Even within the soft-wheat region, VC declined in the soft-wheat states west of the Mississippi (Missouri and Arkansas). Although, in 2013, VC was highest in the Great Lakes, in 2014, the “center of complexity” shifted to the southeast, and VC skewed higher when averaged across the entire sampled area. Although the current study did not include B. graminis f. sp. tritici isolates from Kentucky or Indiana, according to our predictions, those states should have among the highest-VC populations in the United States. In fact, in spring 2016, eight single-pustuled B. graminis f. sp. tritici isolates from Lexington, KY were derived, purified, and virulence tested in the same manner as in the present study. For the 13 Pm genes used in the present study, the 2016 Kentucky isolates had a mean VC of 56% (data not shown). The isolates all had distinct multilocus virulence profiles, indicating they were not clones. This level of VC conforms to the predictions for Kentucky in Figure 3A and B. Interestingly, wheat powdery mildew epidemics are rare in Kentucky (D. Van Sanford, personal communication).

In the same vein, a sample was also obtained in spring 2016 from a spring wheat field near Geraldine, MT, which is 1,300 to 1,600 km northwest of the 2013 and 2014 Plains locations and, thus, geographically well outside the present sampling area. For the 13 Pm genes used in the present study, a set of 17 Montana isolates had a mean VC of 36% (data not shown), which is similar to the levels observed in the Plains locations in the current study (Fig. 3A and B). This suggested that the trend toward lower VC in the Plains may hold constant in B. graminis f. sp. tritici populations much farther to the west. Of course, that would need to be tested with a larger sample. Furthermore, the Montana isolates exhibited the same pattern (data not shown) of avirulence to Pm3a and Pm3b but not to Pm2 or Pm4a that distinguished the Plains population from the eastern-U.S. population in the present study.

These observations support the characterization of a B. graminis f. sp. tritici population that has evolved to maximum VC in the eastern United States, and remains avirulent in the central and perhaps also the western United States to certain Pm genes that are defeated farther east. It also supports the findings of Cowger et al. (2016) that the U.S. B. graminis f. sp. tritici population is not random mating across its full geographical extent but, rather, is geographically subdivided, and lacks westward migration out of the eastern soft-wheat population.

The study also revealed a curious pattern in which certain Pm genes recently introgressed from diploid wild wheat relatives were ineffective against soft-wheat B. graminis f. sp. tritici isolates in tests on detached seedling leaves, yet effective against a soft-wheat field population of the pathogen. With three of these four genes, the Plains B. graminis f. sp. tritici population originating from hard wheat did not evoke this anomaly. In the case of the Kentucky and Montana isolates from 2016 mentioned above, Kentucky isolates performed similarly to Mid-Atlantic isolates on Pm25, Pm34, Pm35, and NCA6, with high levels of intermediate and full virulence (≤0.75, data not shown) in the detached-leaf tests. The Montana isolates performed just as the 2013-to-2014 Plains isolates against Pm 25, Pm34, and NCA6 but displayed universal virulence to Pm35 in detached leaves (56% of isolates were intermediate and 44% fully virulent).

The reasons for the partial susceptibility of these four recently introgressed genes in the laboratory detached-leaf setting are unclear. Pm17 also exhibited high levels of intermediate responses but this gene is defeated in the Mid-Atlantic and Southeast and perhaps in other regions as well. It has been deployed in several regions, including the Plains (for greenbug resistance), the Southeast, and the Mid-Atlantic. Its defeat in commercial deployment was widely observed in the years 2009 to 2012. By contrast, Pm25, Pm34, Pm35, and NCA6 have not yet been broadly utilized in commercial production. These genes were introgressed from T. monococcum subsp. aegilopoides (AA), A. tauschii subsp. tauschii (DD), and A. tauschii subsp. strangulata (DD). The Pm genes recently introgressed from wild relatives that were consistently resistant not only under North Carolina field conditions but also against all isolates tested on detached leaves (Pm37, MlAG12, NCAG13, and MlUM15) were introgressed from tetraploid donors T. timopheevii (AAGG) and A. neglecta (UUMM) and from diploid A. speltoides (SS). However, it is not known whether these varying origins account for the different performance of the resistance sources.

The inconsistency between detached-leaf and field results in the performance of recently introgressed genes Pm25, Pm34, Pm35, and NCA6 illustrates the importance of maintaining field nurseries of single-gene differential lines to complement laboratory tests. If those genes are used in breeding programs outside the Mid-Atlantic region, where they perform well, it would be valuable to plant their single-gene differential lines in the field for confirmation of their efficacy against other regional B. graminis f. sp. tritici populations.

The present results are the first to compare mildewed wheat fields across a broad range of U.S. growing areas with respect to the diversity of pathotypes found within them. They suggest that epidemics are started by widely varying numbers of genetic individuals: in some fields by just a few strains and in other fields by larger, well-mixed populations of strains. It is also possible that within-season competition affected the numbers of pathotypes found, with different outcomes in different fields. In our study, within-season fitness of strains was probably determined by factors other than pathotype, because samples came from susceptible crops. Previously, nearly every sampled B. graminis f. sp. tritici individual was a different genotype (Cowger et al. 2016) by simple-sequence repeat markers, if not necessarily by pathotype, and, thus, samples such as the present one are likely not the product of repeated resampling of the same bulked-up clone to any significant degree.

It should also be kept in mind that most epidemics probably do not start just once in time but through multiple inoculations, and this may have contributed to the variability in pathotype richness that was observed. Indeed, this factor may have played a role in the different levels of whole-field pathotype homogeneity that we observed (e.g., some epidemics may have been started in one part of the field by one or more specific pathotypes, different from those that initiated disease development, perhaps at another time, in other parts of the field). The fields with high levels of pathotype richness had variable degrees of pathotype homogeneity by site, suggesting that there was variability in how well mixed the initiating pathotypes were, the timing and location at which they arrived, or both.

The number of genetic individuals that initiate a B. graminis f. sp. tritici epidemic probably depends considerably on the amount of wheat stubble present, the mildew susceptibility and consequent level of disease development in nearby previous wheat crops, oversummering conditions, and weather conditions at epidemic onset. The fact that mean pathotype richness was lower in the Southeast than in the other sampled regions suggests that the B. graminis f. sp. tritici population in that region undergoes genetic bottlenecks, perhaps due to patchier wheat cultivation or the greater difficulties of oversummering in a hotter climate. This would result in fewer virulence haplotypes. It is also possible that fewer Pm genes have historically been deployed in the Southeast, where powdery mildew has not generally been a significant problem in wheat production.

In conclusion, the survey highlighted genes that should be useful for mildew resistance breeding in all geographic regions of the United States where mildew is present. Among the older genes, Pm1a and Pm1b were universally effective, and another gene, Pm4b, also retained at least some efficacy across the sampled area. The research also confirmed that older genes defeated in the eastern soft-wheat region such as Pm2, Pm3a, Pm3b, and Pm4a continue to retain efficacy in the Plains environment, likely due to differing Pm gene selection pressures and west-to-east migration.

Furthermore, several recently introgressed mildew resistance genes from wild wheat relatives (Pm37, Pm53, MlAG12, NCAG13, and MlUM15) exhibited full resistance to all local B. graminis f. sp. tritici populations from across the sampled area. Pm25, Pm34, Pm35, and NCA6 are also likely to be effective over a wide geographic area, despite the odd results from their detached-leaf tests. These new Pm genes, which are not yet in widespread commercial use, should have greater durability if they are commercially deployed in stacks or in genetic backgrounds with high levels of quantitative resistance (Burdon et al. 2014; Quenouille et al. 2013). It is recommended to use these genes in that manner, given the difficulty of locating new sources of resistance and the propensity of B. graminis f. sp. tritici to rapidly overcome Pm genes. Germplasm is available from the authors upon request.

ACKNOWLEDGMENTS

We thank E. T. Cole, M. Hargrove, R. Parks, and R. Whetten for virulence assays and assistance with data; and collaborators B. Allen, T. Allen, G. Bergstrom, A. Blount, B. Bockus, K. Bowen, A. Collins, K. Davies, J. DeDecker, R. Ethredge, T. Gafnea, B. Hunger, J. Kichler, G. Milus, D. Moore, M. Nagelkirk, B. Olson, P. Paul, G. Roth, A. Sawyer, M. Stanyard, L. Sweets, B. Verbeten, and S. Wegulo for providing mildew samples from the field.