A review on systems biology approaches in agricultural entomology

Introduction

The study of host plant resistance to insects has always sat uneasily between physiology and genetics. Painter's (1951) [15] original classification of resistance into antixenosis, antibiosis and tolerance gave breeders a working vocabulary, but it said little about the biochemical machinery underlying any of the three categories, and for several decades resistance screening remained largely phenotypic, built on damage scores, survival counts and development times recorded under controlled infestation. The biochemical turn came gradually, as researchers began to associate resistance with measurable enzyme activities and secondary metabolite concentrations rather than with damage symptoms alone (Kogan and Ortman, 1978) [12] . What has changed more recently is not the recognition that biochemistry underlies resistance, which was already accepted by the 1980s, but the scale at which biochemical and molecular data can now be collected and related to one another. Systems biology, as a discipline distinct from molecular biology, was defined by its insistence that biological function emerges from interactions among components rather than from any single component in isolation (Kitano, 2002) [11] . Ideker et al. (2001) [8] described this as a shift from asking what a gene does to asking how a network of genes, proteins and metabolites produces a phenotype, and this reframing travelled into plant-insect biology once genome sequences for major crop pests became available. The sequencing of the pea aphid genome by the International Aphid Genomics Consortium (2010) [9] and of the brown planthopper genome by Xue et al. (2014) [21] gave entomologists, for the first time, a reference against which transcriptomic and proteomic data from resistant and susceptible host genotypes could be interpreted mechanistically rather than descriptively. The appeal of a systems framework in agricultural entomology is not merely technical. Resistance to a given pest is rarely governed by one pathway. Howe and Jander (2008) [7] argued that plant immunity to insect herbivores draws on the same signalling networks, principally jasmonate and salicylate mediated pathways, that plants use against pathogens, which means that resistance to a chewing insect and resistance to a phloem feeding insect on the same plant can share upstream regulators while diverging sharply in their downstream biochemical output. Reviewers such as Wu and Baldwin (2010) [20] and, later, Erb and Reymond (2019) [4] extended this argument to show that the same herbivore can elicit different transcriptional and metabolic responses depending on genotype, tissue and prior exposure history, a level of context dependence that single gene studies were not designed to capture. It is this context dependence, more than any single discovery, that has made a networked, multi-layered view of resistance attractive to researchers working on rice, cotton, pulses and other field crops where breeding programmes must contend with several pests simultaneously across variable seasons. The present review does not attempt to catalogue every omics study published on crop pests, a task better suited to a database than to a narrative review. Instead it asks a narrower question relevant to breeders and physiologists working on host plant resistance, namely how far the genomic, biochemical and network based evidence currently available actually converges on a coherent, testable model of resistance, and where the evidence base is thinner than the volume of publication might suggest.

Genomic and Transcriptomic Foundations of Insect-Plant Interactions

Reference genomes for insect pests changed what a transcriptomic comparison between resistant and susceptible host genotypes could claim to demonstrate. Before a pest genome existed, differential expression studies on the host plant could report which host genes were induced or suppressed following infestation, but they could say almost nothing about the corresponding molecular events inside the insect itself. Once genomes such as those of the pea aphid, Acyrthosiphon pisum (International Aphid Genomics Consortium, 2010) [9] , and the brown planthopper, Nilaparvata lugens (Xue et al., 2014) [21] , became available, it became possible to ask whether host derived defensive compounds were being detoxified, sequestered or simply tolerated at the molecular level, rather than inferring insect physiology indirectly from survival and fecundity data. Detoxification gene families, particularly cytochrome P450 monooxygenases, carboxylesterases and glutathione Stransferases, feature prominently in this literature, and Feyereisen (2015) [5] traced how expansions in the P450 gene family across insect lineages correlate with the breadth of host plants a species can exploit, a correlation that holds reasonably well for polyphagous pests such as Helicoverpa armigera but less consistently for narrowly host specific insects. Oakeshott et al. (2010) [14] reached a related conclusion working with esterase gene families, noting that constitutive overexpression of specific esterase loci, rather than the mere presence of the gene, distinguishes insecticide resistant and host adapted populations from susceptible ones. Taken together, these findings support the view that insect side gene expression, not gene content alone, is what should be measured when comparing pest performance across host genotypes, a point that is sometimes lost in resistance screening programmes that sequence pest populations without accompanying expression data. On the plant side, transcriptomic comparisons of resistant and susceptible genotypes under insect challenge have consistently implicated the jasmonate signalling pathway, but the consistency is at the pathway level rather than at the level of individual transcripts. Mithöfer and Boland (2012) [13] and Erb and Reymond (2019) [4] both note that while jasmonate biosynthesis genes are reliably upregulated across cereal, legume and solanaceous hosts following herbivory, the specific downstream defence genes activated differ enough between species, and even between genotypes of the same species, that a defence transcript signature identified in one rice variety cannot be assumed to transfer to another without independent verification. This is a more cautious conclusion than some transcriptomic papers imply in their discussion sections, where a shared upstream pathway is occasionally presented as though it guarantees a shared resistance mechanism. The distinction matters for breeding programmes that hope to use transcriptomic markers for early selection, since a marker validated in one genetic background may simply not generalise. A further complication, raised explicitly by Walling (2000) [18] in one of the earlier syntheses of herbivore induced plant responses and echoed in subsequent work, is that chewing and phloem feeding insects elicit substantially different transcriptional programmes even on the same host plant, with chewing damage favouring jasmonate dominated responses and phloem feeding, through its more subtle wounding signature, favouring salicylate associated responses that can in some cases antagonise the jasmonate pathway. For crops such as rice, where yellow stem borer, leaf folder, brown planthopper and green leafhopper coexist as a pest complex, this antagonism has practical consequences, because a genotype bred for strong antibiosis against a chewing pest through jasmonate pathway upregulation may not confer, and in principle could even compromise, resistance to a co-occurring phloem feeder. Few multi-pest transcriptomic studies have tested this tradeoff directly in rice, which remains a conspicuous gap given how often stem borers and planthoppers are screened in the same nursery.

Proteomic and Metabolomic Layers in Biochemical Defense

The biochemical parameters most commonly reported in host plant resistance screening, peroxidase (POD), superoxide dismutase (SOD), catalase (CAT), polyphenol oxidase (PPO), malondialdehyde (MDA) as an index of lipid peroxidation, and total phenolic content, form the backbone of what might be called the classical resistance biochemistry literature, and their elevation in resistant genotypes following insect challenge has been reported with considerable regularity. War et al. (2012) [19] , in a widely cited synthesis of induced resistance mechanisms, described this suite of oxidative and phenolic responses as a coordinated defensive front, in which reactive oxygen species generated during the early stages of feeding are managed by antioxidant enzymes while phenolic oxidation, catalysed by PPO and peroxidase, produces quinones that are directly toxic or antinutritive to the feeding insect. This framework has been reproduced often enough, across pulses, cereals and cotton, that the general direction of the response, higher POD, SOD, CAT, PPO and phenol content in resistant material relative to susceptible checks, can reasonably be treated as an established pattern rather than a novel finding whenever it recurs in a new genotype screening study.

What is less settled is the interpretation of MDA, which behaves differently from the enzymes it is usually reported alongside. MDA is a product of membrane lipid peroxidation and its accumulation signals oxidative damage rather than defensive capacity, so a resistant genotype is generally expected to show lower MDA accumulation under stress, the inverse of the pattern expected for the antioxidant enzymes. Chen (2008) [3] made this distinction explicit, cautioning that treating MDA as though it simply tracked the other biochemical markers, rising uniformly with resistance, is a misreading of its biological role, yet this inversion is not always handled carefully in applied screening papers, where MDA is occasionally tabulated alongside POD and SOD without comment on the fact that a lower value, not a higher one, should indicate the more resistant genotype. This is a minor point of methodology, but it recurs often enough across published genotype comparisons to be worth stating plainly. Beyond the oxidative enzyme suite, structural and nutritional traits contribute independently to resistance, and their biochemical basis is comparatively underexplored relative to the attention given to POD and phenolics. Silica deposition in graminaceous tissue has long been associated with resistance to stem boring and leaf feeding insects through simple mechanical deterrence, abrading mandibles and reducing digestibility, and this mechanism does not depend on inducible signalling in the way enzymatic defences do, since silica accumulation is largely a function of transpiration driven passive uptake and genotype specific silicification capacity. Amino acid and free sugar content of host tissue, similarly, influence antixenosis independently of any inducible biochemical cascade, since insects use these compounds as feeding stimulants, and genotypes with lower concentrations of preferred amino acids can show reduced settling and oviposition without any measurable difference in oxidative enzyme activity. Smith (2005) [17] , reviewing molecular approaches to arthropod resistance, argued that resistance breeding programmes have historically overweighted inducible biochemical defences relative to these constitutive nutritional and structural traits, in part because enzyme assays are cheaper and faster to run at scale than tissue silica or amino acid profiling, a practical bias in method choice rather than a reflection of relative biological importance. A methodological weakness that runs through much of this literature, and one that a systems perspective is well placed to expose, is the reliance on single time point sampling. Most genotype screening studies assay biochemical parameters at one or two fixed intervals after infestation, commonly reported as measurements taken at a small number of days after treatment, and treat the resulting values as representative of the genotype's defensive capacity. Karban and Baldwin (1997) [20] , in their earlier treatment of induced plant responses, and later Mithöfer and Boland (2012) [13] both point out that induced biochemical responses are transient, rising and falling over a time course that differs between enzymes and between genotypes, so a single snapshot risks comparing genotypes at different phases of their respective response curves rather than at comparable physiological states. The practical consequence is that a genotype ranked as biochemically resistant based on one sampling day could, in principle, rank differently if sampled a few days earlier or later, a possibility that time course studies occasionally confirm but that routine screening protocols rarely test for.

Network Biology and Systems Integration in Resistance Breeding

Network based reasoning entered biology as a way of formalising the intuition that a phenotype is the output of interacting components rather than the sum of independently acting genes. Barabási and Oltvai (2004) [2] set out the conceptual case for treating cellular processes as networks with identifiable hubs and modules, arguing that a small number of highly connected nodes disproportionately determine network behaviour, and this framing has since been imported, with varying degrees of rigour, into plantinsect interaction research. In principle, a network model of host plant resistance would integrate transcriptomic, proteomic and metabolomic data from the same genotypes under the same infestation regime, identify the nodes, whether genes, enzymes or metabolites, whose variation best predicts resistance phenotype, and generate testable hypotheses about which of those nodes are causal rather than merely correlated with resistance. In practice, agricultural entomology has adopted the vocabulary of network and multi-omics integration more readily than the underlying analytical rigour. Hasin et al. (2017) [6] , writing on multi-omics approaches more broadly, set out the statistical difficulties that any such integration must confront, particularly the problem that transcriptomic, proteomic and metabolomic datasets are generated at different scales, with different noise structures, and are frequently collected from different biological replicates or even different experiments, which undermines the validity of correlating them directly. Applied to crop pest research, this caution is rarely observed in full, and a number of studies that describe themselves as integrating multi-omics data in reality report transcriptomic and biochemical results from the same genotypes side by side, in adjacent tables or figures, without the statistical integration, such as correlation network construction or pathway enrichment linking the two datasets, that the term implies. This is not a fatal weakness, since side by side reporting still has descriptive value, but it does mean that claims of systems level integration in the applied literature should be read carefully against what was actually done statistically. Where genuine network integration has been attempted, the results have generally supported the pathway level convergence described by Howe and Jander (2008) [7] , with jasmonate responsive transcription factors and their downstream enzymatic targets, including PPO and proteinase inhibitors, forming recognisable co-expression modules that correlate with resistance phenotype across genotype panels. Erb and Reymond (2019) [4] reviewed this body of work and concluded that network approaches have been more successful at confirming known pathway architecture than at discovering genuinely novel resistance mechanisms, a modest but not trivial contribution, since confirmation across independent genotype sets is itself valuable for breeders deciding which biochemical markers merit routine screening. Whether network biology will eventually deliver markers with predictive power beyond what individual enzyme assays already provide remains, on the evidence reviewed here, an open question rather than a settled achievement.

Comparative Findings and Unresolved Tensions

Not all of the literature synthesised above points in the same direction, and some of the disagreement is substantive rather than merely a matter of emphasis. One point of genuine tension concerns the relative importance of constitutive versus induced defences. Karban and Baldwin (1997) [20] argued, on the basis of comparative studies across wild and cultivated plant species, that induced responses are favoured by natural selection specifically because constitutive defence is metabolically costly to maintain in the absence of attack, implying that inducibility itself should be treated as an adaptive trait worth selecting for in breeding programmes. Other researchers working more directly with cultivated cereals and pulses have placed greater weight on constitutive traits, silica content, trichome density, cell wall thickness and baseline phenolic concentration, as the more reliable basis for resistance breeding precisely because these traits do not depend on the timing or intensity of pest pressure in a given season and are therefore more stable across environments (Panda and Khush, 1995) [16] . Neither position is simply wrong, but they imply different breeding priorities, and reviews that cite both traditions without acknowledging the tension between them understate a genuine disagreement in the field. A second area of disagreement concerns specialist versus generalist herbivores and whether host plant resistance mechanisms generalise across pest species. Ali and Agrawal (2012) [1] argued, based on a comparative synthesis across plant-herbivore systems, that specialist insects, having coevolved with a narrower range of host defences, are frequently unaffected or even stimulated by defensive compounds that deter generalist feeders, which means that a single resistance trait cannot be expected to confer broad spectrum protection against a mixed pest complex. This view sits uneasily against the practice, common in applied rice and cotton breeding papers, of screening a genotype against one target pest and then presenting the resulting biochemical profile as evidence of general pest resistance. The disagreement is less about the underlying biology, which most researchers accept, and more about how cautiously applied breeding studies extrapolate from single pest screening to claims of broader resistance. A third tension, already touched on above, concerns the reliability of transcript abundance as a predictor of the metabolite pools that actually mediate resistance. Wu and Baldwin (2010) [20] documented cases in which strong transcriptional induction of a defence related gene was not matched by a correspondingly large change in the downstream metabolite, attributing the mismatch to posttranscriptional regulation, enzyme kinetics and metabolic flux constraints that transcript level data cannot capture. This finding sits in some tension with the substantial body of applied research that uses differential gene expression alone, without accompanying biochemical or metabolomic confirmation, as sufficient evidence of a resistance mechanism. Neither body of work is necessarily incorrect, since transcript induction remains informative even where it is an imperfect proxy, but the disagreement over how much weight transcript data alone should carry is real and, in the applied breeding literature, frequently glossed over rather than argued through explicitly.

Limitations

The most consistent limitation across the literature reviewed here is the near absence of biomarker validation across independent genotype panels and across seasons. A biochemical or transcriptomic marker identified as resistance associated in one screening trial, involving a fixed set of genotypes grown at one location in one season, is regularly presented in the discussion section of the originating paper as though it were a general property of the resistant phenotype, without a follow up study in a different genetic background or growing season to test whether the association holds. Erb and Reymond (2019) [4] raised this concern directly with respect to transcriptomic markers, and the same criticism applies with at least equal force to enzyme based markers such as POD and PPO activity, which are known to be sensitive to plant age, nutrient status and ambient temperature at the time of sampling, variables that are inconsistently reported or controlled for across studies. A related gap concerns the treatment of pest complexes rather than single target pests. Field crops such as rice are rarely attacked by a single insect species in isolation, yet the great majority of biochemical and omics based resistance studies screen genotypes against one pest at a time, whether yellow stem borer, leaf folder, brown planthopper or green leafhopper, and report resistance mechanisms as though they operate independently of co-occurring pest pressure. Given the documented antagonism between jasmonate and salicylate associated defence pathways discussed earlier in this review, a genotype's biochemical response to simultaneous or sequential attack by a chewing and a phloem feeding pest could plausibly differ from its response to either pest alone, yet this scenario has been tested directly in only a limited number of published studies, most of them working with model systems rather than field crops of direct agronomic importance. A third limitation is statistical and analytical rather than experimental. As Hasin et al. (2017) [6] note in the broader multi-omics context, integrating datasets of different types requires methods, correlation network construction, multivariate ordination, pathway enrichment analysis, that are not yet routinely applied in agricultural entomology, where biochemical and transcriptomic data are still commonly analysed and reported separately even within the same study. Analytical capacity, rather than data availability, increasingly appears to be the binding constraint on progress, since the sequencing and mass spectrometry platforms required to generate multi-layered data are now widely accessible even to programmes with moderate budgets, while the statistical expertise required to integrate the resulting datasets meaningfully is less evenly distributed.

Conclusion

The evidence surveyed in this review points to a field that has accumulated a large volume of biochemical and molecular data on insect-plant interactions without yet developing the analytical infrastructure to integrate that data into a genuinely predictive systems model. The oxidative and phenolic enzyme suite, POD, SOD, CAT, PPO and total phenols, together with structural and nutritional traits such as silica deposition and amino acid content, form a reasonably well established biochemical vocabulary for describing resistance, and genomic resources for major pest species now allow this plant side biochemistry to be interpreted against the insect's own molecular response for the first time. What remains underdeveloped is validation, both across independent genotype panels and across the pest complexes that field crops actually encounter, and the statistical machinery needed to move from side by side reporting of transcriptomic and biochemical data to genuine network level integration. Future work would be better served by fewer, more thoroughly validated biomarkers than by an expanding catalogue of single study associations. A biochemical or transcriptomic marker that is measured at multiple time points after infestation, tested against more than one cooccurring pest, and confirmed across at least two independent genotype panels, would carry considerably more weight for breeding decisions than the single season, single pest screening study that currently dominates the literature. Systems biology offers agricultural entomology the conceptual tools to build such an evidence base, but the discipline has, on present evidence, made greater use of the vocabulary of systems integration than of its statistical substance, and closing that gap should be treated as a methodological priority rather than a secondary refinement.

References

1. Ali JG, Agrawal AA. Specialist versus generalist insect herbivores and plant defense. Trends in Plant Science. 2012;17(5):293-302.
2. Barabási AL, Oltvai ZN. Network biology: Understanding the cell's functional organization. Nature Reviews Genetics. 2004;5(2):101-113.
3. Chen MS. Inducible direct plant defense against insect herbivores: A review. Insect Science. 2008;15(2):101114.
4. Erb M, Reymond P. Molecular interactions between plants and insect herbivores. Annual Review of Plant Biology. 2019;70:527-557.
5. Feyereisen R. Insect P450 enzymes. In: Cohen E, editor. Advances in Insect Physiology. Vol. 49. London: Academic Press; 2015. p. 1-31.
6. Hasin Y, Seldin M, Lusis A. Multi-omics approaches to disease. Genome Biology. 2017;18:83.
7. Howe GA, Jander G. Plant immunity to insect herbivores. Annual Review of Plant Biology. 2008;59:41-66.
8. Ideker T, Galitski T, Hood L. A new approach to decoding life: Systems biology. Annual Review of Genomics and Human Genetics. 2001;2:343-372.
9. International Aphid Genomics Consortium. Genome sequence of the pea aphid Acyrthosiphon pisum. PLOS Biology. 2010;8(2):e1000313.
10. Karban R, Baldwin IT. Induced responses to herbivory. Chicago: University of Chicago Press; 1997.
11. Kitano H. Systems biology: A brief overview. Science. 2002;295(5560):1662-1664.
12. Kogan M, Ortman EF. Antixenosis: A new term proposed to define Painter's "nonpreference" modality of resistance. Bulletin of the Entomological Society of America. 1978;24(2):175-176.
13. Mithöfer A, Boland W. Plant defense against herbivores: Chemical aspects. Annual Review of Plant Biology. 2012;63:431-450.
14. Oakeshott JG, Johnson RM, Berenbaum MR, Ranson H, Cristino AS, Claudianos C. Metabolic enzymes associated with xenobiotic and chemosensory responses in Nasonia vitripennis. Insect Molecular Biology. 2010;19(Suppl 1):147-163.
15. Painter RH. Insect resistance in crop plants. New York: Macmillan; 1951.
16. Panda N, Khush GS. Host plant resistance to insects. Wallingford: CAB International; 1995.
17. Smith CM. Plant resistance to arthropods: Molecular and conventional approaches. Dordrecht: Springer; 2005.
18. Walling LL. The myriad plant responses to herbivory. Journal of Plant Growth Regulation. 2000;19(2):195216.
19. War AR, Paulraj MG, Ahmad T, Buhroo AA, Hussain B, Ignacimuthu S, et al. Mechanisms of plant defense International Journal of Advanced Biochemistry Research https://www.biochemjournal.com against insect herbivores. Plant Signaling and Behavior. 2012;7(10):1306-1320.
20. Wu J, Baldwin IT. New insights into plant responses to the attack from insect herbivores. Annual Review of Genetics. 2010;44:1-24.
21. Xue J, Zhou X, Zhang CX, Yu LL, Fan HW, Wang Z, et al. Genomes of the rice pest brown planthopper and its endosymbionts reveal complex complementary contributions for host adaptation. Genome Biology. 2014;15:521.