lable at ScienceDirect

Trends in Analytical Chemistry 118 (2019) 43e49

Contents lists avai

Trends in Analytical Chemistry

journal homepage: www.elsevier .com/locate/ t rac

Advanced spectroscopic techniques for plant disease diagnostics. Areview

Charles Farber a, Mark Mahnke a, Lee Sanchez a, Dmitry Kurouski a, b, *

a Department of Biochemistry and Biophysics, Texas A&M University, College Station, TX, 77843, United Statesb The Institute for Quantum Science and Engineering, Texas A&M University, College Station, TX, 77843, United States

a r t i c l e i n f o

Article history:Available online 23 May 2019

Keywords:Plant diseasesPathogen diagnosticsRaman spectroscopyInfrared spectroscopySurface-enhanced Raman spectroscopyPlant pathology

* Corresponding author. Department of BiochemistrUniversity, College Station, TX, 77843, United States.

E-mail address: dkurouski@tamu.edu (D. Kurousk

https://doi.org/10.1016/j.trac.2019.05.0220165-9936/© 2019 Elsevier B.V. All rights reserved.

a b s t r a c t

Timely plant pathogen diagnostics can save up to 50% of the total agricultural yield worldwide. Currentmolecular and imaging methods for detection and identification of plant diseases have many limitations.This sparked interest in the development of minimally invasive and substrate general spectroscopictechniques that can be used directly in the field for confirmatory plant disease diagnostics. This reviewdiscusses recent progress in development of reflectance, infrared, Raman and surface-enhanced Ramanspectroscopy for detection and identification of plant diseases. It also shows advantages and disadvan-tages of these optical spectroscopy methods compared to the most common molecular and imagingtechniques.

© 2019 Elsevier B.V. All rights reserved.

1. Introduction

As the human populations grows from its current size of 7billion to the projected 9.7 billion in 2050 and 11.2 billion in 2100,we will need to produce 70e100% more food in the near future [1].One way to address this issue is to increase the amount of agri-cultural areas. Although many developing countries follow thisstrategy, it has limited applicability and essentially is destructive tonature [2]. The second strategy is focused on increasing the yield ofexisting crops. This includes directed plant breeding and timelydiagnostics of plant biotic and abiotic stresses [3]. Currently, bioticstress accounts for about 11.5% of crop loss in the United States, upto about 1e2% higher in other places around the world [4]. Totalbiotic stress caused by various pathogens, pests, and weeds ac-counts for 20e40% of agricultural productivity loss worldwide [5].To help minimize these massive losses in food, several molecularand imaging methods are employed to enable detection of diseasesat early stages. However, when it comes to disease, time and reli-ability are of the essence and these requirements are not both metby current methods of analysis. These issues have cultivated in-terest in alternative methods, such as infrared (IR) and Raman (RS)spectroscopies, for rapid confirmatory detection and identificationof plant pathogens in the field.

y and Biophysics, Texas A&M

i).

Various measures to prevent and contain crop diseasesincluding treatment with pesticides, use of genetically modifiedplants and timely removal of diseased plants have had variablesuccess [6]. Chemical treatment by pesticides and fungicides can beused to prevent development of plant diseases as well as reduceamount of pest that feed on them. One study showed that in 22crops, application of fungicides improved yield by ~50%, equivalentto about 44 billion kilograms of food or 12.8 billion U.S. dollars’worth [7]. Using pesticides, however, can potentially harm bothhumans and the environment. Development of genetically modi-fied organisms (GMOs) is an alternative approach which allowsproduction of transgenic crops with pathogen-specific resistancecoded into their DNA. Although using transgenic crops has raisedyields by 22%, increasing profits by 68%, according to a review of147 studies [8], the potential risk of introducing transgenic plansinto the ecosystem is not yet fully understood. Lastly, removinginfected plants in order to contain the spread of the disease may becostly, and failure to detect a disease outbreak at its onset oftenmakes this containmentmeasure impractical. The limitations of thecurrent disease prevention and containment practices furtherelevate the demand for reliable, fast and cost-effective methods forearly-stage detection of crop diseases.

2. Current methods in plant disease detection

A variety of methods are currently used in the agricultural in-dustry to detect plant pathogens (Table 1). These techniques can be

Table 1Currently available molecular, imaging and spectroscopic methods for plant pathogen diagnostics.

Method Sensitivity Specificity Limitations Advantages Time ofanalysis

Portability

Molecular MethodsPolymerase chain reaction (PCR) and

quantitative polymerase chain reaction(qPCR)

High Yes Requires DNA sequencing and design ofprimers. Sensitive to organiccontaminants. Labor intensive.

Requires small sample amount.Provides high specificity andaccuracy

Long No

Enzyme-linked immunosorbent assay(ELISA)

Medium Yes Photobleaching Easy to operate. Low cost Long No

Fluorescence in-situ hybridization (FISH),immunofluorescence imaging (IFI), flowcytometry (FCM)

Medium Yes Autofluorescence and photobleaching High specificity and accuracy ofdetection

Long No

Imaging MethodsThermography Low No High cost. Poor specificity. Requires

calibration.Probes large agricultural areas.Can be used to detect “danger”areas.

Long Yes

RGB and fluorescence imaging Low No Poor specificity. Requires calibration. Probes large agricultural areas.Can be used to detect “danger”areas.

Long Yes

Hyperspectral Imaging Low No High cost, complexity, and long dataacquisition. Sophisticated and time-consuming data analysis.

Probes large agricultural areas. Long Yes

Spectroscopic MethodsVisible (VIS) and Near-Infrared (NIR)

Transmittance and Reflectancespectroscopy

Low No Poor specificity Fast and inexpensive. Fast Yes

Infrared Spectroscopy (IR) High Yes Presence of water Fast and accurate. Fast YesRaman Spectroscopy (RS) High Yes Highly fluorescent samples. Non-invasive and non-

destructive. Fast and accurate.Fast Yes

Surface-enhanced Raman Spectroscopy(SERS)

High Yes Signal fluctuations. Destructive. Single-molecule sensitivity.Fast and accurate

Fast Yes

C. Farber et al. / Trends in Analytical Chemistry 118 (2019) 43e4944

divided on three groups: molecular, imaging and spectroscopicmethods. Sensitivity is the capacity of a method to correctly iden-tify its target (in this case, a particular pathogen), while specificityis the method's ability to correctly identify when its target is notpresent [9].

2.1. Molecular methods

Molecular methods used for the detection and identification ofplant diseases include polymerase chain reaction (PCR), enzyme-linked immunosorbent assay (ELISA), fluorescence in situ hybridi-zation (FISH), immunofluorescence, aptamer-based diagnostics andflow cytometry. These methods are based on either direct detectionof the pathogen or associated molecules, such as DNA or toxins. Thecurrent reviewwill only briefly discuss the first twomethods due totheir broad application in plant pathology. Detailed discussions ofmolecular methods can be found in several excellent reviews byFang and Ramasamy [10], Martinelli and co-authors [11], as well asDonoso and Valenzuela [12].

PCR and, in particular, quantitative (q)PCR, are the work horsesof plant pathology [13,14]. PCR has high sensitivity, which enablesearly stage disease detection, and high specificity, which ensuresprecise identification of fungal, bacterial and viral pathogens[15e17]. However, PCR is a labor-intensive and destructive tech-nique that requires extensive sample preparation and is generallyperformed in a laboratory by highly skilled personnel. The methodinvolves sample homogenization, extraction and purification of theDNAmaterial, as well as gel electrophoresis and gel imaging, whichadd to the cost and duration of analysis of individual samples.Furthermore, PCR requires primer preparation and consequentlydetermination of the DNA sequence of the pathogen of interest.Despite great progress in optimization of PCR-based disease di-agnostics over the last few decades, this technique remainsexpensive for routine detection and identification of plant diseases.

Also, PCR is highly sensitive to contamination from environmentalDNA, can be inhibited by small quantities of organic solvents [18].

ELISA is based on using specially raised bodies to bind antigensof interest to confirm their presence; it is typically a colorimetricassay [19]. Like PCR, this assay is invasive and destructive, but ELISAgenerally has lower sensitivity. ELISA is also highly sensitive tophotobleaching of chemical dyes used in this technique. Since thismethod's inception in the early 1970s [20], many modifications tothis method have been made such as voltammetric detection andthe use of electrical-impedance spectroscopy (EIS). Others retainedthe optical approach using Lateral Flow Immunoassays (LFI) orsurface plasmon resonances (SPR) [21].

2.2. Imaging methods

Imaging methods include thermography, hyperspectral, as wellas RGB and fluorescence imaging. Their coupling to unmannedaerial vehicle (UAV) allows for monitoring of large agriculturalareas. These techniques enable indirect disease diagnostics viadetected changes in color, texture or temperature of the plant.

Thermography detects the heat emitted by objects and surveylarge areas in a single image. Because pathogens induce metabolicand structural changes in the host, both associated with temper-ature, thermography can be used to detect plant diseases andother stresses [22]. For instance, infection of tobacco by tobaccomosaic virus causes closing of the plant stomata. This affectstranspiration rates and results in overheating of the plant [22,23].The change in plant temperature, relative to either the tempera-ture of adjacent plant parts or the temperature of a referencehealthy plant, can be detected by thermal imaging cameras. Thistechnique has also been shown to be able to detect stress-inducedthermal changes in plants [24]. Thermography, while capable atsurveying large areas at a time, is non-specific and only sensitiveto some types of pathogens, is entirely non-invasive. It requires

Fig. 1. Reflectance Vis-near IR spectra of healthy and HLB infected citrus leaves.Reproduced with permission from Ref. [28].

C. Farber et al. / Trends in Analytical Chemistry 118 (2019) 43e49 45

only a thermal scan of the plant surface, which requires nocontact.

RGB and fluorescence imaging detect pathogens by probingchanges in transmittance or fluorescence of various plant partscaused by pigmentation. In RGB imaging, a series of photographsare consecutively taken is through red, green and blue filters. Likethermography, images taken of a plant at multiple points in timecan be compared to each other in order to see potential differencesin the levels of these colors. These differences can be caused by achange in intensity in the transmittance of these color by thevarious pigments in a plant [25] and therefore can be used toidentify certain pathogens.

Chlorophyll fluorescence imaging measures the difference influorescence intensity across plant area or relative to the signalfrom surrounding plants. Plant fluorescence typically decreasesunder stressful conditions [26]; this method is therefore non-invasive and non-destructive; however, it is also non-confirmatory as many very similar changes in plant color andtexture can be induced by abiotic and biotic stresses.

A more advanced imaging technique known as hyperspectralimaging utilizes reflectance data collected over a wide spectralrange, typically 350e2500 nm, to reconstruct a spatial image of theplant leaf analyzed with special image processing methods [22].Being an information rich method, it enables detection of a widerrange of diseases compared to RGB imaging. Among disadvantagesof hyperspectral imaging are high cost, complexity, long dataacquisition times and complex data analysis, which all togetherlimit its practical application especially in cases where quickresponse times or screening of large areas are desired [27].

3. Spectroscopic methods

Visible and Near-Infrared (Vis-NIR) Transmittance andReflectance spectroscopy requires easily accessible and inexpen-sive experimental setup which is essentially based on a lamp andthe photodetector. In reflectance mode (RM), the detector and lightsource are located on the same side of the sample, whereas intransmittance mode (TM) the detector is located on the oppositeside of the illumination. Depending on broadness of the range ofelectromagnetic radiation of the lamp, spectral acquisition can beachieved in visible and near-infrared regions. Vis-NIR spectroscopyis based on determination of the wavelengths of light a sample bestabsorbs, reflects or transmits in the Vis and NIR ranges, which areassociated with the color of the sample. This simplicity and smallsize of currently available commercial VIS spectrometers attractedresearchers to explore suitability of this technique for detection andidentification of plant diseases. Sankaran and co-workers used VIS-NIR RM spectroscopy for detection of Huanglongbing (HLB), orcitrus greening, on citrus trees [28]. HLB is a devastating disease ofcitrus trees caused by the gram-negative Candidatus Liberibacterasiaticus, a phloem-limited species vectored by the Asian citruspsyllid, Diaphorina citri, and the African citrus psyllid, Trioza ery-treae. Infected trees exhibit yellowing of leaves, premature leaf andfruit drop, and ultimately, death of the entire plant. The researchersshowed that in combination with multivariate statistical analysis,this approach was able to provide high accuracy of HLB diagnostics.Additionally, several claims have been made that RM VIS-NIRspectroscopy could detect bacterial contamination of spinachleaves, spot disease on wheat and olive leaf spot on olives [29e32].These works showed the potential of VIS-NIR spectroscopy as anon-invasive and non-destructive approach for plant disease di-agnostics. However, several very important questions remainunanswered: 1) can VIS-NIR spectroscopy detect diseases beforetheir visual appearance on plants; 2) can this technique disentangleplant disease and results of abiotic stresses, such as drought, if they

have similar visual appearance on the plant; and 3) can severaldiseases be simultaneously detected on the plant.

Based on the careful analysis of the reported results, one canspeculate that it is impossible to answer these questions with TMand RM VIS-NIR spectroscopy. This is primarily because VIS-NIRdoes not probe structural changes in the plant but rather detectschanges in colors which are not necessarily disease specific. Sec-ondly, TM and RM VIS-NIR spectra have broad bands with fluctu-ating intensities (Fig. 1). This drastically decreases the accuracy ofthe disease diagnostics.

Vibrational spectroscopy is a class of analytical techniques thatincludes IR and RS. Both provide information about the chemicalstructure of analyzed samples. They are also, in general, label-free,non-invasive and non-destructive, which makes them ideal forvarious applications in food science, biochemistry, electrochem-istry andmedicine. Thesemethods offer an unparalleled capacity toprobe structural changes in plants associated with stressors such asdisease. Application of vibrational spectroscopy for plant science ingeneral has been previously described in a review by Skolik and co-workers [33].

Infrared spectroscopy, including near- and mid-IR, is broadlyused for analysis food, grains, animal feeds, minerals and soils. It isbased on absorption of infrared radiation corresponding in energyto molecular vibrations. Consequently, the IR absorption spectrumwill reflect chemical and structural organization of the sample[34e36]. For instance, Sankaran and co-workers recently showedthat mid-IR could be used to detect HLB and nutritional deficiencyon citrus trees [37]. The researchers achieved nearly 90% accuracyof prediction based on the collected spectra. Similar results havebeen independently reported by Hawkins and co-workers [38].Liaghat and co-workers utilized mid-IR to detect basal stem rot(BSR) disease caused by Ganoderma boninense [39]. Principalcomponent analysis of the first and second derivatives of the IRspectra enabled overall classification accuracy of 92% in predictionof mild, moderate and severe stages of BSR disease. Kos and co-workers proposed to use diffuse reflection and attenuated totalreflection (ATR) IR to detect Fusarium graminearum in maize ker-nels [40]. The researchers observed significant spectral differencebetween healthy and Fusarium-infectedmaize and 100% predictionaccuracy in ATR-IR approach. One of the major shortcomings of IR-based techniques is the need for extensive sample preparation.Water is highly IR active, so removal of water is a typical step inthese experiments. However, in some cases, intact samples can beused: Liang and co-workers demonstrated detection of a bacterialinfection of potato tubers known as zebra chip disease on intacttubers with NIR [41].

Fig. 2. Photographs (A) of healthy tomato (aec) and tomato fruit infected byG. candidum (dee). The corresponding ATR-FTIR spectra shown in B (healthy) and C(infected). Reproduced with permission from Ref. [42].

Fig. 3. Raman spectra of healthy maize kernels (green) and maize kernels infected byAspergillus niger (brown), A. flavus (blue), Diplodia spp. (black) and Fusarium spp. (red).Reproduced with permission from Ref. [48].

C. Farber et al. / Trends in Analytical Chemistry 118 (2019) 43e4946

Skolik and co-workers showed that ATR mid-IR could be usedfor non-invasive and highly accurate diagnostics of sour rotinfection caused by Geotrichum candidum on tomato fruits, Fig. 2[42]. The researchers showed that mid-IR could detect structuralchanges in a cuticle, which could be assigned to lignin or cutin, aswell as pectin and other polysaccharides. These spectral changescould be evidence of the sour rot disease at early and late stages ofits progression on the fruit. The reported results indicate thatdirect application of mid-IR can be challenging because spectralchanges were observed not only for sour rot-infected tomatoes,which took place upon pathogen proliferation on the fruit, butalso for healthy tomatoes (Fig. 2). Such spectral changes verylikely reflect metabolomic activity of the fruit itself or bacterialiving on or in the fruit.

Raman Spectroscopy (RS) is a modern analytical technique thatprovides information about molecular vibrations and, like IR, thestructure of the analyzed specimen [43,44]. In contrast to IR, Ramanspectra are generated by collection of inelastically scattered pho-tons from samples of interest. Additionally, water has a very smallRaman cross-section, allowing for spectral acquisition in live bio-logical systems such as cells and tissues [34,45]. Raman instrumentdesign can enable unique applications: addition of a telescope canincrease the spectra collection range to over 60 meters [46,47].

These advantages make RS one of the most promising tools inagriculture for detection and identification of plant diseases.

3.1. Raman-based detection and identification of fungal diseases

Recently, Farber and Kurouski showed that RS can be used todetect and identify several different fungal pathogens on maizekernels [48]. For this study, the researchers used a hand-heldRaman spectrometer equipped with a 1064 nm laser. Farber andKurouski found that spectral changes associated with Fusariumspp., Diplodia spp., Aspergillus niger, and A. flavus diseases wereobserved in lignin (1600 and 1633 cm�1), a structural moleculeimportant for a hard cell wall, carotenoids (1523e1547 cm�1),pigments that gives kernels their gold color, proteins that areidentified by their characteristic amide peptide bond(1640e1670 cm�1), and carbohydrates, including monomericsugars (1115, 1082, 1052, 1043, and 1024 cm�1) and starch (1153,938, and 864 cm�1) (Fig. 3). With lignin, maize kernels showedalmost no band after Fusarium proliferation, while these band in-tensities were lower in A. niger and A. flavus.The intensity of thisband did not change in Diplodia-infected kernels. This showspathogen-specific spectroscopic responses to disease in plants. Theresearchers utilized orthogonal partial least-squares discriminantanalysis (OPLS-DA) to demonstrate that RS has high accuracy ofplant disease detection and identification. It was found thatcoupling of OPLS-DA with RS allowed for 100% accurate detectionand identification of these four pathogens on maize kernels.

Fig. 4. Leaf samples collected from fields for qPCR assay and Raman spectrum (A).Raman spectra generated from leaves of healthy (green), HLB-infected on late (blue)and early (purple) stages, as well as nutritional deficiency (ND) symptoms (red)grapefruit (B) and orange (C) trees. Spectra normalized on cellulose vibrational bands(marked by asterisks (*)). Reproduced with permission from Ref. [51].

C. Farber et al. / Trends in Analytical Chemistry 118 (2019) 43e49 47

This work was followed by Egging and co-workers who showedthat this spectroscopic approach could be extended on detection offungal diseases onwheat and sorghum grain [49]. It was found thatin the spectra of ergot-infected wheat, the amide I band had twomaxima at 1650 and 1667 cm�1. These bands were not evident inthe spectra of healthy wheat or the black tip-infected wheat. Thisobservation indicated that ergot infection in the wheat may beassociated with expression and deposition of a-helical and b-sheetproteins.

In the same work, Egging and co-authors also investigatedwhether the proposed spectroscopic approach could be used forinvestigating disease development. The researchers collected

spectra from mold-infected sorghum seeds at different stages ofdisease progression, including healthy, middle and late stage grain.They observed a decrease in intensities of lignin, carotenoids andcarbohydrates vibrational bands in healthy, middle, and late stageprogression of mold on sorghum seeds.

3.2. Raman-based detection of pests in beans

Many pests or pathogens of plants live within their hosts. Toaddress whether RS can detect these hidden threats, Sanchez andcolleagues studied Callosobruchus maculatus, an insect pest of cowpeas [50]. These insects lay their nearly invisible eggs on the surfaceof the peas; upon hatching, the larvae burrow into the seeds andconsume them before emerging as adults. One breeding pair ofinsects can quickly destroy a store of cow peas, so rapid detection ofthe larvae is essential. The researchers scanned cow pea seedscontaining the insect larva at different stages of development. Itwas shown that RS could detect spectroscopic peaks specific to theinsect larvae and their excrements within the peas. Using PLS-DA,Sanchez and co-workers demonstrated that not only could thehealthy and infected peas be distinguished from one another, butalso the developmentally early and late stage larvae with high ac-curacy. These results together suggest that Raman is not limited topests and pathogens on the surface of plants, but also within them.

3.3. Raman-based detection and identification of bacterial and viraldiseases

Sanchez and co-workers recently showed that RS in combina-tion with multivariate statistical analysis enables highly accuratedetection of HLB on early and late stages as well as nutritionaldeficiency (ND) on orange and grapefruit citrus trees [51]. The ac-curacy of detection of Raman-based diagnostics of healthy vs. HLBinfected vs. nutritional is ~98% for grapefruit and ~87% for orangetrees, whereas the accuracy of early vs. late stage HLB infected is~85% for grapefruits and ~94% for oranges.

The researchers observed significant increase in the intensity of1601e1630 cm�1 bands, which has been assigned to lignin (Fig. 4).The intensities of these bands in Raman spectra collected fromgrapefruit leaves with late stage HLB infection appeared to behigher compared to the spectra of early stage HLB-infected leaves.In contrast, the intensities of these bands in Raman spectracollected from orange leaves with late and early stage HLB werenearly identical. The increase in the intensities of these bands iseven more pronounced in the spectra of both grapefruit and orangeleaves that have ND symptoms. Thus, one can envision that NDmayhave a unique vibrational fingerprint related to lignin that isdistinctly higher from Raman signatures of both healthy and HLB-infected leaves.

This work made two groundbreaking discoveries in diagnosticsof plant diseases. It showed that RS can 1) distinct biotic (diseases)and abiotic (ND) stresses on plants and 2) enable pre-symptomaticdiagnostics of HLB with high accuracy.

It should be noted that Vallejo-P�erez and co-workers alsoattempted to demonstrate RS-based detection of HLB in 2016 [52].In that study, the researchers observed only changes in the back-ground of spectra collected from HLB-positive and healthy leaves,whereas no specific bands or relative changes in intensities ofvibrational bands have been observed. Background-based spectralchanges may not be reliable and specific for confirmatory HLB di-agnostics. The work by Sanchez and co-authors revealed that bothHLB and ND had specific vibrational fingerprint in the Ramanspectra of citrus leaves. These spectral differences enable highlyaccurate detection and identification of HLB and ND.

C. Farber et al. / Trends in Analytical Chemistry 118 (2019) 43e4948

In 2017, Yeturu and co-workers investigated the possibility ofRS-based detection of Abutilon mosaic virus (AbMV) on Abutilon, aflowering shrub found in the tropics [53]. Infection by the virusmanifests as yellow spots, or mosaic patterns, on the leaves of theplant. The researchers observed a continuous decrease in the in-tensity of all vibrational bands in the Raman spectra collected fromthe plants infected with the virus. Statistical analysis performed bythe authors showed that the differences in intensities between thespectra was 99%, confirming that they were able to detect the virusbefore the symptoms appeared. At the same time, such approachshould be taken with a grain of doubt the main symptom of AbMVis the yellow mosaic pattern on the leaves. It is possible that acoherent decrease in the intensity of all bands could represent thechange in the leaf color rather than be associated with the viraldisease itself. This evidence suggests that RS based diagnosticscould be considered only if relative change in band intensities hasbeen observed. These illustrate issues with Raman-based detection,primarily related to fluorescence and absorbance. Different colorpatches will have different fluorescent backgrounds and absorbdifferent amounts of light, thus changing the spectra. These factorsshould be considered in Raman analysis of plant tissue.

3.4. Surface-Enhanced Raman Spectroscopy (SERS)-based detectionof mycotoxins

In 1977, Van Duyne discovered that Raman scattering can beamplified up to 108 by coherent oscillations of an electron cloudat the surface of nanoparticles [54]. This phenomenon, known assurface-enhanced Raman spectroscopy (SERS), allows for thesingle-molecule detection and therefore has been broadly uti-lized for detection of toxins associated with fungi [55e57].Additionally, SERS has been applied to related problems such asdetection of environmental pollutants and identification of bac-teria [58,59].

While RS and IR investigate primarily bulk-volume samples ofplant tissue for spectroscopic evidence of infection, SERS can detecttoxic metabolites produced by a pathogen directly; however, thesetoxins must first be extracted from a sample, meaning that thismethod is often invasive and destructive. This significantly restrictsutilization of SERS directly in the field. Additionally, fluctuations ofthe signal intensities, which is typical for SERS, remain the majordrawback of this spectroscopic approach [60]. Nevertheless, Wuand co-workers demonstrated that using silver nanorod (AgNR)array substrates SERS-based detection of aflatoxin (AF) B1, B2, G1and G2 could be achieved [61]. They determined the limits ofdetection (LODs) reach 5 � 10�5 mol/L for AFB1, 1 � 10�4 mol/L forAFB2, and 5 � 10�6 mol/L for both AFG1 and AFG2 in bulk solution.Lee and co-workers used SERS coupled to various methods ofmultivariate statistical analysis to detect aflatoxin in maize atconcentrations of 0e1206 mg/kg [62,63]. Zhao and co-workersproposed to use aptamer-coated Ag@Au coreeshell (CS) nano-particles (NPs) for double detection of ochratoxin A (OTA) andaflatoxin B1 in maize meal [64]. This enabled significant improve-ment in the LOD, which has been found to be as low as 0.006 ng/mLfor OTA and 0.03 ng/mL for AFB1.

Galarreta and co-workers developed aptamer-based SERSmicrofluidic sensor for the detection of OTA [65]. A thiol-modifiedaptamer of OTA was immobilized onto the 2D SERS platformenabling detection of OTA. Additionally, Li and co-workers con-structed similar aptamer-based SERS chip for the ultrasensitivedetermination of AFB1 in spiked peanuts [66]. Presence of AFB1 inthe sample, released AFB1 aptamers from the hybridization duplexstrands of AFB1 aptamers and their partial complementary strands.Next, this aptamer was immediately hybridized with hairpin DNAon the Au film of chip surface. After this, the hairpin DNA was

hydrolyzed by exonuclease III at a restriction site, leaving shortsingle-stranded DNA to capture Raman signals via hybridization onAu film and releasing complementary DNA for recycling. Althoughthis approach is very sophisticated, it demonstrated high sensitivityand good selectivity for the detection of AFB1 enabling detection ofAFB1 in the range from 1 � 10�6 mg/L to 0.4 � 10�6 mg/L. Thus,demonstrating that such aptamer-SERS sensing chips could providean important application for more rapid and portable determina-tion of mycotoxins. Development of more robust and reliable SERSsubstrates with high enhancement factors should increase sensi-tivity to toxins. Additionally, development of capture layers wouldtremendously increase SERS specificity [60].

4. Conclusions

This review critically discussed spectroscopic methods, as wellas modern molecular and imaging methods, used for plant diseasediagnostics. It showed a great potential of both RS and IR forconfirmatory, non-invasive and non-destructive detection andidentification of plant diseases. The review also revealed advan-tages and disadvantages of RS and IR, as well as other spectroscopictechniques such as SER and VIS/NIR-TM and RM spectroscopy, forthe plant disease diagnostics. One can envision that a decrease inthe cost of portable spectrometers and development of Ramantelescopes, which are capable of a standoff spectral acquisition, willfoster direct practical applications of RS in plant pathology andother areas of agriculture.

Acknowledgments

The authors are grateful for the financial support of AgrilifeResearch at Texas A&Mand acknowledge the Governor’s UniversityResearch Initiative (GURI) grant program of Texas A&M, GURI GrantAgreement No. 12-2016, M1700437.

References

[1] UN Projects World Population to Reach 8.5 Billion by 2030, Driven by Growthin Developing Countries, Arabia, 2000, p. 2015.

[2] H.K. Gibbs, A.S. Ruesch, F. Achard, M.K. Clayton, P. Holmgren, N. Ramankutty,J.A. Foley, Tropical Forests Were the Primary Sources of New Agricultural Landin the 1980s and 1990s, PNAS, 2010, p. 200910275.

[3] H.C. Godfray, J.R. Beddington, I.R. Crute, L. Haddad, D. Lawrence, J.F. Muir,J. Pretty, S. Robinson, S.M. Thomas, C. Toulmin, Food security: the challenge offeeding 9 billion people, Science 327 (2010) 812.

[4] P.S. Teng, W.W. Shane, Crop losses due to plant pathogens, Crit. Rev. Plant Sci.2 (1) (1984) 21e47.

[5] S. Savary, A. Ficke, J.-N. Aubertot, C. Hollier, Crop losses due to diseases andtheir implications for global food production losses and food security, FoodSecur. 4 (4) (2012) 519e537.

[6] M. Kogan, Integrated pest management: historical perspectives and contem-porary developments, Annu. Rev. Entomol. 43 (1998) 243.

[7] L.P. Gianessi, N. Reigner, The Value of Fungicides in US Crop Production,CropLife Foundation, Washington, DC, 2005.

[8] W. Klümper, M. Qaim, A meta-analysis of the impacts of genetically modifiedcrops, PLoS One 9 (2014) 1.

[9] R. Parikh, A. Mathai, S. Parikh, G. Chandra Sekhar, R. Thomas, Understandingand using sensitivity, specificity and predictive values, Indian J. Ophthalmol.56 (2008) 45.

[10] Y. Fang, R.P. Ramasamy, Current and prospective methods for plant diseasedetection, Biosensors (Basel) 5 (2015) 537.

[11] F. Martinelli, R. Scalenghe, S. Davino, S. Panno, G. Scuderi, P. Ruisi, P. Villa,D. Stroppiana, M. Boschetti, L.R. Goulart, C.E. Davis, A.M. Dandekar, Advancedmethods of plant disease detection. A review, Agron. Sustain. Dev. 35 (2014) 1.

[12] A. Donoso, S. Zalenzuela, In-field molecular diagnosis of plant pathogens:recenttrends and future perspectives, Plant Pathol. 67 (2018) 1451.

[13] A.B. Stephen, How to speed up the polymerase chain reaction, Biomol. Detect.Quantif. 12 (2017) 10.

[14] S. Derveaux, J. Vandesompele, J. Hellemans, How to do successful geneexpression analysis using real-time PCR, Methods (San Diego Calif. Print) 50(2010) 227.

[15] N. Schaad, R. Frederick, Real-time PCR and its application for rapid plantdisease diagnostics, Can. J. Plant Pathol. 24 (3) (2002) 250e258.

C. Farber et al. / Trends in Analytical Chemistry 118 (2019) 43e49 49

[16] M. Liu, E. McCabe, J.T. Chapados, J. Carey, S.K. Wilson, R. Tropiano, S.A. Redhead,C.A. Levesque, S. Hambleton, Detection and identification of selected cereal rustpathogens by TaqMan® real-time PCR, Can. J. Plant Pathol. (2015) 92.

[17] W. Zhang, Z. Zhang, G. Fan, Y. Gao, J. Wen, Y. Bai, C. Qiu, S. Zhang, Y. Shen,X. Meng, Development and application of a universal and simplifiedmultiplex RT-PCR assay to detect five potato viruses, J. Gen. Plant Pathol.(2017) 33.

[18] C. Schrader, A. Schielke, L. Ellerbroek, R. Johne, PCR inhibitors e occurrence,properties and removal, J. Appl. Microbiol. 113 (2012) 1014.

[19] M.F. Clark, A.N. Adams, Characteristics of the microplate method of enzyme-linked immunosorbent assay for the detection of plant viruses, J. Gen. Virol.34 (1977) 475.

[20] E. Engvall, P. Perlmann, Enzyme-linked immunosorbent assay (ELISA) quan-titative assay of immunoglobulin G, Immunochemistry 8 (1971) 871.

[21] M. Khater, A. de la Escosura-Mu~niz, A. Merkoçi, Biosensors for Plant PathogenDetection, Elsevier Science B.V., Amsterdam. Great Britain, 2017, p. 72.

[22] A.K. Mahlein, E.C. Oerke, U. Steiner, H.W. Dehne, Recent advances in sensingplant diseases for precision crop protection, Eur. J. Plant Pathol. (2012) 197.

[23] S.-e.-A. Raza, G. Prince, J.P. Clarkson, N.M. Rajpoot, Automatic detection ofdiseased tomato plants using thermal and stereo visible light images, PLoSOne 10 (2015) e0123262.

[24] L. Chaerle, W.V. Caeneghem, E. Messens, H. Lambers, M. Van Montagu, D. VanDer Straeten, Presymptomatic visualization of plantevirus interactions bythermography, Nat. Biotechnol. 17 (1999) 813.

[25] H.E. Nilsson, Remote sensing and image analysis in plant pathology, Annu.Rev. Phytopathol. 33 (1995) 489.

[26] A.M. Mutka, R.S. Bart, Image-based phenotyping of plant disease symptoms,Front. Plant Sci. 5 (2015) 734.

[27] A.A. Gowen, C. Odonnell, P. Cullen, G. Downey, J. Frias, Hyperspectral imaginge an emerging process analytical tool for food quality and safety control,Trends Food Sci. Technol. 18 (2007) 590.

[28] S. Sankaran, A. Mishra, J.M. Maja, R. Ehsani, Visible-near infrared spectroscopyfor detection of Huanglongbing in citrus orchards, Comput. Electron. Agric. 77(2011) 127.

[29] N. Abu-Khalaf, M. Salman, Detecting plant diseases using visible/near infraredspectroscopys, NIR News 24 (2013) 12.

[30] A. Larsolle, H. Hamid Muhammed, Measuring crop status using multivariateanalysis of hyperspectral field reflectance with application to disease severityand plant density, Precis. Agric. 8 (2007) 37.

[31] N. Abu-Khalaf, Sensing tomato's pathogen using Visible/Near infrared (VIS/NIR) spectroscopy and multivariate data analysis (MVDA), Palest. Tech. Univ.Res. J. 3 (2015) 12.

[32] Y. Zhao, F. Qin, P. Cheng, X. Li, Z. Ma, L. Zhao, J. Li, H. Wang, Quantitativedetermination of germinability of Puccinia striiformis f. sp. tritici urediosporesusing near infrared spectroscopy technology, J. Spectrosc. 2015 (2015) 1.

[33] P. Skolik, M.R. McAinsh, F.L. Martin, in: J. Lopes, C. Sousa (Editors), ChapterTwo - Biospectroscopy for Plant and Crop Science, Elsevier, 2018, p. 15.

[34] N.B. Colthup, L.H. Daly, S.E. Wiberley, Introduction to Infrared and RamanSpectroscopy, third ed., Academic Press, 1990.

[35] Z. Chu, F. Xuping, W. Jian, L. Fei, H. Yong, Z. Weijun, Mid-infrared spectroscopycombined with chemometrics to detect Sclerotinia stem rot on oilseed rape(Brassica napus L.) leaves, Plant Methods 13 (2017) 1.

[36] V. Erukhimovitch, L. Tsror, M. Hazanovsky, M. Huleihel, Direct identification ofpotato's fungal phyto-pathogens by Fourier-transform infrared (FTIR) micro-scopy, J. Spectrosc. 24 (2010).

[37] S. Sankaran, R. Ehsani, E. Etxeberria, Mid-infrared spectroscopy for detectionof Huanglongbing (greening) in citrus leaves, Talanta 83 (2010) 574.

[38] S.A. Hawkins, B. Park, G.H. Poole, T. Gottwald, W.R. Windham, K.C. Lawrence,Detection of citrus Huanglongbing by Fourier transform infrared-attenuatedtotal reflection spectroscopy, Appl. Spectrosc. 64 (2010) 100.

[39] S. Liaghat, S. Mansor, R. Ehsani, H. Zulhaidi, M. Shafri, S. Meon, S. Sankaran,Mid-infrared spectroscopy for early detection of basal stem rot disease in oilpalm, Comput. Electron. Agric. 101 (2014) 48.

[40] G. Kos, H. Lohninger, R. Krska, Development of a method for the determina-tion of Fusarium fungi on corn using mid-infrared spectroscopy with atten-uated total reflection and chemometrics, Anal. Chem. 75 (2003) 1211.

[41] P.-S. Liang, R.P. Haff, S.-S.T. Hua, J.E. Munyaneza, T. Mustafa, S.B.L. Sarreal,Nondestructive detection of zebra chip disease in potatoes using near-infrared spectroscopy, Biosyst. Eng. 166 (2018) 161.

[42] P. Skolik, M.R. McAinsh, F.L. Martin, ATR-FTIR Spectroscopy Non-destructivelyDetects Damage-Induced Sour Rot Infection in Whole Tomato Fruit, Planta,2018.

[43] R. McCreery, Raman Spectroscopy for Chemical Analysis, Wiley, New York,2000.

[44] D.A. Long, The Raman Effect: A Unified Treatment of the Theory of RamanScattering by Molecules, West Sussex, England, 2002.

[45] K. Kong, C.J. Rowlands, S. Varma, W. Perkins, I.H. Leach, A.A. Koloydenko,H.C. Williams, I. Notingher, Diagnosis of tumors during tissue-conservingsurgery with integrated autofluorescence and Raman scattering microscopy,Proc. Natl. Acad. Sci. U. S. A. 110 (2013) 15189.

[46] S.K. Sharma, New trends in telescopic remote Raman spectroscopic instru-mentation, Spectrochim. Acta A Mol. Biomol. Spectrosc. 68 (2007) 1008.

[47] S.M. Angel, N.R. Gomer, S.K. Sharma, C. McKay, Remote Raman spectroscopyfor planetary exploration: a review, Appl. Spectrosc. 66 (2012) 137.

[48] C. Farber, D. Kurouski, Detection and identification of plant pathogens onmaize kernels with a hand-held Raman spectrometer, Anal. Chem. 90 (2018)3009.

[49] V. Egging, J. Nguyen, D. Kurouski, Detection and identification of fungal in-fections in intact wheat and sorghum grain using a hand-held Raman spec-trometer, Anal. Chem. 90 (2018) 8616.

[50] L. Sanchez, C. Farber, J. Lei, K. Zhu-Salzman, D. Kurouski, Noninvasive andnondestructive detection of cowpea bruchid within cowpea seeds with ahand-held Raman spectrometer, Anal. Chem. 91 (2019) 1733.

[51] L. Sanchez, S. Pant, Z. Xing, K. Mandadi, D. Kurouski, Rapid and noninvasivediagnostics of Huanglongbing and nutrient deficits on citrus trees with ahandheld Raman spectrometer, Anal. Bioanal. Chem. (2019). https://doi.org/10.1007/s00216.

[52] M.R. Vallejo-P�erez, M.G. Mendoza, M.G. Elias, F.J. Gonzalez, H.R. Contreras,C.C. Servin, Raman spectroscopy an option for the early detection of citrushuanglongbing, Appl. Spectrosc. 70 (2016) 829.

[53] S. Yeturu, P. Vargas Jentzsch, V. Ciobot�a, R. Guerrero, P. Garrido, L.A. Ramos,Handheld Raman spectroscopy for the early detection of plant diseases:Abutilon mosaic virus infecting Abutilon sp, Anal. Meth. 8 (2016) 3450.

[54] D.L. Jeanmaire, R. Van Duyne, Surface Raman spectroelectrochemistry: Part I.Heterocyclic, aromatic, and aliphatic amines adsorbed on the anodized silverelectrode, J. Electroanal. Chem. Interfac. Electrochem. 84 (1977) 1.

[55] J. Yuan, C. Sun, X. Guo, T. Yang, H. Wang, S. Fu, C. Li, H. Yang, A rapid Ramandetection of deoxynivalenol in agricultural products, Food Chem. 221 (2017)797.

[56] T.-t. Pan, D.-W. Sun, H. Pu, Q. Wei, Simple approach for the rapid detection ofalternariol in pear fruit by surface-enhanced Raman scattering with pyridine-modified silver nanoparticles, J. Agric. Food Chem. 66 (2018) 2180.

[57] K.-M. Lee, T.J. Herrman, Determination and prediction of fumonisin contam-ination in maize by surfaceeenhanced Raman spectroscopy (SERS), FoodBioprocess Tech. 9 (2016) 588.

[58] S. Pahlow, S. Meisel, D. Cialla-May, K. Weber, P. R€osch, J. Popp, Isolation andidentification of bacteria by means of Raman spectroscopy, Adv. Drug. Deliv.Rev. 89 (2015) 105.

[59] D.-W. Li, W.-L. Zhai, Y.-T. Li, Y.-T. Long, Recent progress in surface enhancedRaman spectroscopy for the detection of environmental pollutants, Micro-chim. Acta 181 (2014) 23.

[60] D. Kurouski, H. Lee, F. Roschangar, C. Senanayake, Surface-enhanced Ramanspectroscopy: from concept to practical application, Spectroscopy 32 (2017)36.

[61] X. Wu, S. Gao, J.S. Wang, H. Wang, J.W. Huang, Y. Zhao, The surface-enhancedRaman spectra of aflatoxins: spectral analysis, density functional theorycalculation, detection and differentiation, Analyst 137 (2012) 4226.

[62] K.M. Lee, J. Davis, T.J. Herrman, S.C. Murray, Y. Deng, An empirical evaluationof three vibrational spectroscopic methods for detection of aflatoxins inmaize, Food Chem. 173 (2015) 629.

[63] K.M. Lee, T.J. Herrman, Y. Bisrat, S.C. Murray, Feasibility of surface-enhancedRaman spectroscopy for rapid detection of aflatoxins in maize, J. Agric. FoodChem. 62 (2014) 4466.

[64] Y. Zhao, Y. Yang, Y. Luo, X. Yang, M. Li, Q. Song, Double detection of myco-toxins based on SERS labels embedded Ag@Au coreeshell nanoparticles, Appl.Mater. Interfaces 7 (2015) 21780.

[65] B.C. Galarreta, M. Tabatabaei, V. Guieu, E. Peyrin, F. Lagugne-Labarthet,Microfluidic channel with embedded SERS 2D platform for the aptamerdetection of ochratoxin A, Anal. Bioanal. Chem. 405 (2013) 1613.

[66] Q. Li, Z. Lu, X. Tan, X. Xiao, P. Wang, L. Wu, K. Shao, W. Yin, H. Han, Ultra-sensitive detection of aflatoxin B1 by SERS aptasensor based on exonuclease-assisted recycling amplification, Biosens. Bioelectron. 97 (2017) 59.