Detection of biotic stress caused by apple stem grooving virus in apple trees using hyperspectral reflectance analysis

Abstract

The applicability of hyperspectral remote sensing technique, leaf spectral reflectance, for detection of latent infection in young apple trees caused by apple stem grooving virus (ASGV) was investigated. Apple trees, cultivar Florina, planted at the end of 2012 in the Kostinbrod region were used for the analyses. They developed without symptoms of viral infection of the aerial parts during the vegetation period. In the spring of 2013 twenty three trees of the orchard were checked for presence of ASGV infection using enzyme immunosorbent assay test DAS-ELISA. Five of the trees were not infected and were adopted as control. Also, non-destructive hyperspectral analyses of the reflected radiation from the leaf samples of the investigated trees were carried out. Spectral measurements were conducted with a portable fibre-optics spectrometer in the visible and near infrared spectral ranges (450–850 nm). Presence of ASGV was revealed by changes in the reflectance spectra in the green, red, red-edge, and near infrared spectral regions. Statistical analyses (Student t-test and cluster analysis) and first derivative were applied to assess the differences between reflectance data of control and infected trees. A clearly expressed correlation was found between the results from the spectral reflectance and serological analyses.

Full text

Доклади на Българската академия на науките
Comptes rendus de l’Acad´emie bulgare des Sciences
Tome 68, No 2, 2015
PHYSIQUE
Spectroscopie
DETECTION OF BIOTIC STRESS CAUSED BY APPLE
STEM GROOVING VIRUS IN APPLE TREES USING
HYPERSPECTRAL REFLECTANCE ANALYSIS
Dora Krezhova, Antony Stoev∗, Svetla Maneva∗
(Submitted by Corresponding Member Ch. Stoyanov on December 13, 2014 )
Abstract
The applicability of hyperspectral remote sensing technique, leaf spectral
reflectance, for detection of latent infection in young apple trees caused by apple
stem grooving virus (ASGV) was investigated. Apple trees, cultivar Florina,
planted at the end of 2012 in the Kostinbrod region were used for the analyses.
They developed without symptoms of viral infection of the aerial parts during
the vegetation period. In the spring of 2013 twenty three trees of the orchard
were checked for presence of ASGV infection using enzyme immunosorbent as-
say test DAS-ELISA. Five of the trees were not infected and were adopted as
control. Also, non-destructive hyperspectral analyses of the reflected radia-
tion from the leaf samples of the investigated trees were carried out. Spectral
measurements were conducted with a portable fibre-optics spectrometer in the
visible and near infrared spectral ranges (450–850 nm). Presence of ASGV was
revealed by changes in the reflectance spectra in the green, red, red-edge, and
near infrared spectral regions. Statistical analyses (Student t-test and cluster
analysis) and first derivative were applied to assess the differences between re-
flectance data of control and infected trees. A clearly expressed correlation was
found between the results from the spectral reflectance and serological analyses.
Key words: hyperspectral leaf reflectance, apple trees, apple stem groov-
ing virus (ASGV), DAS-ELISA
Introduction. During the vegetation period crop plants are exposed to
different kinds of stress. Most biotic stress factors like plant pathogens (fungi,
bacteria, viruses, nematodes, insects, etc.) as well as abiotic stress factors like wa-
ter and nutrient deficiency, temperature anomalies, etc., affect the photosynthetic
175

apparatus observably [1,2 ]. Nowadays, about one fourth of global agricultural pro-
duce is lost due to biotic stresses [3]. Plant viral infections in particular cause
enormous economic losses in many crops worldwide. Many studies have investi-
gated the development of viral diseases and changes in the plant-virus interactions
on tissue level [4]. Most of these studies were focused on viral infections of herba-
ceous plants producing visible disease symptoms [5]. However, the interactions
of viruses with woody species, including economically important fruit trees, are
poorly studied. The integration of modern technologies and information manage-
ment systems has the potential for early detection of biotic stress in plants which
is particularly important for precision crop protection.
Apple (Malus domestica) is one of the most widely grown fruit crops in the
world. It is an important source of energy, vitamins and minerals in human diet.
Apple is susceptible to infection by pathogens, especially viruses such as Apple
stem grooving virus (ASGV), Apple chlorotic leaf spot virus (ACLSV) and Apple
stem pitting virus (ASPV), all of which are economically important and common
pathogens in commercial cultivars [6,7 ]. These viruses do not usually induce
visible disease symptoms in the infected trees and fruits, although the infection
eventually does lead to significant reduction in fruit yield and quality [8]. Further,
the infection is permanent in apple trees due to vegetative propagation [9].
ASGV, the type member of the genus Capillovirus, is wide spread in its oc-
currence and has been described as a latent virus of apple because it does not pro-
duce symptoms on most commonly used commercial rootstocks and varieties [10].
However, ASGV produces stem grooving, swelling of graft unions and graft union
necrosis symptoms upon apple seedling rootstock [11]. Symptoms such as chlorotic
leaf spot and brown line may develop immediately above graft union [12]. ASGV
infection may increase susceptibility to infection by other viruses and/or other
pathogens. The mixed infections with other pome fruit viruses are frequent in
apple orchards and cause significant yield reduction [10]. The widespread distribu-
tion of the virus and potential to cause economics losses [12] are the circumstances
that bring about the screening for ASGV as being an important component of
pome fruits virus certification programs.
The laboratory assays, including enzyme linked immunosorbant assay
(ELISA) and reverse transcription polymerase chain reaction (RT-PCR), are
widely used for the detection of ASGV. However, little is known whether the
above detection assays are reliable using different tissues and during any time
of the year [13]. These methods require specific laboratory diagnostic reagents,
laboratory fitting-up and equipment (installations and appliances). Therefore,
their use is not always justified from economic point of view.
The use of non-destructive methods for detection of plant disease holds great
potential for optimization of the agricultural management of commercially impor-
tant orchards and can improve fruit quality [14]. Remote sensing (RS) technology
is broadly defined as a method of obtaining information about properties of an
176 D. Krezhova, A. Stoev, S. Maneva

object without coming into physical contact with it. A more specific definition of
RS relates to studying the environment from a distance. Nowadays the hyperspec-
tral RS techniques based on reflectance measurements, acquired in a high number
of contiguous spectral bands, allows pre-symptomatic monitoring of changes in
the physiological state of plants on large areas with high spectral resolution. The
spectral reflectance analyses proved to be very useful in detecting plant stress and
disease due to changes in the absorption of incident light in the visible (VIS) and
near-infrared (NIR) ranges of the electromagnetic spectrum [15, 16 ]. The adverse
growing conditions give rise to morphological, physiological and/or biochemical
changes that affect the manner in which plants interact with light. Moreover, the
high spatial and spectral resolution of hyperspectral reflectance data increases
the potential to detect anomalies in the normal plant production processes at an
early stage, thereby enabling researchers to model and monitor plant production
systems. This is especially important for capital-intensive perennial crops, such
as various fruit species [17].
This research aims at investigating the applicability of hyperspectral remote
sensing technique of leaf reflectance for detection of latent infection in young
apple trees caused by apple stem grooving virus (ASGV), as well as for developing
an approach to explore the potential of this technique being used in early viral
screening.
Plant material and methods. Plant material.The objects of study
were apple trees, cultivar Florina, propagated onto rootstocks MM106. Trees
were planted in two rows 3 m ×3 m in the autumn of 2012 in a small non-
commercial orchard near the town of Kostinbrod, West Bulgaria.
During the vegetation period in the next year the trees were without symp-
toms of viral infection of the aerial parts and organs. In the summer of 2013 leaves
from twenty three trees were checked trough enzyme linked immunosorbent assay
(ELISA) for ASGV. Five of the studied set of trees were not infected and were
adopted as control trees for the analyses.
Leaf spectral reflectance.Terrestrial materials reflect and absorb the inci-
dent light differently at the wavelengths and may be differentiated by their spec-
tral reflectance signatures revealed through the measured reflected electromag-
netic radiation at varying wavelengths in VIS (400–700 nm), NIR (700–1200 nm),
and short wave IR (1200–2500 nm) spectral ranges. All green vegetation species
have unique spectral features, mainly because of the chlorophyll, carotenoid, and
other pigments, and water content [18]. The spectral reflectance is a property used
to quantify the spectral signatures of the leaves and is described as a ratio of the
intensity of reflected light to the illuminated light on wavelength. The amount of
reflected light depends on a number of leaf-related factors, such as external mor-
phology, internal structure, internal distribution of biochemical components, etc.,
and makes it possible for hyperspectral remote sensing to detect deviations from
optimally functioning plant systems, i.e., stress-induced physiological changes will
Compt. rend. Acad. bulg. Sci., 68, No 2, 2015 177

affect leaf biochemical constituents such as chlorophyll concentration (Chl) and
water content [19].
Spectral measurements.Spectral reflectance measurements were carried
out on ten apple trees, cultivar Florina. Three of them were healthy (control)
and the rest were infected with ASGV. Hyperspectral reflectance data were col-
lected from fresh detached leaves by a portable fibre-optics spectrometer USB2000
(Ocean Optics) in the VIS and NIR spectral ranges (450–850 nm) at a spectral
resolution (halfwidth) of 1.5 nm. The measurements were carried out using an
experimental setup in laboratory. The light source was a halogen lamp providing
homogeneous illumination of the leaf surfaces. The spectral reflectance charac-
teristics (SRC) of the investigated plants were determined as the ratio between
the radiation reflected from the leaves and the one reflected from the diffuse
reflectance standard. Specialized software was used for data acquisition and pro-
cessing.
Data analysis.The Student’s t-criterion, cluster and first derivative anal-
yses were applied for determination of the statistical significance of differences
between the means of sets of the values of the reflectance spectra of healthy (con-
trol) and infected trees as well as the position of the inflection points in the red
edge region. The spectral reflectance analyses were performed in four most in-
formative for investigated plants spectral ranges: green (520–580 nm, maximal
reflectivity of green vegetation), red (640–680 nm, maximal chlorophyll absorp-
tion), red edge (680–720 nm, maximal slope of the reflectance spectra) and the
NIR (720–770 nm). The statistical significance of the differences between SRC of
control and infected trees was examined in ten wavelengths (λ1= 475.22 nm, λ2=
489.37 nm, λ3= 524.29 nm, λ4= 539.65 nm, λ5= 552.82 nm, λ6= 667.33 nm,
λ7= 703.56 nm, λ8= 719.31 nm, λ9= 724.31 nm, and λ10 = 758.39 nm) chosen
to be disposed uniformly over these ranges.
Serological analysis.DAS-ELISA (double antibody sandwich enzyme
linked immunosorbent assay) analyses were made using a commercial kit (LOE-
WE Biochemica GmbH, Sauerlach, Germany). Polyclonal antibodies (IgG) ex
rabbit specific for ASGV were used.
The extinction values were measured by a spectrophotometer SUMAL PE
(Karl Zeiss, Jena, Germany) at a wavelength of 405 nm. All samples with extinc-
tion values exceeding at least three times the negative control were considered
virus positive.
Results and discussion. The averaged SRC over all measurements (up to
25 areas of investigated leaves for each tree) of one of the control trees (numbered
17) and infected with ASGV seven apple trees (numbered 3, 4, 9, 12, 14, 16, 18)
are shown in Fig. 1. It is seen that the values of SRC of leaves of all of the infected
trees are positioned close to the control SRC values. The differences occurred in
the green (520–580 nm), red (640–680 nm) and NIR (720–850 nm) spectral ranges
where the values of the SRC of infected trees were increased against the control.
178 D. Krezhova, A. Stoev, S. Maneva

Fig. 1. Averaged spectral reflectance characteristics of control and seven apple trees infected
with ASGV. The inset presents the differences in the spectral region 510–690 nm

Fig. 3. a) Averaged SRC of control and two infected with ASGV apple trees from each of the
two clusters; b) Maxima of the first derivatives on SRC of control and two infected with ASGV
apple trees

Fig. 2. Hierarchical cluster analysis of the spectral reflectance data of ten investigated apple
trees: control and infected with ASGV
To discriminate the spectral behaviour of the investigated trees a hierarchical
cluster analysis (HCA) was applied to the averaged reflectance data of each of
the trees in the green spectral range (520–580 nm). HCA is an unsupervised
similarity-based [20] cluster algorithm, which creates a hierarchical cluster tree
given a set of input data objects. This cluster tree is obtained using a linkage
algorithm based on prior calculations of pairwise distance between all objects
included in each input data set. Figure 2 shows that three separate clusters are
discerned for spectral data of the ten investigated trees as one member hangs
having properties of both classes. The first cluster includes the spectral data of
the control trees, the second and third ones – the data of 4, 9, and 14 trees and
3, 12, 16, 18 trees, respectively.
To assess the statistical significance of the differences between the analysed
sets of spectral data the Student’s t-test was applied in the ten above mentioned
wavelengths (λ1to λ10). The results are displayed in Table 1. With compared
pairs it is possible to take each measurement in a given SRC and pair it sensibly
with one measurement in the control SRC. The differences between SRC of control
trees are not statistically significant at all except one of the wavelengths. In
contrast, the differences between SRC of control and infected trees from the third
cluster (3, 12, 16, and 18) are statistically significant at all except one of the
wavelengths. For the second cluster, composed of SRCs lying close to the control,
the number of statistically significant differences decreased.
Compt. rend. Acad. bulg. Sci., 68, No 2, 2015 179

T a b l e 1
p-values of the Student’s t-test in the case of apple trees, cultivar Florina,
infected with ASGV
Pairs
compared Control p < Tree 3 p < Tree 4 p < Tree 5 p < Tree 9
λ1/λ1c 3.64 *** 5.72 ns 4.62 ns 3.74 ns 3.51
λ2/λ2c 4.15 *** 6.19 *** 4.97 ns 4.29 *** 4.15
λ3/λ3c 8.16 *** 9.70 ns 9.45 ns 8.57 ns 8.19
λ4/λ4c 10.43 ** 11.61 ns 12.04 ns 11.05 ** 10.43
λ5/λ5c 11.01 ** 12.11 ns 12.72 ns 11.71 ns 11.04
λ6/λ6c 4.53 *** 6.58 *** 5.25 ns 4.67 ns 4.57
λ7/λ7c 12.72 * 13.55 ** 14.76 * 13.90 ** 12.95
λ8/λ8c 33.85 ** 37.72 ns 37.61 ns 36.82 ns 35.16
λ9/λ9c 41.66 *** 45.05 ns 45.80 ns 44.87 * 43.21
λ10/λ10c 67.12 *** 78.69 ns 71.56 ** 70.13 ** 69.24
Pairs
compared Control p < Tree 12 p < Tree 14 p < Tree 16 p < Tree 18
λ1/λ1c 3.64 *** 4.37 ** 3.69 *** 5.46 * 3.87
λ2/λ2c 4.15 *** 4.82 ns 4.33 *** 5.68 ** 4.42
λ3/λ3c 8.16 ** 8.87 * 8.86 *** 9.83 *** 8.94
λ4/λ4c 10.43 *** 11.20 ns 11.43 ** 12.11 *** 11.69
λ5/λ5c 11.01 *** 11.82 * 12.14 *** 12.65 *** 12.43
λ6/λ6c 4.53 *** 5.12 ns 4.79 *** 5.85 ** 4.82
λ7/λ7c 12.72 ** 13.92 ns 14.18 ** 14.77 *** 14.72
λ8/λ8c 33.85 ns 37.55 ** 38.32 *** 36.41 *** 37.48
λ9/λ9c 41.66 *** 46.45 ns 47.43 *** 44.15 *** 45.15
λ10/λ10c 67.12 ** 75.60 *** 78.29 ** 68.54 ns 68.16
ns – no significance between obtained differences (p > 0.05); * – p < 0.05; ** – p < 0.01;
*** – p < 0.001
The averaged SRC of control tree 17 and of one tree from each of the two
clusters, shown in Fig. 2 are displayed in Fig. 3a. Figure 3billustrates the red-
edge analysis performed on the same set of trees. In the derivative spectrum
the red-edge inflection point is defined from the maximum of the peak lying be-
tween 680 and 770 nm. The first derivative was calculated using a first-difference
transformation of the reflectance spectrum obtained from the polynomial fit. The
maximum of the derivative of control SRC is located at 721.65 nm while the cor-
responding peak for the infected trees 4 (cluster 2) and 18 (cluster 3) is positioned
at 719.98 nm and 717.98 nm, respectively. The steady shift in peak position to
shorter wavelengths is an indicator for the presence of the infection.
The results from DAS-ELISA test, shown in Fig. 4, revealed presence of
ASGV in all infected groups of leaves. According to extinction values (EV)
the trees were divided into three groups, healthy (EV <0.400), feebly infected
180 D. Krezhova, A. Stoev, S. Maneva

Fig. 4. Results of DAS-ELISA test on leaf samples from apple trees, cultivar Florina, infected
with ASGV
(0.400 <EV <0.500) and infected (EV > 0.500). Three of the trees had extinc-
tion values near to the negative control (−K). These were control trees. Four of
the trees had EV near the cut off value (0.400) and for the rest three cases EV
was higher than 0.500.
Conclusions. Hyperspectral remote sensing technique of leaf reflectance
was tested for detection of latent infection in young apple trees caused by apple
stem grooving virus (ASGV). It was demonstrated that the spectral behaviour of
the leaves from healthy and infected apple trees was clearly different. Hyperspec-
tral reflectance data were analyzed by means of statistical (Student’s t-criterion,
cluster) and first derivative analyses. In the red edge region a shift of the SRC
values of infected leaves towards the shorter wavelengths was observed. This is
reliable indicator for presence of viral infection. For assessment of the presence
and the degree of the viral infections serological analysis via DAS-ELISA test
was applied on leaf samples from the same trees. The results from the two ap-
plied techniques were subjected to comparative analysis. It was found that the
spectral behaviour of the investigated group of apple leaves corresponded to the
assessed degree of the infection. The results show the efficiency and sensitivity of
hyperspectral reflectance for revelation of stress and diseases, as well as for early
diagnosis of symptoms in plants at different stages of infections.
Compt. rend. Acad. bulg. Sci., 68, No 2, 2015 181

REFERENCES
[1]Carter G. A., A. K. Knapp (2001) Am. J. Bot., 88, No 4, 677–684.
[2]Meier U. (2001) In: Growth stages of mono and dicotyledonous plants (ed.
U. Meier). Pome Fruits Federal Biological Research Centre for Agriculture and
Forestry, 158.
[3]Pawar S. S. (2009) In: State of Indian Agriculture. An address for Foundation
Day Celebration 2009, National Academy of Agricultural Sciences, New Delhi.
[4]Yanase H. (1983) Acta Hortic., 130, 117–122.
[5]Tiziano C. et al. (2003) Crop Prot., 22, 1149–1156. doi: 10.1016/s0261-
2194(03)00156-x.
[6]Kundu J. K. (2003) Plant Prot. Sci., 39, 88–92.
[7]Shim H. et al. (2004) Moleculars and Cells, 18, 192–199.
[8]Cembali T. et al. (2003) Crop Prot., 22, 1149–1156.
[9]Chen S. et al. (2014) PLoS ONE, 9, No 4, e95239, doi:10.1371/journal.pone.
[10]Nemeth M. (1986) Virus, Mycoplasma and Rickettsia Diseases of Fruit Trees.
Martinus Nijhoff/Dr. W. Junk Publ., Boston, Lancaster, Dordrecht, 841.
[11]Lister R. M. (1970) CMI/AAB Descriptions of Plant Viruses, No 31. Common-
wealth Agricultural Bureaux, Kew, U.K.
[12]Welsh M. F., F. A. Van der Meer. (1989) In: Virus and Virus-like Diseases of
Pome Fruits and Simulating Noninfectious Disorders (ed. W. A. Pullman). College
of Agriculture and Home Economics, Washington State University, 253–267.
[13]Kundu J. et al. (2003) Plant Prot. Sci., 39, No 3, 93–96.
[14]Tagliavini M., A. D. Rombola (2001) Eur. J. Agron., 15, 71–92.
[15]Smith K. L. et al. (2005) Int. J. Remote Sens., 26, 4067–4081.
[16]Krezhova D. D. et al. (2012) Proc. SPIE 8531, Remote Sensing for Agriculture,
Ecosystems, and Hydrology, XIV, 85311H, doi:10.1117/12.974722.
[17]Delalieux S. et al. (2007) Eur. J. Agron, 27, 130–143, http://dx.doi.org/10.
1016/j.eja.2007.02.005
[18]Carter G. A., A. K. Knaap (2001) Am. J. Bot., 88, 677–684.
[19]Delalieux S. et al. (2006) Airborne Imaging Spectroscopy Workshop BruHyp
2006, 1–16.
[20]Duin R. P. W. et al. (1997) Pattern Recognition Letters, 18, 1159–1166.
Space Research and Technology Institute
Bulgarian Academy of Sciences
Acad. G. Bonchev St, Bl. 1
1113 Sofia, Bulgaria
e-mail:dora.krezhova@gmail.com
∗Institute of Soil Science
Agrotechnology and Plant Protection
Agricultural Academy, Sofia
Division of Plant Protection
2230 Kostinbrod, Bulgaria
182 D. Krezhova, A. Stoev, S. Maneva