Background

Plant hormonomics, a term coined recently [1,2,3,4], shares a similar objective with other omics sciences: to provide comprehensive characterization of specific cellular complements. Well-established branches of omics include genomics, transcriptomics, proteomics and metabolomics [5, 6]. Systems biology combines and integrates these approaches [7]. However, the term ‘omics’ extends beyond these fields. One example is phenomics [8], which is a key discipline of plant sciences that considers plant phenotypes, primarily using high-throughput phenoty**. Another example is lipidomics, which is a distinct branch of metabolomics that deals with analysis of the lipidome. Furthermore, other fields based on the integration of several omics sciences have emerged, such as glycomics or foodomics [9,10,11,12]. Plant hormonomics, a subdivision of metabolomics, aims to achieve the qualitative or quantitative characterization of all plant hormones in a given sample.

Plant hormones are low molecular weight naturally occurring plant growth regulators. Interestingly, their production is not exclusive to plants as they are also found in microorganisms and fungi. These substances govern virtually all essential processes in a plant’s lifecycle, including germination, plant development and growth, interaction with the biotic and abiotic environment, the reproductive phase and fruit development and seed formation [13,14,15,16,17,18]. However, it is difficult to classify plant hormones definitively. Although they are low molecular weight compounds, they cannot be classified as either primary or specialized (secondary) metabolites. Primary and specialized metabolites also transfer signals; however, plant hormones are present at much lower levels, and are not dispensable like specialized metabolites [19, 20], and thus remain a distinct group of metabolites. Currently, plant hormones can be categorized into nine major classes: (I) abscisic acid (ABA) and its metabolites (collectively abscisates—ABAs), (II) auxins (Aux), (III) brassinosteroids (BRs), (IV) cytokinins (CKs), (V) ethylene (ET), (VI) gibberellins (GBs), (VII) jasmonic acid (JA, jasmonates—JAs), (VIII) salicylic acid (SA, salicylates—SAs) and (IX) strigolactones (SLs) [1, 13, 20]. Nonetheless, new sets of potent growth regulators that have hormone-like effects are being (re)discovered and are attracting increasing attention, such as indoleamines (melatonin) [21, 22], numerous apocarotenoids (anchorene, blumenols, β-cyclocitral, β-ionone, loliolide, mycorradicins, zaxinone) [23], fairy compounds [24, 25] and karrikins of exogenous origin [26] (Fig. 1). Further research and testing of these sets of compounds are necessary to unravel their function and mechanism of action.

Fig. 1
figure 1

Sets of selected plant hormones (ethylene, auxins, abscisic acid, brassinosteroids, cytokinins, gibberellins, jasmonic acid, strigolactones, and salicylic acid), their biosy nthetic precursor and catabolites (1–29) and selected plant hormone-like compounds (indoleamines, apocarotenoids, fairy compounds, karrikins) (30–40). (1) 1-aminocyclopropanecarboxylic acid (ACC); (2) malonyl-ACC; (3) indole-3-pyruvic acid; (4) indole-3-acetic acid; (5) 2-oxindole-3-acetic acid; (6) phenylpyruvic acid; (7) xanthoxin; (8) abscisic acid; (9) dihydrophaseic acid; (10) abscisic acid β-d-glucopyranosyl ester; (11) castasterone; (12) 24-epi-brassinolide; (13) 26-hydroxy-24-epi-brassinolide; (14) trans-zeatin riboside monophosphate; (15) trans-zeatin; (16) trans-zeatin-9-glucoside; (17) GA-12; (18) GA4; (19) GA34; (20) 12-oxo-phytodienoic acid; (21) jasmonic acid; (22) jasmonoyl-l-isoleucine; (23) carlactone; (24) 5-deoxystrigol; (25) orobanchol; (26) methyl carlactoate; (27) salicylic acid; (28) isochorismic acid; (29) salicylic acid glucoside; (30) tryptamine; (31) serotonin; (32) melatonin; (33) 3-hydroxymelatonin; (34) 2-aza-8-oxohypoxanthine; (35) 2-azahypoxanthine; (36) zaxinone (37) beta-ionone (38) mycorradicin; (39) karrikin 4; (40) karrikin 1

Full size image

Each class of plant hormones performs a different role and induces different responses and changes in plants. Previously, these groups were categorized into two main groups, namely growth and stress hormones. However, this categorization is invalid [27] as stress-related hormones are also involved in growth and development and vice versa. Hormones engage in complex mutual crosstalk [14, 16, 28,29,30] as well as crosstalk with reactive oxygen species [31] and other signaling compounds, resulting in moderately attenuated responses. Thus, methods capable of comprehensive plant hormone analysis could help unravel such complex interactions.

Metabolomics utilizes different workflow designs of untargeted (semiquantitative) analyses and targeted (quantitative) analyses such as metabolic profiling, fingerprinting or footprinting [32]. Plant hormonomics is essentially targeted analysis that requires preexisting knowledge about analytes of interest and their biological significance. The difficulty of targeted analysis rapidly increases with each plant hormone class monitored, together with all related precursors, intermediates and catabolites, reaching higher hundreds maybe thousands of compounds [20, 33]. Furthermore, the majority of these compounds are present only in trace amounts in vivo. Therefore, the method used must strike the right balance to enable a wide range of compounds to be detected but also high-throughput analysis of large numbers of samples [34]. Meeting such requirements may allow effective integration with other areas of plant sciences, e.g., high-throughput phenoty** (indoor and outdoor), a newly established and integral part of plant research [8, 35,36,37,38,39,40,41].

In this review, we summarize recent methodologies employed in multiple-class plant hormone analysis (with focus on target analysis of active hormones and few of their metabolites), explaining their possibilities and weaknesses. By revisiting the basic workflow of these methods, we discuss and suggest changes required for hormonomics and outline possible future trends in the field of plant hormone analysis. We aim to provide a nuanced perspective on the legitimacy of the term “plant hormonomics”.

Plant hormone significance and utility

The prominent role of plant hormones provides opportunities for a wide variety of uses and applications in research, agriculture and biotechnology. Identification of genes associated with biosynthesis, catabolism or perception and their modification can enable the development and improvement of crops with agronomically valuable traits (e.g., yield predictors, resistance to stresses and pathogens, morphology, chemical composition, nutritional composition, sensory qualities, technological properties) [42, 43]. Crop domestication has involved many changes at the genetic level directly linked to plant hormone action, such as reduced CK dehydrogenase activity, which increases the grain number in rice [44]. The use of semi-dwarf rice genotypes, caused by changes in the metabolism of GBs [45], is one of the key successes of the green revolution. Another agronomic achievement is the discovery of a gene providing resistance to long-term flooding, which is associated with the action of ET [46]. In the context of climate issues, stress-resilient crops that maintain a high yield during conditions unfavorable for cultivation have gained increasing attention [41, 47,48,49,50]. Several works have investigated modification of plant hormone-related genes to increase drought tolerance without penalizing growth [51] or changes to the plant architecture to improve water management strategy during drought [52]. Kudo et al. [53] focused on modifying genes associated with ABA and GBs metabolism to develop drought-tolerant plants. Such knowledge can be used to develop and breed superior crop varieties at a fast pace in conjunction with molecular biology and genetic tools [54,55,56,57].

In addition, plant growth regulators (i.e., plant hormones and their synthetic analogs) are used as chemicals in industry or directly in agriculture. For example, in plant biotechnology, they have been used in plant in vitro manipulation and propagation [58, 59] or directly applied to influence seed germination, regulation of growth, flower and fruit set, regulation of senescence, abscission and fruit ripening or achieve post-harvest manipulation [60, 61].

Analysis of plant hormones

Understanding the complex interaction of hormonal crosstalk requires extensive information about as many as possible plant hormones in a given sample. Crosstalk between hormones can take different forms, such as regulation of biosynthesis, inter-tissue transport, catabolism, signal perception and signal transduction of other hormones [62]. Information about only active forms does not fully represent ongoing processes. Thus, determination of levels of biosynthetic precursors, transport forms, storage forms and catabolites (also known as “hormone profiling”) is important [1, 2, 63] (Fig. 1) and might provide supporting information or clues that are not deducible from just examining levels of active hormones.

The notion of broad-scale plant hormone analysis is hampered by the problematic selection of which hormone metabolites should be included in the analysis as some prior knowledge is required. Good candidates are compounds unique to the pathway of interest that play a significant role or which are a part of a regulatory rate-limiting step. CKs profiling serves as a suitable example. Methods for CKs analysis commonly include CK nucleotides (biosynthetic precursors), nucleosides (transport forms), free bases (active hormones, transport forms) and glucosides (catabolites, possible storage forms [3, 64,65,66]. However, some pathways are still not fully elucidated as map** them requires decades of research and is often a complicated task. Pathways can be characterized by physicochemical studies monitoring the turnover and flux of isotope-labeled (stable or radioactive) compounds or molecular and genetic studies characterizing genes and enzymes involved [67,68,69,70,71]. Additionally, some pathways are known but the key compounds are not commercially available and require in-house synthesis or the precursors are involved in other pathways.

Such challenges can be exemplified by the SA, ET and Aux pathways. Major and minor SA biosynthetic pathways are known. The short major pathway includes three biosynthetic steps—the first committed step formation of isochorismic acid (IC) from chorismic acid, conjugation of IC with glutamate (IC-9-Glu) and subsequent hydrolysis to form SA. IC is not commercially available and IC-9-Glu is unstable [72]. Therefore, inspection of SA major pathway is difficult. Meanwhile, the minor pathway and SA catabolism have not been fully explored, hampering the selection of analytes and scale of analysis. In the case of ET and Aux, the biosynthetic pathways begin with precursors that are shared intermediates of other pathways. 1-aminocyclopropane-1-carboxylic acid (ACC), an ET precursor is generated by ACC synthase from S-adenosyl methionine (SAM). Consumed SAM is regenerated via the Yang cycle, which is also involved in polyamine and nicotianamine biosynthesis [73]. Similarly, tryptophan, a precursor of the major auxin, indole-3-acetic acid (IAA), is a common intermediate in pathways of specialized metabolites, protein biosynthesis and degradation [67, 74, 75]. Thus, changes in levels of biosynthetic precursors cannot easily be attributed to a single pathway, especially when levels differ by orders of magnitude.

Notably, evolution of the ET and JA biosynthetic and signaling pathways in the plant kingdom has led to the formation of intriguing networks. These pathways are characterized by the presence of multiple active compounds in biosynthetic pathways, e.g., 1-amino-cyclopropane-1-carboxylic acid (ACC) in the ET signaling pathway [76]. Biosynthesis of ACC has been confirmed in land plants, but only angiosperms and gymnosperms utilize ACC as a precursor for ethylene, whereas lower plants utilize different precursors. Further evidence that ACC is a standalone signaling compound has been gathered [76, 77]. Similarly, biosynthetic precursors of JA include 12-oxophytodienoic acid (OPDA) and dinor-12-oxophytodienoic acid (dnOPDA), which have regulatory functions distinct from JA itself [78]. These pathways present a challenge for isolating and understanding the specific roles of individual compounds and call for methods capable of comprehensive analysis.

Although plant hormone analysis targets a relatively small number of analytes, hormones and their metabolites (currently hundreds), compared to metabolomics (tens of thousands) [79,80,81], many of their problems are similar. The physicochemical properties (e.g., polarity, volatility, stability and solubility) of the nine plant hormone classes vary considerably, imposing different requirements for extraction and subsequent analysis. This issue is even more pronounced when considering hormones together with their metabolites.

Sample matrix

Currently, liquid chromatography–tandem mass spectrometry (LC–MS/MS) dominates as the method of choice for hormonal analysis [63, 82]. The main difficulty in the analysis of these compounds is their low endogenous concentrations (apart from a few individual compounds). To overcome this, analyte enrichment is necessary to ensure that the amount of analyte injected is sufficient to detect its signal. All analyses of complex samples (e.g., plant extracts) utilizing MS detection are inherently prone to matrix effects (ME), i.e., the signal of an analyte is influenced by coeluting compounds present in the sample matrix, which can either suppress or (less often) enhance the ionization process of the analyte, resulting in lower or higher signals respectively [83]. The design of the entire analytical procedure is heavily modified to prevent the suppression of ionization. The magnitude of ME is influenced by a number of factors. Different sample matrices, such as different plant organs and tissues (leaves, roots, flowers, fruits, tubers,…), have contrasting chemical compositions, resulting in different patterns of coeluting compounds. On the other hand, sufficient sample purification reduces the number of coeluting compounds. Finally, the ME is influenced by the chromatographic separation (a high number of coeluting substances lowers ionization efficiency) and analyte properties [84,85,86].

The problem of ME is especially pronounced in plant hormone analysis because enrichment of sample extracts without sufficient purification also leads to enrichment of the matrix components. There are two possible ways to mitigate or eliminate ME. One way is to employ intensive sample purification (“Sample purification” section) prior to analysis, whereas the other approach is to improve the chromatographic separation (“Chromatographic separation” section) [83, 87, 88], but both approaches require the utilization of internal standards to determine the magnitude of matrix effects ("Selection of internal standards" section). Some protocols for the analysis of nearly crude extract have been published [89,90,91,92]. However, the majority of published and validated protocols employ various procedures for sample purification to prevent ME during MS analysis.

Sample extraction

The nature of the extraction solvent and extraction conditions determines which analytes are efficiently extracted from the sample matrix. So far, no analytical method provides a universal solution and each of them introduces bias. The nine plant hormone classes form a set of compounds with widely different physicochemical properties, stability and volatility. Common practice is to use aqueous mixtures with a high content of organic solvents, e.g. methanol (MeOH), acetonitrile (ACN) and isopropanol (IPA). Typical solvents utilized for the extraction of plant hormones are listed in Table 1. Organic solvents are capable of extracting a wide spectrum of small molecules [93] and precipitate proteins (ACN is more effective than MeOH) [94, 95]. In addition, in correct concentrations, they can substantially decrease enzymatic activity to prevent enzymatic changes during extraction—e.g. a strong inhibitory effect has been found when using 40–50% aqueous ACN, whereas at higher concentrations of ACN enzyme activity increases [96, 97]. Appropriate sampling and metabolism quenching prevent undesirable chemical changes in sample composition. Additionally, mechanical intervention [98] and changes linked to circadian rhythms [66, 99] can alter plant hormone levels. Therefore, thorough planning and careful sampling should precede any analysis of plant material [100].

Table 1 Overview of recently developed methods for plant hormone analysis (continued on the next page)
Full size table

The most frequently used solvents include modified Bieleski solvent (15:4:1; MeOH:H2O:formic acid) and aqueos mixtures of MeOH or ACN (Table 1). Originally, Bieleski solvent was developed to prevent phosphatase activity during the sample extraction of CKs [101]. Modified Bieleski solvent avoids the use of chloroform in the extraction mixture and has been shown to provide the same or higher extraction efficiency as the original solvent [102]. In addition to CKs, this solvent has been used for the extraction of other plant hormone classes, e.g. acidic hormones ABAs, JAs, SAs, GBs, Aux [103,104,105,106,107] and ACC [108]. MeOH and ACN can be used for extraction under milder conditions including analytes with a wide range of polarities such as CKs, GBs and BRs [4, 149] On the other hand, the BEH C18 column provides better mechanical and chemical stability.

In contrast, alternative non-RP chromatographic systems are scarce in the field, with only two hydrophilic interaction chromatography (HILIC) based methods for CK analysis published to date [136, 150]. Despite this, HILIC offers several advantages: it provides an orthogonal separation mechanism to RP, allows separation of mid-polar and polar compounds (such as ACC) and typically provides lower detection limits [151], a crucial parameter for hormonal analysis. Analyte retention is achieved via analyte partitioning between the bulk mobile phase and a stagnant water-rich layer formed on the stationary phase, hydrogen bonding and electrostatic interactions. Available stationary phases include bare silica and bonded phases, such as amino, cyano, (cross-linked) diol, amide, zwitterionic, polyethylene glycol and other variants [152, 153] each providing unique selectivity based on the dominant retention mechanism of the particular phase. A better understanding of the fundamental principles of HILIC and a decade of technological progress may facilitate its use as a valuable complementary system to RP. Other chromatographic systems are also improving, such as ultra-high performance supercritical fluid chromatography, which combines advantages of both normal phase and RP chromatography [154].

As mentioned earlier, there are essentially two ways to mitigate ME: (I) sample purification, and (II) improved chromatographic separation [83, 87]. The already established complex SPE protocols provide one or more fractions of a single sample [103, 107]. Each fraction, of relatively high purity, undergoes a very short analytical run. This has allowed separate analyses of 4, [162,163,164]. Great benefits can be obtained from the use of new chemical modifications of stationary phases, diverse functional groups and multi-modal phases, and new column designs. Packed bed columns remain the gold standard, outlasting even their supposed successors, monolithic sorbents. It is not yet clear whether new additive methods (3D printing) for column production [165, 166] or application of multidimensional LC [167] will contribute to a radical shift in separation technology. It will be interesting to see whether these technologies find wider application and dominance for the analysis of natural substances and plant hormones.

Mass spectrometry detection

The first step performed in MS detection is ionization, which takes place in the ion source. In the context of small molecule analysis, different types of ionization can be used, e.g., electrospray ionization (ESI), atmospheric pressure chemical ionization (APCI) and atmospheric photoionization (APPI). The efficiency of ionization is profoundly affected by coeluting compounds, contributing to ME. Among the different types of ionization, ESI is the most widely used but also the most prone to ME. The mechanism of ME in ESI is not entirely clear but likely involves competition among analytes co-occurring in the ion source for a charge, which is limited because the ESI source is in a sense an electrolytic cell operating at the same current [86, 168].

Other ionization techniques APCI and APPI are important complements that are less prone to ME. The main reason is the different mechanisms of ionization. At the APCI interface, sample vapor is ionized in the reaction zone of the corona discharge needle, whereas at the APPI interface, ionization occurs when an 8–12 eV photon is absorbed by the molecule and an electron is ejected [169,170,171]. The different ionization mechanisms also result in different selectivity. APCI allows more efficient ionization of nonpolar compounds that are difficult to ionize using ESI, and for example, has been used for the analysis of oxylipins [172] BRs [173] and several acidic hormones after derivatization [174]. On the other hand, APPI favors ionization of larger molecules, reduces ionization of molecules of solvent, which generally have a high ionization potential, and overall generates cleaner spectra [169]. Despite the higher susceptibility to ME, ESI remains the most popular choice because it allows ionization of a wider spectrum of analytes.

Tandem MS (MS/MS) detection is frequently used as a routine method for plant hormone analysis as emphasis is placed on ensuring a sufficient selectivity as well as high measurement sensitivity. Triple quadrupole (QqQ) mass analyzers are favored for several reasons. An inherent feature of quadrupole mass analyzers is the quantitative data output. In addition, the multiple reaction monitoring (MRM) mode provides the necessary selectivity and electron multiplier detectors ensure a high sensitivity [175, 176].

However, QqQ mass analyzers have some limitations, e.g., the acquisition speed of these instruments allows only a certain number of analytes (MRM transitions) to be monitored during the time-frame of a chromatographic run. Multiple transitions are monitored to provide additional confirmation of the identity of analytes. This is based on relative ratios of multiple signal intensities that are unique for each compound [177, 178]. Monitoring one MRM transition is sufficient for quantification. However, it does not provide sufficient selectivity and increases the risk of false identification and quantification of the targeted analyte. While it is possible to decrease the number of monitored MRM transitions per analyte, this may be problematic when studying complex or previously uncharacterized matrices. On the other hand, monitoring multiple MRM transitions is not always feasible since small molecules generate simple fragmentation spectra after collision-induced dissociation. A high signal intensity is measurable for a single fragment, whereas additional transitions provide order(s)-of-magnitude lower signal intensities, often rendering them irrelevant for trace analysis, e.g., ACC ionization in positive mode with m/z 102 → 56, ionization in negative mode with SA m/z 137 → 93 and JA m/z 209 → 59 [108, 179].

Rapidly develo** high-resolution mass spectrometry (HRMS) has opened new possibilities for plant hormone research by facilitating analysis of high molecular weight targets of plant hormones, e.g., proteins of signaling pathways, leading to better characterization of protein interaction or post-translational modifications. This, in turn, improves the description of plant hormone signal cross-talk [180]. HRMS (based on quadrupole time-of-flight and QExactive Orbitrap analyzers) offers numerous advantages over MS/MS based on QqQ [181], such as better selectivity, a substantially higher number of acquired analytes/features, no requirement for standard compounds. Additionally, analyses of small molecules can take advantage of the comparable or higher sensitivity of Orbitrap mass analyzers [182,183,184,185,186]. Nevertheless, the adoption of HRMS in quantitative plant hormone analysis remains somewhat reserved. Several studies have employed HRMS for trace analysis of Aux, CKs or multiple-class analyses [71, 92, 104, 120, 187, 1].

Derivatization

Not all plant hormones and related compounds have properties suited for MS detection. A poor ionization efficiency, thermal instability, low molecular weight, or their combination, can result in a low signal response and subsequent failure of detection or proper quantification of plant hormones in samples. In such cases, chemical derivatization could be used during sample preparation to introduce moieties that enhance ionization, increase the stability or molecular weight of the analytes, or improve the retention in a given chromatographic system [71, 82, 160, 203].

Derivatization is a necessary step when using gas chromatography (GC) as it requires volatile analytes. However, in plant hormones analyses GC–MS has mostly been displaced in favor of LC–MS (apart from analyses of volatile hormones) [82, 114] and derivatization in LC–MS remains an option rather than a necessity. Several plant hormone classes exhibit low sensitivity due to a poor ionization efficiency. In general, this applies to acidic plant hormones ionized in the negative mode at the low pH values of mobile phases used in RP systems or other classes, such as Aux, SA and BRs. Thus, the use of a derivatization step in sample preparation may be inevitable. However, in the context of plant hormonomics, the justification for a derivatization step is debatable for several reasons. Firstly, the extreme conditions of derivatization may lead to substantial changes in the profile of the compounds of interest. Secondly, it is already clear that it is not sufficient to just use one derivatization reagent [132]. The use of multiple reagents may be undesirable, require a more demanding protocol or generate hidden changes or artifacts.

Reagents typically employed for the analysis of plant hormones are listed in Table 2. Due to their extremely low endogenous concentrations and poor ionizability, BRs are often derivatized using analogs of phenylboronic acid targeting the vicinal diol moiety. The advantages of these reagents are selectivity and mild reaction conditions. Various reagents have been utilized, e.g., such as 2-methyl-4-phenylaminomethyl-benzeneboronic acid [132], 4-phenylaminomethyl-benzeneboronic acid [144], 3-(dimethylamino)-phenylboronic acid [224, 225]. Protocols for the separation of cell organelles have also been developed. Cells can be separated into plastids, cytoplasm, vacuoles and endoplasmic reticulum. Subsequent analysis then provides metabolite profiles in each organelle [226,227,228,229].

These unique protocols open further possibilities for destructive type analyses and offer an alternative to biosensor [2, 230], microscopic and molecular biology methods for the study of processes occurring at the cellular level. However, such MS techniques have significant drawbacks. Further applications of single-cell analysis and MS imaging are severely hindered by the trace levels of analytes present in samples. Flow cytometry cell type sorting requires upfront GFP labeling unique for each cell type and protoplast preparation. Also, maintaining hormonal homeostasis and minimizing changes during such sample preparation may be problematic. The main difficulty in the case of plant hormone analysis is the need to obtain enough material using these techniques for successful detection.

Data analysis

Data analysis is an essential aspect of experiments involving instrumental analyses, capable of extracting valuable insights from the chemical information obtained. Preprocessing workflows differ for targeted and untargeted LC–MS analysis. Essentially, targeted analysis does not require extensive data preprocessing, unlike untargeted LC–MS studies. As the identities of compounds are known, the main aim is their accurate quantification. Therefore, little to no emphasis is placed on the identification of unknown compounds and data are normalized using IS (“Selection of internal standards” section). Nonetheless, even targeted analysis may require data transformation. Plant hormones are often present at extremely low levels and provide signals near the limit of quantification or detection (recommendations for validation of chromatographic methods please see ref [231, 232]). However, endogenous concentrations may vary by several orders of magnitude after stimulation. For multi-order calibration and curve fitting, log–log transformation [log(normalized signal) plotted on the y-axis vs. log(concentration/molar amount) plotted on the x-axis] is preferable to linear regression without any transformation or weighting; log–log transformation improves curve fitting, increases robustness and avoids massive leveraging effects by evenly spacing calibration points (1,3,10,30… or √1, √10, √100, √1000… concentration series) [233,234,235]. Therefore, data transformation should be considered.

The experimental design, whether descriptive or hypothesis-driven, and research objectives should guide the selection of statistical analyses. A crucial consideration is the small sample size in plant hormone analysis, which is challenging to overcome. Limitations stem from the preparation of the plant material and the laboriousness of the analytical procedure. A large sample size is difficult to achieve when working with rare mutants, small plant organs, such as root tips, apical meristems and similar organs, or when monitoring dynamic changes with multiple time-point collections. Despite these problems, some biological questions may only require simple statistical comparison, e.g., differences in levels of relevant chemical species (hormones). For hypothesis testing, tools such as the t-test for two group comparison and ANOVA (including post hoc tests, e.g., Tukey’s range test) for three or more groups are essential. When a normal distribution of data cannot be assumed, the Mann–Whitney test U test and Kruskal–Wallis are non-parametric alternatives to the t-test and ANOVA. However, normality tests (e.g., Shapiro–Wilk W test) have low power when performed on a small sample size (n < 30) and might lead to erroneous assumptions [236, 237]. To conduct statistical analysis, outlier rejection might be necessary. Dixon and Grubb’s tests are commonly used when dealing with small sample sizes or a robust median absolute deviation-based method could be considered to detect outliers before performing statistical analysis [238]. These listed approaches represent just a small fraction of all tools available for outlier detection [239]. Given the high biological variability of metabolite levels and small sample size, outlier rejection has to be performed cautiously.

Multivariate analysis plays a pivotal role when looking for insights within complex datasets. Instrumental analyses allow the collection of large amounts of data, especially untargeted analyses. This also applies to large-scale targeted analyses. Thus, multivariate analysis may provide new insights into the data obtained. In this review, only principal component analysis (PCA) and clustering analysis will be briefly mentioned. As the volume of data increases, so does the difficulty of recognizing patterns. Large datasets often include variables (e.g., concentrations of metabolites/hormones) that are correlated, making some information redundant. PCA is frequently used to visualize 2D (or 3D) plots of datasets with large sets of variables for visual inspection and pattern recognition [238]. This is performed by finding the principal components to reduce redundant information and retain the variance of uncorrelated variables. The first principal component (PC1), a combination of original variables, is oriented in the direction of maximum variation. The second principal component (PC2) is oriented in the direction of the next greatest variation, while remaining uncorrelated to PC1. In order to capture the majority of variance by PC1 and PC2, a dataset has to contain variables with high covariance, otherwise PCA is not suitable. Cluster analysis again helps with identifying patterns and groups within a dataset. Hierarchical cluster analysis (HCA) is a simple and useful non-supervised clustering method. The iterative process of HCA sorts and links objects (samples) by their similarity, and as a result, a dendrogram is plotted that is straightforward and easy to understand. However, simplicity and elegance are not always possible when handling very complex datasets (either many samples or many variables) as the dendrogram becomes crowded and difficult to interpret. PCA and HCA just scratch the surface of multivariate statistical analysis. The topic has been the subject of several reviews; for further reading, see [238, 240,241,242].

Summary

A new popular term has been coined—plant hormonomics. Similarly to other omics sciences, plant hormonomics aims to achieve qualitative and quantitative analysis of the hormone complement in a given sample. Hormones represent a standalone set (several hundred) of low molecular weight compounds alongside primary and secondary metabolites. They possess high biological activity and govern all processes during a plant’s lifecycle, attracting the attention of many researchers. Not all related compounds exhibit high biological activity, but data on the biosynthetic precursors, transport forms, storage forms and catabolites of biologically active hormones provide additional and valuable information that can help to elucidate the internal processes.

In general, plant hormone analysis shares some common ground with metabolomics. The key difference is the several orders of magnitude lower endogenous levels of plant hormones in vivo, in contrast to other metabolites. To counteract matrix effects hindering correct signal responses, samples for hormonal analysis are typically purified to remove matrix components and enriched to obtain the highest possible signal response. However, comprehensive hormonomics analysis requires little or no purification, otherwise some analytes may be removed during sample purification. Therefore, efficient and robust chromatographic separation is necessary to minimize matrix effects. Notably, current methods do not fully explore the possibilities available in the field of liquid chromatography, such as other chromatographic systems (hydrophilic interaction chromatography, supercritical fluid chromatography, multimodal), and generally incline toward extremely short analysis of highly purified samples. Furthermore, technological improvements in HRMS have led to the development of instruments that offer equal or greater sensitivity compared to triple quadrupoles. At the same time, they offer the advantages associated with high resolution. However, the advent of these technologies has been so far reserved. Sample derivatization could help when dealing with small sample sizes or low signal responses in general. The improved physicochemical properties of derivatized analytes are often unparalleled when compared to the native state. However, despite the increased hormonome coverage, complications may arise, such as artifact formation and degradation of analytes under harsh derivatization conditions. Therefore, derivatization is mainly reserved for non-hormonomics applications.

Destructive plant hormone analysis remains an indispensable tool in plant sciences. However, conventional single hormone class analysis employs a purification protocol that eliminates the majority of other matrix components, including other hormone classes. Comprehensive multiple/all-class plant hormonomics analysis cannot be achieved using a minimal purification protocol, fast chromatographic separation and triple quadrupole tandem mass spectrometry.