crop science, vol. 54, january–february 2014 www.crops.org 1

ReseaRch

The existence of substantial genotypic variation for Fe and Zn is required for increasing Fe and Zn concentration of cereal grain

by plant breeding (Cakmak et al., 2004). Many researchers have reported genotypic variation in wheat seed for Fe and Zn (Peterson et al., 1986; Graham et al., 1999; Rengel et al., 1999; Cakmak et al., 2002; Velu et al., 2012). Genotypic variations reported in cultivated genotypes are greater and it is highly possible to develop Zn- and Fe-enriched wheat varieties with high yields (Velu et al., 2012).

Inductively coupled plasma–optical emission spectrometry (ICP–OES), atomic absorption spectrometry, and, more recently, X-ray fluorescence spectrometry (Paltridge et al., 2012a, 2012b) analysis of plant tissues for nutrient concentration are essential technologies underpinning many plant nutrition and biofortifi-cation studies. Within the HarvestPlus biofortification network, the main nutrients of interest are Fe and Zn, and the tissues of primary interest are mature grains. Both Fe and Zn are micronu-trients, present in grain at very low levels (generally between 10 and 50 mg kg-1 [Paltridge et al., 2012a, 2012b]). Major seed crops of interest in biofortification include wheat, rice (Oryza sativa

Measuring Genotypic Variation in Wheat Seed Iron First Requires Stringent Protocols

to Minimize Soil Iron Contamination

Zarina Yasmin, Nicholas Paltridge, Robin Graham, Bao-Lam Huynh and James Stangoulis*

AbstrActMeasuring genotypic variation in Fe is an impor-tant task for plant breeders trying to biofortify wheat (Triticum aestivum L.), but levels of Fe are often enhanced in wheat seed due to soil derived contamination. Seed Al levels can be used to identify genotypes that are con-taminated with Fe. The sources of contamina-tion were identified and removed by milling the wheat grains in a modified Kett mill for 5 s before analysis. Thirty contaminated lines from Australia and 240 contaminated lines from Paki-stan were cleaned for the validation of the 5 s milling method. Before cleaning, average Al concentrations were 10 mg kg-1. After cleaning, 183 out of the 240, or 76%, had Al less than 5.0 mg kg-1, and average Al levels were 4.2 mg kg-1. Therefore, the cleaning procedure was highly successful in reducing Al levels in contaminated samples. The ratio between Al removed and Fe removed was fairly consistent and followed a 1:1 relationship across two environments. Twenty genotypes of a wheat population were evalu-ated for their Fe concentration before and after cleaning and ranking changed significantly and cleaning also improved the precision of quanti-tative trait loci (QTL) analysis, with QTL QGFe.ta-3B having a logarithm of the odds score of 1.1 before cleaning, which rose to 3.4 after clean-ing. The method provides a valuable cleaning procedure for plant breeders wanting to ana-lyze wheat seed and also provides evidence for reducing the influence of soil Fe contamination on seed by simply adjusting the Fe concentra-tions based on the level of Al in the grain.

Z. Yasmin, N. Paltridge, R. Graham, and J. Stangoulis, School of Biological Science, Flinders Univ., Bedford Park, SA 5042; and B.-L. Huynh, Dep. of Nematology, Univ. of California–Riverside, River-side, CA 9252. Received 3 Apr. 2013. *Corresponding author ( james.stangoulis@flinders.edu.au).

Abbreviations: Cat, category; DH, double haploid; HPAN, Harvest-Plus Advance Nursery; ICP–OES, inductively coupled plasma – optical emission spectrometry; LOD, logarithm of the odds; NIST, National Institute of Standards and Technology; QTL, quantitative trait locus/loci; RAC, Roseworthy Agricultural College.

Published in Crop Sci. 54:1–10 (2014). doi: 10.2135/cropsci2013.04.0214 Freely available online through the author-supported open-access option. © Crop Science Society of America | 5585 Guilford Rd., Madison, WI 53711 USA

All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher.

2 www.crops.org crop science, vol. 54, january–february 2014

L.), maize (Zea mays L.), pearl millet [Pennisetum glaucum (L.) R. Br.], and bean (Phaseolus vulgaris L.) (Nestel et al., 2006). For wheat grain, Fe and Zn are present at very low levels (generally between 10 and 50 mg kg-1) and a major obstacle in wheat breeding is that analysis of Fe and Zn in grain is often confounded by the presence of soil derived contamination particularly for Fe rather than Zn (Pfeiffer and McClafferty, 2007; Sillänpaa, 1982).

Soil-borne micronutrients adhere to the outer epider-mal surfaces (Djingova and Kuleff, 1994; Hall, 1995; Jones, 2001), which leads to the contamination. Soil typically con-tains around 4% Fe (40,000 mg kg-1) (Sparks, 2003), several orders of magnitude more than is found in grain, so traces of soil on grain lead to erroneously high reports of grain Fe concentration. Soil contamination is also mostly considered a problem for the plants grown and collected from windy areas (Cherney and Robinson, 1983; Jones, 1991, 2001).

Other common sources of Fe and Zn contamination are residues from grain handling equipment or threshers and residues from human hands or even gloves used during the preparation of seed samples (Sager and Mittendorfer, 1997; Jones, 2001). There are two main methods used to identify Fe and Zn contamination in wheat. First, major sources of contamination may be seen as concentration “spikes” in the analysis, which are generally not reproduced in replicated analysis. Second, since Al is even more abun-dant in soil than Fe (Ibia, 2002; Rauch and Pacyna, 2009) but is not found in clean plant tissue and is easily analyzed by ICP–OES, it can be used as an indicator of soil contami-nation. Typically, soil contamination of grain is considered to be acceptably low in grain analysis reports where Al is present at <5 mg kg-1 (Pfeiffer and McClafferty, 2007).

There is very little information available in the litera-ture as to how contaminated wheat grain can be cleaned of contamination before analysis. In one unpublished study, attempts were made to remove contamination by manual wiping of the seeds with Chux cloth (Clorox Australia Pty. Ltd.) ( J. Stangoulis, unpublished data, 2012); how-ever, these methods were only partially successful, with Al levels on grain contaminated with soil only reduced to 5 to 10 mg kg-1, even after wiping for 90 s, and the method was considered too labor intensive to pursue further.

Washing experiments of wheat and maize grains have been conducted to remove the Fe contamination and have been successful to some extent although there is concern that washing of samples can leach out the micronutrients and this then distorts results further (Djingova and Kuleff, 1994; Hall, 1995; Jones, 2001).

The Kett “Pearlest” Polisher (Kett mill) (Kett Electric Laboratory) has been shown previously to be suitable for the light milling (“polishing”) of small rice samples (e.g., 10-g samples) (Bautista and Siebenmorgen, 2002). As sup-plied by the manufacturer, the Kett mill is unsuitable for micronutrient work since the rubber spinning disk and

walls have been shown to introduce Zn contamination to rice samples ( J. Stangoulis, unpublished data, 2010). However, modified Kett mills, in which the rubber disk and gasket in the polishing chamber are replaced with polyurethane PU40 90A, have been shown to be suitable for rice micronutrient work, that is, to mill rice without introducing detectable micronutrient or Al contamina-tion ( J. Stangoulis, unpublished data, 2010).

The present work was divided into three main parts. The first experiment investigated what milling time should be used to clean wheat and whether the process of cleaning itself introduces contamination. The second experiment, a validation, investigated the effect of clean-ing on highly contaminated lines from two environments. The third experiment investigated the effect of cleaning on the precision of quantitative trait locus (QTL) analysis.

MAteriAls And MethodMilling time studyInitial work was conducted using lines from the Berkut × Krichauff double haploid (DH) population grown in 2007 at the Roseworthy Agricultural College (RAC) in South Australia. The Berkut × Krichauff DH population was generated by cross-ing the Australian cultivar Krichauff with the CIMMYT cultivar Berkut. This population has already been used in other studies (McDonald et al., 2008; Huynh et al., 2008; Nguyen et al., 2011).

Plots from the available three replicates of all genotypes were screened to identify contaminated plots. From this data, three contaminated genotypes, DH35, DH57 and DH135, were selected and these showed Al concentrations in the range of 8 to 13 mg kg-1 (Table 1). Two uncontaminated genotypes, Har-vestPlus Advance Nursery (HPAN) 1 and HPAN 2, were also selected, which were taken from the HarvestPlus irrigated wheat trials at Obregon, Mexico, and showed very low levels of dusti-ness. It was not possible to select uncontaminated genotypes from RAC year 2007, because all lines from that site were slightly or highly contaminated. HarvestPlus Advance Nursery 1 and HPAN 2 showed Al concentrations of 1.6 and 1.5 mg kg-1.

All samples were dried overnight at 60°C before the mill-ing treatment. Milling times used were 5, 10, 20, and 40 s, in addition to an unmilled control. Approximately 2 g of wheat grain were cleaned for each treatment. The experiment was conducted with four replications; that is, four different sub-samples of each genotype were milled at each milling time. It was not necessary to further clean the milled samples with a sieve after each run as the Kett mill is equipped to separate out bran from the polished rice, but the Kett mill was thoroughly cleaned between samples. Samples were milled in random order and each treatment was replicated four times. Wheat samples were carefully weighed to the nearest 0.001 g before and after milling to allow the calculation of dry weight removed and retention percentages for each genotype across the entire range of genotypes. To observe the changes brought about by mill-ing, photographs of seed samples were taken before and after milling. After milling, samples were analyzed for Al, Fe, and Zn at Waite Analytical Services (The University of Adelaide, Waite Campus in Netherby, South Australia). To investigate

crop science, vol. 54, january–february 2014 www.crops.org 3

the concentrations of Zn, approximately 0.6 to 0.8 g of oven-dried wheat seeds were weighed into 75 mL Pyrex tubes and digested in 11 mL mixture of HNO3 and HClO4 (10:1 v/v). This mixture was wrapped with polyethylene and kept over-night in a fume cupboard for predigestion. Tubes were then removed from cupboard and digested to reduce the sample to around 1 mL HClO4. Then, samples were vortexed after adding water and analyzed (Wheal et al., 2011). Quality control and quality assurance was assessed using control sample data gener-ated when performing routine analyses. Duplicates of at least one National Institute of Standards and Technology (NIST) reference material (wheat flour 1567a) were prepared per batch. Averaged results were compared to published values from either NIST certified values (mean ± 95% confidence interval) or Australasian Soil and Plant Analysis Council consensus values (median ± mean absolute deviation) (Wheal et al., 2011).

Validation of Five second Milling Method on contaminated lines From two environmentsTo validate the cleaning method on highly contaminated lines, a set of 30 highly contaminated wheat lines from the RAC site were randomly selected and analyzed both as “contaminated” samples and as “cleaned” samples. This enabled the calculation of the amount of Fe and Zn removed per unit of Al removed in a larger group of lines. Finally, validation was conducted on a larger set of 240 lines from Pakistan. This population of 240 lines had been grown in the field at Islamabad in Pakistan during 2010 as part of HarvestPlus trials. From initial analysis by ICP–OES, it was clear that all 240 samples contained Al in the range of 5 to 20 mg kg-1. These lines were milled for 5 s and then reanalyzed by ICP–OES, again allowing investigation of the ratios of elements removed and the efficacy of the cleaning method.

Precision of Quantitative trait loci AnalysisTo investigate whether cleaning leads to greater precision of QTL analysis, lines from the Berkut × Krichauff DH popula-tion grown in 2009 at RAC were analyzed for QTL by using the newest version of inclusive composite interval mapping, before and after cleaning.

resultsMilling time studyThere was a steady reduction in dry weight retention as milling time increased, with approximately 0.45% dry weight reduction after 5 s milling, 1% reduction after 10 s milling, 2% reduction after 20 s milling, and 3% reduction after 40 s milling (Fig. 1).

For clean lines, with Al concentration approximately 1.5 mg kg-1, Al concentration decreased by around 1 mg kg-1 after just 5 s of milling and then decreased steadily by an additional 0.1 to 0.2 mg kg-1 at each subsequent time point (Table 2). In percentage terms, 5 s removed 60 to 70% of the Al and then subsequent milling periods of 5 to 20 s only removed an additional 5% per time point. For Fe concentra-tion, a 5 s cleaning caused a decrease of similar (although slightly greater) magnitude. For Zn concentrations, levels were only decreased very slightly by cleaning, at about 0.2

Tab

le 1

. Dry

wei

ght

(g) (

DW

) ret

enti

on

(%) a

mo

ng t

wo

unc

ont

amin

ated

(UC

) and

thr

ee c

ont

amin

ated

(C) g

eno

typ

es a

fter

mill

ing

fo

r d

iffe

rent

tim

es.

Gen

oty

pe

Al

(mg

kg-

1 )C

lass

Mill

ing

tim

e (s

)

P

erce

nt

rete

ntio

n

05

1020

40

DW

b

efo

re

clea

nP

erce

nt

rete

ntio

n

DW

b

efo

re

clea

n

DW

af

ter

cl

ean

Per

cent

re

tent

ion

DW

b

efo

re

clea

n

DW

af

ter

cl

ean

Per

cent

re

tent

ion

DW

b

efo

re

clea

n

DW

af

ter

clea

nP

erce

nt

rete

ntio

n

DW

b

efo

re

clea

n

DW

af

ter

clea

n

HPA

N 1

†1.

6U

C2.

1810

02.

212.

299

.55

2.12

2.1

99.0

62.

222.

1898

.20

2.2

2.1

95.4

5

HPA

N 2

1.5

UC

1.97

100

1.98

1.97

99.4

92.

032.

098

.52

2.14

2.1

98.1

32.

082.

096

.15

DH

3512

.1C

2.02

100

2.07

2.06

99.5

21.

971.

9498

.48

2.07

2.02

97.5

81.

861.

896

.77

DH

579.

1C

2.02

100

2.39

2.38

99.5

82.

32.

2899

.13

2.23

2.19

98.2

12.

242.

1897

.32

DH

135

6C

2.01

100

2.28

2.27

99.5

62.

212.

299

.55

2.35

2.31

98.3

02.

182.

1297

.25

Mea

n10

099

.55

98.9

598

.08

96.5

9† H

PAN

, Har

vest

Plu

s A

dva

nce

Nur

sery

.

4 www.crops.org crop science, vol. 54, january–february 2014

to 0.5 mg kg-1 over the first three time points and then more after that. No elemental concentrations were increased by milling, indicating that the milling method does not intro-duce significant amounts of contamination.

For contaminated lines (average Al level of 9 mg kg-1), Al concentration was reduced, on average, by 5.6 mg kg-1 by the first 5 s milling, and then a further 0.5 to 1 mg kg-1 was removed at each subsequent time point. In percentage terms this translated to 60% Al removal in the first 5 s and then removal of an additional 3 to 7% at each time point. In absolute terms, Al was reduced to <5 mg kg-1 in all contaminated genotypes after just 5 s of milling but was not ever reduced to below 1 mg kg-1, even after 40 s of milling. The detection limits of the ICP–OES are around 3 mg kg-1 so we consider very low levels of Al to be ambiguous and mostly baseline noise.

The Fe concentration in clean lines (average Fe was around 31 mg kg-1) was decreased by around 1.5 mg kg-1 after 5 s of milling and then decreased by an additional 0.2 to 2.5 mg kg-1 at each subsequent time point. In per-centage terms 5 s removed 5% of Fe, and then subsequent milling periods of 5 to 20 s removed an additional 0.7 to 6.5% per time point. For dirty lines (average Fe level 48.8 mg kg-1), Fe concentration was reduced, on average, by around 4 mg kg-1 by the first 5 s of milling and then a further 0.7 to 2.3 mg kg-1 at each subsequent time point. In percentage terms, this translated to 8% Fe removal in the first 5 s and then removal of an additional 1.5 to 4.6% after 10 to 20 s and then 8% after 40 s.

Zinc concentration in clean lines (average Zn 36.5 mg kg-1) was decreased by around 0.3 mg kg-1 after 5 s of milling and either stayed similar or decreased slightly up to 20 s and then decreased by about 3 mg kg-1 at 40 s of milling. With the contaminated lines (average Zn concentration of 30 mg kg-1), the Zn concentration decreased by about 1 to 1.5 mg kg-1, except after 40 s milling when the decrease was around

2 mg kg-1. In percentage terms, a similar trend was observed as with Fe, except that less Zn was removed than Fe.

Overall, the generalization can be made that most of the Al that can be removed by milling was removed after the first 5 s and that additional removal of Al after the first 5 s was slight but increased progressively with time. Con-taminated lines tended to retain more Al than clean lines. For Fe, a quick reduction of 2 mg kg-1 (clean lines) or 4 mg kg-1 (contaminated lines) was observed after the first 5 s of milling, and subsequent removal was again slight but steady. For Zn, there was no sudden decrease of Zn after the first 5 s; instead, the decrease was minor and steady across all time points. Another generalization is that Al and Fe were removed in similar amounts over the first 20 s of milling (at approximately a 1:1 ratio). With 40 s of milling, however, in clean lines, more Fe was removed than Al. Over the first 20 s of milling much less Zn was removed than Al and Fe, with Zn typically removed at about 0.2 mg kg-1 for every 1 mg kg-1 of the other elements.

Visual inspection of the five genotypes revealed the milling action to remove a majority of seed hairs, even after 5 s (for examples, see Berkut and DH135 before and after milling [Fig. 2]). Milling for longer times led to a more severe erosion of the seed coat.

Validation of the Five second Milling Method on contaminated linesTo confirm typical removal rates of Al, Fe, and Zn from contaminated wheat samples, 30 genotypes of the Berkut × Krichauff DH population that was grown in the field at Roseworthy in 2007 were selected and divided into three categories of contamination: Category (Cat) 1 (samples with 4 to 6 mg Al kg-1), Cat 2 (samples with 6 to 8 mg Al kg-1), and Cat 3 (samples with 8 to 10 mg Al kg-1). These samples were all subjected to 5 s of milling. Removal of Al was greatest from the most contaminated lines (Table 3): 6

Figure 1. Dry weight (DW) retention (%) among two uncontaminated genotypes (HarvestPlus Advance Nursery [HPAN] 1 and HPAN 2) and three contaminated genotypes (DH35, DH57, and DH135) after milling for different times

crop science, vol. 54, january–february 2014 www.crops.org 5

mg kg-1 was removed from lines that origi-nally contained 9 mg kg-1, 5 mg kg-1 from lines that contained 7 mg kg-1, and 3.6 mg kg-1 from lines that contained 5.4 mg kg-1. Similar amounts of Fe were removed as Al. For Zn, however, much less was removed and the amount of Zn removed was not highest in the most contaminated lines, consistent with the idea that Zn levels are not proportional to soil contamination level. Therefore, the generalization may be made that, in these highly contami-nated lines, Fe was removed by cleaning at approximately a 1:1 ratio with Al but that Zn removal only occurred at about 1.4 mg kg-1 per 5 mg kg-1 (equivalent to 0.2 mg kg-1 per 1 mg kg-1 of Al).

For final validation of the 5 s milling method, a further 240 highly contaminated lines were cleaned by milling (average Al before cleaning was 10 mg kg-1). This study was conducted first as a service to HarvestP-lus, who did not consider data obtained from the original ICP–OES analysis to be reli-able and wanted the analysis repeated after cleaning of samples, and second because it provided a good opportunity to see how effective cleaning was on a large collection of lines and to investigate ratios of Al, Zn, and Fe removal in another environment.

Before cleaning, average Al concentra-tions were 10 mg kg-1, and only 9 out of 240 or 4% of wheat lines showed Al con-centrations <5.0 mg kg-1 (i.e., were clean enough for results to be considered by Har-vestPlus to be reliable). After cleaning, 183 out of 240 or 76% had Al <5.0 mg kg-1, and average Al levels were 4.2 mg kg-1. There-fore, the cleaning procedure was highly successful in reducing Al levels in contami-nated samples but only partially successful at reducing Al to acceptable levels (<5 mg kg-1). Iron levels were reduced from an average of 49.4 mg kg-1 to an average of 42.8 mg kg-1 in the cleaned set, and Zn levels were reduced from an average of 38.1 mg kg-1 to 35.8 mg kg-1 in the cleaned set.

Regarding the ratios of elements removed, a ratio of 1:1 for Al:Fe was evi-dent, with slightly more Fe removed than Al, with much less Zn removed than Fe (approximately 0.4 concentration units of Zn removed per unit of Al).

Tab

le 2

. Co

ncen

trat

ion

(mg

kg-

1 ) o

f A

l, F

e, a

nd Z

n in

tw

o c

lean

and

thr

ee c

ont

amin

ated

gen

oty

pes

aft

er m

illin

g f

or

dif

fere

nt t

imes

.

Ele

men

tG

eno

-ty

pe

Co

ntam

- in

atio

n cl

ass

Mill

ing

tim

e (s

)

05

1020

40

Ori

gin

al

conc

en-

trat

ion

Per

cent

re

tent

ion

Co

ncen

-tr

atio

nD

ecre

ase

Per

cent

re

tent

ion

Co

ncen

-tr

atio

nD

ecre

ase

Per

cent

re

tent

ion

Co

ncen

-tr

atio

nD

ecre

ase

Per

cent

re

tent

ion

Co

ncen

-tr

atio

nD

ecre

ase

Per

cent

re

tent

ion

Al

HPA

N 1

†C

lean

1.58

100

0.59

0.99

370.

441.

1428

.00.

441.

1428

0.34

1.24

21A

lH

PAN

2C

lean

1.52

100

0.50

1.0

330.

421.

1028

.00.

301.

2220

0.23

1.29

15A

lM

ean

Cle

an1.

5510

00.

541.

0134

.90.

431.

1227

.70.

371.

1823

.70.

291.

2718

.3A

lD

H35

Con

tam

inat

ed12

.110

03.

98.

232

3.8

8.3

313.

68.

529

.82.

59.

620

.7A

lD

H57

Con

tam

inat

ed9.

110

03.

75.

441

2.5

6.6

272.

56.

627

.52.

36.

825

.3A

lD

H13

5C

onta

min

ated

610

02.

83.

247

2.4

3.6

402.

13.

935

.01.

44.

623

.3A

lM

ean

Con

tam

inat

ed9.

110

03.

55.

639

.92.

96.

233

.02.

76.

330

.72.

17.

023

.1Fe

HPA

N 1

Cle

an30

.32

100

29.5

00.

897

.329

.20

1.1

96.3

28.6

61.

794

.526

.90

3.43

89Fe

HPA

N 2

Cle

an32

.02

100

29.8

92.

193

.429

.72

2.3

92.8

29.6

82.

392

.727

.40

4.62

85.6

FeM

ean

Cle

an31

.17

100

29.6

91.

595

.329

.46

1.71

94.6

29.1

72.

093

.627

.15

4.02

87.1

FeD

H35

Con

tam

inat

ed50

.710

043

.57.

285

.844

.95.

888

.644

.46.

387

.642

.28.

583

.2Fe

DH

57C

onta

min

ated

58.4

100

54.5

3.9

93.3

52.6

5.8

90.1

54.1

4.3

92.6

51.4

788

.0Fe

DH

135

Con

tam

inat

ed37

.210

036

.11.

197

.034

.52.

792

.734

.23.

091

.932

.44.

887

.1Fe

Mea

nC

onta

min

ated

48.8

100

44.7

4.1

92.1

44.0

4.8

90.5

44.2

4.5

90.7

42.0

6.8

86.1

ZnH

PAN

1C

lean

33.5

210

033

.46

0.1

99.8

33.2

40.

399

.233

.21

0.3

99.1

31.0

12.

592

.5Zn

HPA

N 2

Cle

an39

.54

100

39.0

50.

598

.838

.77

0.8

98.1

39.3

60.

299

.536

.24

3.3

91.7

ZnM

ean

Cle

an36

.53

100

36.2

60.

2799

.336

.00

0.53

98.6

36.2

80.

2599

.333

.62

2.9

92.1

ZnD

H35

Con

tam

inat

ed29

100

26.1

2.9

9026

.82.

292

.427

.11.

993

.426

.12.

990

ZnD

H57

Con

tam

inat

ed38

.710

037

.80.

997

.738

0.7

98.2

38.6

0.1

99.7

36.6

2.1

94.6

ZnD

H13

5C

onta

min

ated

21.6

100

210.

697

.220

.61.

095

.420

.41.

294

.419

.52.

190

.3Zn

Mea

nC

onta

min

ated

29.8

100.

028

.31.

595

.028

.51.

395

.328

.71.

195

.927

.42.

491

.6† H

PAN

, Har

vest

Plu

s A

dva

nce

Nur

sery

.

6 www.crops.org crop science, vol. 54, january–february 2014

Finally, a highly important attribute of genotypes that can be of interest in micronutrient screening programs is “rank” of grain concentration for the element of interest. To investigate if cleaning affects the ranking of lines, 21 lines were chosen randomly from the 240 line population and ranked for Fe and Zn concentration.

Data showed that after cleaning, the ranking of geno-types for Fe concentration changed markedly (Table 4); for example, genotype HPAN 2 changed from 3 to 21, HPAN 4 from 4 to 10, and HPAN 5 from 5 to 20. For Zn, rankings only changed slightly after cleaning; the top 10 lines retained the same rank after cleaning, but there were some changes in rank after cleaning in lines ranked 11 to 21.

Finally, the 240 wheat lines were divided into different categories depending on the concentration of Al, namely, Cat 1, with Al concentrations of 4 to 6 mg kg-1, Cat 2, with Al concentrations of 6 to 8 mg kg-1, Cat 3, with Al concen-trations of 8 to 10 mg kg-1, and Cat 4, with Al concentra-tions of 10 to 12, 12 to 14, 14 to 16, and 16 to 20 mg kg-1. Data showed that the higher the contamination class of Al, the more Al was removed and the higher the proportion of Al removed (Fig. 3). Similarly, the higher the contamina-tion class, the more Fe was removed. However, removal of Zn was not highest in the most highly contaminated classes.

It was also apparent that milling can only be expected to reduce samples with “medium” levels of contamina-tion (approximately 10 mg kg-1 to below 5 mg kg-1, the

acceptable limit for contamination with HarvestPlus sam-ples). Above that level of Al contamination, the samples could not be cleaned to <5 mg kg-1 within the 5 s mill-ing period.

Precision of Quantitative trait loci AnalysisIn contaminated lines of Berkut × Krichauff (2009), two QTL for Fe were found, one QTL on chromosome 2B with a logarithm of the odds (LOD) score of 3.6 and flanking markers cfa2278 to gwm55 and a second on chromosome 3B with a LOD score of 1.1 and flanking markers wPt-4209 to wPt-5390 (Table 5). These contam-inated lines were cleaned by the polishing method and analyzed again. The same two QTL on chromosomes 2B and 3B were found with the same flanking markers but with greater precision. The QTL on chromosome 2B had a LOD score of 3.4 after cleaning while the QTL on chro-mosome 3B had a LOD score of 3.4, which identified it as a significant QTL (Table 5).

discussionThe present study has provided useful information about how to clean grain to ensure that soil contamination does not confound Zn and Fe analysis and provides informa-tion about the levels of Zn and Fe contamination that can be expected at given levels of Al contamination.

Results from the milling experiment indicated that 5 s milling removed a large proportion of the Al and Fe that

Figure 2. Photographs of seeds before and after milling for 5 s showing removal of hair.

Table 3. Average changes in concentration (mg kg-1) of Al, Fe, and Zn in three categories of contaminated lines from the Berkut (B) × Krichauff (K) double haploid population grown in the field during 2007.

No. of genotypes

in each category

Mean range of Al (mg kg-1) incontaminated

genotypes

Al Fe Zn

Beforecleaning

Aftercleaning Reduction

Beforecleaning

Aftercleaning Reduction

Beforecleaning

Aftercleaning Reduction

10 4–6 5.37 1.8 3.6 39 35 4.0 23 21 1.6

10 6–8 6.97 2.3 4.7 39 35 4.4 24 22 2.0

10 8–10 8.92 3.0 6.0 41 37 4.1 23 23 0.4

Mean 7 2 4.7 40 35 4.2 23 22 1.4

crop science, vol. 54, january–february 2014 www.crops.org 7

Figure 3. Average changes in concentration of Al, Fe, and Zn in different categories of 240 contaminated wheat lines before and af-ter cleaning.

Table 4. Ranking of 20 genotypes with reference of Fe and Zn concentrations (mg kg-1) before and after cleaning with the modi-fied Kett mill (Kett Electric Laboratory).

Genotypes

Fe Zn Al Fe rank Zn rank

Before After Before After Before After Before After Before After

HPAN† 13 51 42 34 32 13.7 5.4 1 2 1 1

HPAN 100 50 48 43 40 5.7 2.5 2 1 2 2

HPAN 2 48 32 30 27 21.5 4.4 3 21 3 3

HPAN 4 46 36 28 27 15.4 4.6 4 10 4 4

HPAN 5 46 33 28 27 15 3.4 5 20 5 5

HPAN 15 45 38 31 30 13.3 5.3 6 3 6 6

HPAN 10 44 38 31 30 10.1 4.4 7 5 7 7

HPAN 14 44 37 29 27 8.1 3 8 7 8 8

HPAN 6 43 35 30 28 11.7 3.6 9 15 9 9

HPAN 18 42 37 31 30 10.3 3.5 10 8 10 10

HPAN 9 42 38 30 29 9.4 4.9 11 4 11 19

HPAN 1 42 36 34 31 11.6 4.8 12 9 12 12

HPAN 12 42 35 33 31 11.5 5.6 13 12 13 11

HPAN 20 42 36 28 25 9.3 4.2 14 11 14 13

HPAN 8 42 38 28 27 10.2 5.7 15 6 15 15

HPAN 19 41 33 27 25 10.3 3.6 16 19 16 16

HPAN 11 41 35 32 30 11.4 5 17 13 17 17

HPAN 16 40 34 36 34 8.8 3.5 18 17 18 20

HPAN 17 40 34 32 30 8.6 4 19 18 19 18

HPAN 3 40 35 28 29 9.5 4.1 20 14 20 14

HPAN 7 39 34 30 28 9.9 4.6 21 16 21 21†HPAN, HarvestPlus Advance Nursery.

Table 5. Quantitative trait loci (QTL) for Fe concentration (% of dry weight) measured on the dirty versus clean grain of the Berkut × Krichauff doubled-haploid population grown at Roseworthy in 2009.

QTL Chromosome Position (cM) Flanking markers LOD† scorePercentage of variance (%)

Additive effect‡ (mg kg-1)

QGFe.ta-2B (dirty) 2B 116 cfa2278 to gwm55 3.6 10.6 –1.1

QGFe.ta-2B (clean) 2B 112 cfa2278 to gwm55 3.4 8.2 –1.0

QGFe.ta-3B (clean) 3B 63 wPt-4209 to wPt-5390 3.4 7.9 0.9

QGFe.ta-3B (dirty) 3B 63 wPt-4209 to wPt-5390 1.1 2.7 0.6†LOD, logarithm of the odds.‡A positive effect indicates that the allele from Berkut contributes to higher grain Fe levels while a negative effect indicates that the allele from Krichauff contributes to higher grain Fe levels.

8 www.crops.org crop science, vol. 54, january–february 2014

can be removed by milling, with Zn levels only slightly affected. These results are consistent with Al existing mostly on the outside of the grain, probably in dust stuck to grain hairs, with very low levels of Al in the seed coat or endosperm. Iron, in contrast, exists on both the out-side of the grain in contamination and in the seed coat. Therefore, there is a sudden drop in both Al and Fe as soon as seed hairs and associated surficial contamination are removed by milling. After this initial effect, not much more Al is removed—the remaining Al is probably located in cracks in the testa and the grain crease. With Fe, there is an initial drop in Fe values as contaminant Fe is removed and then steady removal of Fe with longer milling time as the seed coat, known to be relatively rich in Fe (Moore et al., 2012; Singh et al., 2012), is eroded. Photographs of seeds milled for 5 s confirmed that most hairs are removed from seeds by the 5 s milling treatment. With Zn, results were quite different in that there was only a slight drop in Zn with 5 s of milling and then steady slight decreases with later time points. This is consistent with Zn being present at low levels in soil (the global average is only 48 mg kg-1, compared to 33,000 mg kg-1 Fe and 62,000 mg kg-1 Al [Rauch and Pacyna, 2009]). Furthermore, this finding is consistent with the fact that there is no particu-lar localization of Zn to the seed coat, so erosion of the seed coat did not particularly decrease Zn levels.

It was important to include “clean” samples in this study, as a way of checking how much “plant” micronutrient may be removed by cleaning. Therefore, two “clean” genotypes were also included in the milling time study. Data indicated that while milling of “clean” seeds did remove some Al, Fe, and Zn, the amount removed was only around 1 mg kg-1 for Al and Fe and 0.3 mg kg-1 for Zn. It is likely that the slight decrease in elemental concentrations was caused in part by the removal of low level contamination from the “clean” lines; the lines probably did still have some dust on them that was removed by milling. It is also possible that the slight decrease was caused by removal of hairs and the slight erosion of the seed coat. It is significant, however, that the removal occurred at <1 mg kg-1. Therefore, it appears that removal of actual “seed” micronutrient (e.g., in seed hairs) only occurs at a maximum of 1 mg kg-1 for Fe and Al and at a maximum of 0.3 mg kg-1 for Zn.

It is also noteworthy that, among all the lines milled (clean and contaminated), elemental concentrations never increased with milling treatment; therefore, it does appear that the modified mill, by contrast with the original mill, can complete the milling process without introducing any contamination itself.

Having established that 5 s milling was the most suit-able milling time, the next step in this study was to apply the milling method to contaminated lines—first on 30 highly contaminated Australian lines and then on 240 highly contaminated lines from Pakistan.

The cleaning method was highly successful in reduc-ing contamination, with Al in the range 5 and 10 mg kg-1; it generally cleaned to the point where Al was 2 to 3 mg kg-1 (in the Australian lines), and among Pakistani lines average Al was reduced from around 10 to around 4 mg kg-1. However, it is noted that the most highly con-taminated lines (24% of the Pakistani collection) were not able to be brought back to the <5 mg kg-1 Al level used by HarvestPlus for acceptably clean. Therefore, the clean-ing method could be considered suitable for bringing con-taminated lines to a sufficiently clean state for analysis but is not suitable for use on highly contaminated lines.

Given the likelihood that some of the contamination that remains on grain after milling for 5 s could be pres-ent as fine dust on the surface of grain, one possibility to improve the cleaning method would be to briefly wipe seed with a damp cloth after milling. However, removal of contamination from the crease would be more difficult due to inaccessibility to cleaning methods, in which case it may not be practically possible to ever bring highly con-taminated lines to the 5 mg kg-1 level.

One very interesting result in the present study was that the ratio between Al removed and Fe removed was fairly consistent at approximately 1:1, across two environ-ments; that is, similar amounts of Fe and Al were removed by cleaning. This confirms that, at levels of contamination that are often seen in the field (of, say, 5–10 mg kg-1 Al), practically significant amounts of Fe will also be present, inflating Fe values by a similar amount. This result was consistent with the finding that the ranking of Fe in lines changed markedly after cleaning; that is, because levels of soil contamination vary in a set of field samples, then lev-els of contamination vary, and ranking of Fe concentra-tions in dirty samples becomes confounded. Therefore, it is clear that cleaning is important for accurate Fe analysis.

A second very interesting result is that the ratio between decreased Zn concentration and decreased Al concentra-tion was fairly consistent at approximately 0.2:1, across two environments; that is, that only about one-fifth as much Zn was removed as Al (and Fe). This indicates that soil con-tamination actually does not much affect grain Zn levels—for example, a line with soil contamination at 10 mg kg-1 probably only has Zn levels increased by the soil at around 2 mg kg-1. Similarly, the ranking of lines did not change much after cleaning—those that were highly ranked in the contaminated set were also highly ranked in the cleaned set.

The finding that Fe values were markedly affected by soil contamination clearly indicates that any program tar-geting Fe should aim to have clean seed. However, the finding that the relationship between Fe levels removed by cleaning and Al levels removed by cleaning seem to be similar (approximately 1:1 ratio) raises the possibility that breeders could still conduct screens on soil contami-nated material and simply adjust Fe concentrations down

crop science, vol. 54, january–february 2014 www.crops.org 9

by the approximate concentration of Al in the sample; for example, if ICP–OES results indicate Al to be present at 8 mg kg-1, breeders would know that Fe values are prob-ably inflated by a similar amount. Therefore, line rankings could be made on this corrected amount and the screen may still be worthwhile. However, it would still be recom-mended that breeders aim to keep lines as clean as possible and also validate that Al and Fe concentrations from their particular environment show the same relationship.

The findings of this study for Zn are different for Fe in the sense that breeders probably do not need to be too concerned that low level soil contamination is altering Zn values. Instead of insisting that all samples in Zn bio-fortification programs contain <5 mg kg-1 Al, breeders

may instead tolerate higher levels of Al and know that the Zn concentration rankings of their lines are not greatly affected by the soil and that absolute Zn levels are prob-ably only inflated by approximately a few milligrams per kilogram (e.g., 1–3 mg kg-1) by the soil contamination. Alternatively, breeders may attempt to correct for Zn con-tamination by adjusting the values down by one-fifth the value of Al concentrations. This study showed that while Fe concentration and ranking is greatly affected by soil contamination, Zn concentration and ranking is not.

The findings of this study also provided evidence to show that cleaning seed before analysis improves the precision of Fe QTL. Seed cleaning either narrowed the QTL peak (chromosome 2B in QTL; Fig. 4) or significantly increased

Figure 4. Chromosomal locations and logarithm of the odds (LOD) scores of Fe quantitative trait loci in Berkut × Krichauff (population grown in year 2009) before and after cleaning.

10 www.crops.org crop science, vol. 54, january–february 2014

the significance of a QTL (i.e., the chromosome 3B QTL in Fig. 4). The method of cleaning seed is both cheap and easy to run and is worthwhile as it improves analytical outcomes.

AcknowledgmentsWe would like to thank HarvestPlus for funding, Dr. Velu Govinden at CIMMYT and Dr. Muhammad Yaqub Mujahid at PARC for providing seed samples for data validation and the Waite Analytical Service for nutrient analyses.

referencesBautista, R.C., and T.J. Siebenmorgen. 2002. Evaluation of labora-

tory mills for milling small samples of rice. Appl. Eng. Agric. 18(5):575−583.

Cakmak, I., R. Graham, and R.M. Welch. 2002. Agricultural and molecular genetic approaches to improving nutrition and pre-venting micronutrient malnutrition globally In: Encyclopedia of life support systems. UNESCO-EOLSS Publishers Co. Ltd., Oxford, UK. p. 1075–1099.

Cakmak, I., A. Torun, E. Millet, M. Feldman, T. Fahima, A. Korol, E. Nevo, H.J. Braun, and H. Ozkan. 2004. Triticum dicoccoi-des: An important genetic resource for increasing zinc and iron concentration in modern cultivated wheat. Soil Sci. Plant Nutr. 50:1047–1054. doi:10.1080/00380768.2004.10408573

Cherney, J.H., and D.L. Robinson. 1983. A comparison of plant digestion methods for identifying soil contamination of plant tissue by Ti analysis. Agron. J. 75:145–147. doi:10.2134/agronj1983.00021962007500010037x

Djingova, R., and I. Kuleff. 1994. On the sampling of vascular plants for monitoring of heavy metal pollution. In: B. Markert, editor, Environmental sampling for trace analysis. VCH Ver-lagsgesellschaft, Weinheim, Germany. p. 395–414.

Graham, R.D., D. Senadhira, S. Beebe, C. Iglesias, and I. Monaste-rio. 1999. Breeding for micronutrient density in edible portion of staple food crops: Conventional approaches. Field Crops Res. 60:57–80. doi:10.1016/S0378-4290(98)00133-6

Hall, G.E.M. 1995. Sample preparation and decomposition. In: R.R. Brooks, C.E. Dunn, and G.E.M. Hall, editors, Biologi-cal systems in mineral exploration and processing. Ellis Hor-wood, New York, NY. p. 427–442.

Huynh, B.L., H. Wallwork, J.C.R. Stangoulis, R.D. Graham, K.L. Willsmore, S. Olson, and D.E. Mather. 2008. Quantitative trait loci for grain fructan concentration in wheat (Triticum aestivum L.). Theor. Appl. Genet. 117:701–709. doi:10.1007/s00122-008-0811-6

Ibia, T.O. 2002. Forms of Fe and Al in soil profiles of inland flood plains of South Eastern Nigeria. Nig. J. Soil Res. 3:72–77.

Jones, J.B., Jr. 1991. Plant tissue analysis in micronutrients. In: J.J. Mortvedt, editor, Micronutrients in agriculture, 2nd edition. SSSA, Madison, WI. p. 477–521.

Jones, J.B., Jr. 2001. Laboratory guide for conducting soil tests and plant analysis. Chapter 3. CRC Press, Boca Raton, FL. p. 191–239.

McDonald, G.K., Y. Genc, and R.D. Graham. 2008. A simple method to evaluate genetic variation in grain zinc concen-tration by correcting for differences in grain yield. Plant Soil 306:49–55. doi:10.1007/s11104-008-9555-y

Moore, K.L., F.J. Zhao, C.S. Gritsch, P. Tosi, M.J. Hawkesford, S.J. McGrath, P.R. Shewry, and C.R.M. Grovenor. 2012. Locali-sation of iron in wheat grain using high resolution secondary ion mass spectrometry. J. Cereal Sci. 55:183–187. doi:10.1016/j.jcs.2011.11.005

Nestel, P., H.E. Bouis, J.V. Meenakshi, and W. Pfeiffer. 2006. Bio-fortification of staple food crops. J. Nutr. 136:1064–1067.

Nguyen, V.L., B.L. Huynh, H. Wallwork, and J. Stangoulis. 2011. Identification of quantitative trait loci for grain arabinox-ylan concentration in bread wheat. Crop Sci. 51:1143–1150. doi:10.2135/cropsci2010.08.0484

Paltridge, N.G., L.J. Palmer, P.J. Milham, G.E. Guild, and J.C.R. Stangoulis. 2012a. Energy-dispersive X-ray fluorescence analy-sis of zinc and iron concentration in rice and pearl millet grain. Plant Soil 361(1–2):251–260. doi:10.1007/s11104-011-1104-4

Paltridge, N.G., P.J. Milham, J.I. Ortiz-Monasterio, G. Velu, Z. Yasmin, L.J. Palmer, G.E. Guild, and J.C.R. Stangoulis. 2012b. Energy-dispersive X-ray fluorescence spectrometry as a tool for zinc, iron and selenium analysis in whole grain wheat. Plant Soil 361(1–2):261–269. doi:10.1007/s11104-012-1423-0

Peterson, C.J., V.A. Johnson, and P.J. Mattern. 1986. Influence of cultivar and environment on mineral and protein concentra-tions of wheat flour, bran, and grain. Cereal Chem. 63:118–186.

Pfeiffer, W.H., and B. McClafferty. 2007. HarvestPlus: Breeding crops for better nutrition. Crop Sci. 47:S88–S105. doi:10.2135/cropsci2007.09.0020IPBS

Rauch, J.N., and J.M. Pacyna. 2009. Earth’s global Ag, Al, Cr, Cu, Fe, Ni, Pb, and Zn cycles. Global Biol. Cycles 23:GB2001. doi:10.1029/2008GB003376

Rengel, Z., G.D. Batten, and D.E. Crowley. 1999. Agronomic approaches for improving the micronutrient density in edible portion of field crops. Field Crops Res. 60:27–40. doi:10.1016/S0378-4290(98)00131-2

Sager, M., and J. Mittendorfer. 1997. Influence of milling or cutting procedures on trace element contents of plant samples. Int. J. Envi-ron. Anal. Chem. 67:59–71. doi:10.1080/03067319708031394

Sillänpaa, M. 1982. Micronutrients and nutrient status of soils: A global study. Soil Bull. 48. FAO, Rome, Italy.

Singh, S.P., K. Vogel-Mikus, P. Pelicon, P. Vavpetic, L. Jeromel, R. Feng, R. Chibbar, J.K. Roy, S.S. Mantri, J. Kumar, and R. Tuli. 2012. New insight into iron transport from maternal tissues to endosperm in mature wheat seed using synchrotron radiation. Qual. Ass. Safety Crops Foods 4:141. doi:10.1111/j.1757-837X.2012.00146.x

Sparks, D.L. 2003. Environmental soil chemistry. 2nd ed. Aca-demic Press, San Diego, CA. p. 269–270.

Velu, G., R.P. Singh, J. Huerta-Espino, R.J. Pena, B. Arun, A. Mahendru-Singh, M.Y. Mujahid, V.S. Sohu, G.S. Mavi, J. Crossa, G. Alvarado, A.K. Joshi, and W.H. Pfeiffer. 2012. Performance of biofortified spring wheat genotypes in target environments for grain zinc and iron concentrations. Field Crops Res. 137:261–267. doi:10.1016/j.fcr.2012.07.018

Wheal, M.S., T.O. Fowles, and L.T. Palmer. 2011. A cost-effec-tive acid digestion method using closed polypropylene tubes for inductively coupled plasma optical emission spectrometry (ICP-OES) analysis of plant essential elements. Anal. Methods 3:2854–2863. doi:10.1039/c1ay05430a