Mutation induction through chemical or physical mutagenesis has long been used in plant breeding as it creates new alleles and can lead to the development of new, agronomically important traits (for further reading see Maluszynski et al., 2017). It is also used in genetic studies for functional analysis of mutated genes. There are two main approaches to such studies: “forward” and “reverse”. The classical forward approach starts with the identification of an interesting phenotype and then the mutated gene responsible for this phenotype can be searched for, usually through mapping and positional cloning. The “reverse” approach is the other way around – first, the mutation within the gene of interest is identified and afterward the phenotype caused by this mutation is analyzed (Griffiths et al., 2004).

TILLING (Targeting Induced Local Lesion IN Genomes) was first developed for Arabidopsis thaliana as an alternative to insertional mutagenesis. It was described as a reverse genetic strategy that combines traditional random chemical mutagenesis with rapid mutational screening for induced lesions in genes of interest (McCallum et al., 2000a; McCallum et al., 2000b). However, once established, a TILLING population may be used for forward genetic studies as well. Immediately after its invention, TILLING has been adapted and deployed for large-scale screening of induced mutations not only in Arabidopsis but also in other plants (reviewed in Colbert et al., 2001; Henikoff et al., 2004; Stemple, 2004). At the beginning of TILLING, one of the biggest limitations of this technology was the necessity of knowing the sequence of the gene of interest (or at least its fragment) and TILLING relied on the emergence of practical genome sequence tools. However, as more and more plant species have been fully sequenced in recent years and there is an enormous amount of sequence data and bioinformatic tools for their analysis, these issues are becoming less limiting. Additionally, throughout the years TILLING is being continuously modified to make it more robust, especially in terms of mutation identification (discussed in the next section). The general scheme of TILLING strategy is described in detail in Szarejko et al, 2017 and Szurman-Zubrzycka et al., 2017, and illustrated briefly in Figure 1 .

The goal of mutagenesis for TILLING is to obtain a highly mutagenized population to increase the probability of finding a mutation in every gene. Thus, it is crucial to select a proper mutagen and its dose. The sensitivity for mutagen treatment varies between plant species and even between various genotypes of the same species. In general, the lower the mutation frequency, the larger population is needed, which leads to higher costs and labor efforts. On the other hand, too high mutation frequency may cause lethality or sterility of M₁ plants. The most popular mutagen used for the creation of TILLING populations is EMS (Ethyl methanesulphonate) which is an alkylating agent causing mainly G/C to A/T transitions (Kurowska et al., 2011). The mutation density in different TILLING populations is very diverse and ranges from ~1/7 Mb to 1/20 kb ( Table 1 ; Till et al., 2018). Once the dose of a mutagen is optimized, the TILLING platform may be created. For generatively propagated crops, the material to be mutagenized is usually the seeds (Jankowicz-Cieslak et al., 2011). This is the best choice because seed mutagenesis is technically easy, there are no temporal or developmental stage restrictions for treatment, relatively high mutagen doses can be applied to dormant seeds, and they are easy to store and transport. Only mutations induced in generative line cells in the seed embryo are inherited, hence all molecular analyses aimed at the identification of induced mutations should be performed starting with the M₂ generation. In regard to vegetatively propagated crops, such as banana (Musa paradisica L.), peppermint (Mentha L.) or sugarcane (Saccharum officinarum), the generation of stable mutants is more problematic. The meristematic buds (e.g. tubers, bulbs, rhizomes) or cell or tissue cultures in vitro are usually target materials for mutagenesis (Jankowicz-Cieslak et al., 2012; Jankowicz-Cieslak and Till, 2016). The chimera dissociation usually has to be performed e.g. by adventitious regeneration from a single somatic cell or by somatic embryogenesis (Geier, 2012). For generatively propagated plants, only one seed from the M₁ plant should be propagated into the next generation (M₂) in order to avoid the repetitions of the same mutations during screening. A very important aspect is the maintenance of mutagenized generations – each M₂ plant and its M₃ progeny should be assigned a unique code and seeds from each plant should be collected separately.

The mutation screening is performed on the M₂ generation, because of the chimerism of the M₁ generation owing to the fact that seeds are multi-cellular and only a subset of the mutagenized cells are responsible for germline development. The DNA should be isolated from each M₂ plant individually (one M₂ plant per M₁ parent) and is usually pooled prior to screening for mutations. There are many screening methods aimed at mutation identification - they differ in sensitivity, capacity, difficulty, and costs. These methods are briefly described in the following section. Depending on the methods of mutation screening used, there is a different schema of DNA sample pooling. The M₂ plants (and further generation) may be also directly phenotypically characterized for a forward genetic approach. Importantly, the proper maintenance of seeds from all generations (in a seed bank) and DNA stocks isolated individually from M₂ plants are crucial to make the TILLING platform long-lasting. In order to make a TILLING platform publicly available, it is indispensable to create a database where all data about phenotypes, mutations identified and seed availability is collected.

The first TILLING population was developed for Arabidopsis thaliana in 2000. Because TILLING is a method that can be easily applied to most organisms, shortly afterwards new TILLING populations for other plant species were produced. Examples of TILLING platforms created for various, diploid and polyploid, plant species are presented in Table 1 .

The discovery that mutations can be induced revolutionized plant breeding. The first crop mutant variety (‘Chlorina’ – a variety of tobacco) was introduced in the 1930s (Tollenaar, 1934; Tollenaar, 1938) and methods of mutation induction with the use of physical or chemical mutagens have long been established for many plant species. In the TILLING approach, mutation induction is followed by mutation identification in the genes of interest. The detection of mutations at the DNA base level became possible with the development of sequencing methods beginning in the 1970s. In recent years DNA sequencing methodologies have experienced a renaissance, with methods continually evolving. This is resulting in higher throughput and lower costs for mutation discovery and other applications.

In the first TILLING experiments performed on Arabidopsis thaliana the method of point mutation identification was based on DHPLC (Denaturing High-Performance Liquid Chromatography) (McCallum et al., 2000a; McCallum et al., 2000b). Arabidopsis seeds were treated with EMS, DNA was isolated from M₂ individuals and pooled, PCR reactions for selected gene regions were performed and after denaturation and annealing (allowing heteroduplex formation in the case of mutation appearance) the DHPLC method was used. This method allows the detection of heteroduplex in a pool as an additional peak in the chromatogram, because of differential melting kinetics of homo- and heteroduplexes of DNA. After the identification of the potential mutations in particular M₂ plants, the mutant PCR products were sequenced to confirm the mutation. The DHPCL method is very sensitive, however, it is relatively low throughput. Another method used for direct analysis of heteroduplex appearance is HRM (High-Resolution Melting) which is very fast but restricted only to fragments of max. 400 bp length (Gundry et al., 2003; Szurman-Zubrzycka et al., 2017). In order to increase the efficiency of mutation discovery, single-strand-specific nucleases began to be used (Colbert et al., 2001). The most commonly used endonuclease is CEL I which recognizes heteroduplexes and cuts them in a position of mismatch (position of induced mutation). The observation of cleaved products was typically performed with the use of LI-COR sequencers, where up to 768 samples may be analyzed on one gel, in the case of 8-fold DNA pooling (Colbert et al., 2001). For the first two decades of TILLING research, this was the most common method of mutation discovery. To reduce the cost of TILLING screening it is possible to visualize the products of digestion on an agarose gel, but this method is not very sensitive and some mutations may go unnoticed. Recently, capillary sequencers (e.g. AdvanceCE 96 FS or Fragment Analyzer) are also becoming the instruments of choice for the identification of heteroduplex cleavage products (Jost et al., 2019).

Currently, the rapid identification of all mutations present in the mutant genome may be done through the NGS approach (‘TILLING by Sequencing’), either directly or using different strategies of sample pooling (one-, two- or three-dimensional) (e.g. Tsai et al., 2011; Fanelli et al., 2021). The huge restriction for broad usage of NGS for TILLING is its cost, however, NGS is becoming noticeably cheaper over time. Another problem is the production of very large datasets for bioinformatic analysis. Nevertheless, the collection of genomic data from individual plants from mutagenized populations opens up new possibilities for TILLING approach. One can simply screen collected data for plants carrying mutations in a gene (regions) of interest and order the seeds from a developer of TILLING population (Krasileva et al, 2017). To reduce the size of obtained data and to minimalize the acquisition of unnecessary data, ‘TILLING by Sequencing’ may be restricted only to coding fragments of DNA, when performing exome capture instead of whole genome sequencing. NGS methods may be also applied to screen a set of pooled PCR amplicons instead of sequencing of whole genome/exome. There are many different NGS technologies described up till now, with Illumina being the most popular in TILLING projects. However, we can suppose that in the future the long-read sequencing technologies, such as PacBio and Oxford Nanopore sequencing, or new technology like Linked-Read (10xGenomics), will also be used in TILLING strategy. This can contribute to increasing sequencing accuracy or improving the balance between costs and depth of sample sequencing. The methods of mutation identification used in different TILLING projects are shown in Table 1 . Recently, Wang et al. (2023) showed a very exciting example of how to boost wheat functional genomics using a mutant population of common winter wheat variety KN9204 generated by EMS treatment, similar to the majority of TILLING population, however, authors did not describe their strategy as TILLING. Studies aimed to obtain gene-indexed mutants in every coding gene in the wheat genome, which is a huge effort from one side, but offers great opportunities from the other. The use of NGS technology, specifically exome capture sequencing of 2,090 mutant lines revealed that for almost all coding genes (99%) this goal was achieved. Potentially, it might be also the direction of progress in the analysis of TILLING population in other plant species.

Another new approach used to screen mutagenized populations is called FIND-IT (Fast Identification of Nucleotide variants by droplet DigITal PCR) (Knudsen et al., 2022). It provides ultrafast screening (10 days) for targeted genetic variants of interest (at single-nucleotide resolution). FIND-IT combines large-scale sample pooling with highly sensitive droplet digital PCR (ddPCR) for genotyping. It was efficiently validated for barley by fast screening of variant libraries from 500,000 individuals and isolating more than 125 targeted gene knockouts and other interesting variants (Knudsen et al., 2022).

Even though the first TILLING service was developed more than 20 years ago, this strategy is still broadly used in plant research studies, especially for agronomically important species. TILLING mutants are used to study the genetic bases of many important traits related to plant development and response to various biotic and abiotic stresses. TILLING is also used to create genotypes with improved important agronomic traits e.g. related to climate change, and hence it holds great promise for addressing global challenges in agriculture and food security and may be helpful in achieving some of the Sustainable Development Goals (SDGs) adopted by the United Nations General Assembly in 2015 as part of the 2030 Agenda for Sustainable Development (https://sdgs.un.org). The newest achievements generated through the TILLING approach are illustrated in Table 2 , where we describe the goal of using TILLING in a particular study, a species and genotype used for TILLING screening, a gene and encoded protein analyzed, its biological function, the created mutants and their characteristics.

The biggest advantage of TILLING, contrary to transgenic techniques, is that it is applicable to all species regardless of their genome size or transformation potential. There is no need to use sophisticated tissue cultures that for some species or genotypes are impossible to maintain. After traditional mutagenesis used for the creation of TILLING population, mutations can be found in every gene of interest. In general, the higher density of mutations the smaller population is needed to be analyzed to find mutations within each gene. With the use of TILLING one can identify the series of alleles (potentially giving “weak” and “strong” phenotypes), not only knock-outs. It is of special importance, especially for genes that are crucial for plants to survive, whose total inactivation may be lethal. For them, the sublethal alleles may be required for phenotypic analysis. Advantage from the reduction of protein activity obtained with TILLING-mutants vs. total turn off of the gene function using RNAi technique can rely also on avoiding some disruption during plant growth and development, which was indicated e.g. in research in tobacco, where RNAi mutants were characterized by necrotic lesions or increased water content in leaves (Liedschulte et al., 2017). Importantly, the analysis of multiple “weak” and “strong” alleles allows the deepening of knowledge on gene function. In barley analysis of the HvHox1 (homeodomain-leucine-zipper1) gene, which controls the row-type character of a spike, showed that three forms carrying missense mutation presented different phenotypes compared with the parent cultivar ‘Barke’. It reflects the different influences of substitutions on the protein function. Two mutants presented intermediate phenotypes, another one a six-row type spike, while cultivar ‘Barke’ was two-rowed (Gottwald et al., 2009). Another example of an increase in knowledge about gene function comes from Arabidopsis and mutant generated by TILLING in ABP (Auxin Binding Protein 1) gene - abp1-5 (Henikoff et al., 2004). This mutant carries a missense mutation leading to substitution in the auxin binding pocket of ABP1, which might alter the auxin binding process. Studies using the TILLING mutant led to the revealing of many auxin-related roles for ABP1, while null alleles, appeared to be embryo-lethal (Enders et al., 2015). TILLING could be also the technique of choice when complete loss of function of the investigated gene is undesired in breeding programs. An example of the necessity to obtain so-called “weak” alleles (that can be provided by TILLING) was shown e.g. in the study of Liedschulte et al. (2017) on the reduction of cadmium in leaves of tobacco (Nicotiana tabacum). The low level of Cd might be „beneficial” for smokers because it leads to lower Cd accumulation in their bodies. The study conducted on inhabitants of the Upper Silesia region in Poland indicated that smokers accumulated a twice higher level of Cd compared with non-smokers (Bem et al., 1993). First, two heavy metal transporter homeologous genes - HMA4.1 and HMA4.2 were silenced using an RNAi (RNA interference) approach. This was successful for Cd reduction in tobacco leaves, however negative effects on plant development, including retarded growth, necrotic lesions, altered leaf morphology, and increased water content have been observed. Application of the TILLING strategy led to removing this impact, and mutants with both, a lower content of Cd in leaves and no negative developmental effects have been identified. In order to minimize the phenotypic effects, the different mutation combinations in analyzed genes have been investigated. It has been proven, that complete functional loss obtained by induction of nonsense mutation in one homeologous HMA4 gene and the functional reduction in the other HMA4 gene obtained by induction of missense mutation led to the best results and these mutation combinations are the best choice for breeding programs in tobacco (Liedschulte et al., 2017).

The TILLING strategy has been used for more than 20 years, during which it was modified and improved. In general, as described in previous sections, the screening methods keep getting better, faster, and cheaper, which makes TILLING a robust tool for functional genomics. This can be seen especially in terms of sequencing technology, but currently, the use of TILLING-by-sequencing is limited to species for which reference genomes are available. The software for mutation analysis is also evolving together with TILLING. As an example, the PARSESNP tool (Taylor and Greene, 2003) was very useful at the beginning of TILLING research, however, it is not active anymore, whereas new tools are arising e.g. Mutation Finder Annotator (https://github.com/bjtill/Mutation-Finder-Annotator-CLI) that can be used for TILLING by sequencing. It is noteworthy that TILLING can also be conducted without any sophisticated apparatus, utilizing only basic laboratory equipment (e.g. with product visualization on agarose gel). Such a low-cost approach is suitable for implementation in developing countries with limited laboratory infrastructure, however, it is important to acknowledge that it may result in reduced sensitivity when it comes to identifying mutations.

A significant advantage of TILLING is that once established, a TILLING platform is long-lasting and may be used for both, forward and reverse screening. Finally, most important from an agronomic point of view, non-GMO mutants are created, as in TILLING strategy there is no need for transformation. Hence, new TILLING alleles may be used not only for functional analysis of genes but also serve as a valuable resource for crop improvement because they can be directly used in breeding programs without regulatory restrictions that exist in some countries for the material labeled as GMO.

TILLING, like any other approach, has also some limitations. Establishing TILLING platform may be time- and labor-consuming, but once established it may serve as a source of mutation for many years. However, the biggest disadvantage of TILLING is, without a doubt, the presence of background mutations that can affect the phenotype and, hence, impede gene function analysis. It was shown that in an abp1-5 mutant carrying change in the ABP (Auxin Binding Protein 1) gene of Arabidopsis, which was identified by TILLING approach, after whole-genome sequencing the additional 8,000 single nucleotide polymorphisms were identified (Enders et al., 2015). It was also calculated that for barley, in a TILLING population with mutational density of 1/500 kb, there are ca. 10,000 mutations in the genome of each plant (Szurman-Zubrzycka et al., 2018). However, it has to be noted that the coding part of a genome is usually very small (in the case of barley, it is 1.4%; Mascher et al., 2017), so most of these 10,000 mutations are in the noncoding sequences, with the lower probability of impacting the phenotype. What is more, some of those in coding regions may be silent or cause amino-acid substitutions that do not affect the protein activity. Taking it all into account, the total number of background mutations that was problematic in terms of functional analysis is low enough to be removed from the background by backcrosses with the parent variety. One backcross of a TILLING mutant with a parent variety reduces the number of background mutations by half, and performing a higher number of backcrosses can be recommended. However, it should be kept in mind that multiple backcrossing will prolong the time before phenotyping. The F₁ generation is heterozygous in terms of analyzed mutation. The homozygous mutants should be selected from the BCF₂ generation. Mutations are very often recessive, which means that only individuals carrying a mutation in a homozygous state should be phenotyped (Okabe et al., 2011). Such backcrossing is often time-consuming and the high density of background mutations is considered to be the main problem in the application of TILLING strategy as a reverse genetic tool. This problem is avoided in the modern targeted genome editing methods, such as CRISPR/Cas9 (reviewed in Gaj et al, 2013; Mei et al., 2016; Liu et al., 2021) and TALEN (reviewed in Becker and Boch, 2021), which are two powerful genome editing (GE) tools that have revolutionized plant biotechnology. In the case of these GE methods, called also New Genomic Techniques (NGTs), the mutation can be introduced specifically to the target position within the gene of interest, and the probability of off-target mutation appearance is relatively low. It is the main advantage of CRISPR/Cas9 and TALEN over TILLING. Both, CRISPR/Cas9 and TALENs have been successfully employed in plant research, and being the precise and efficient methods for modifying plant genomes, they have opened up new possibilities for plant breeding and biotechnology. It has to be kept in mind, that in some countries materials produced through these technologies are labeled as GMOs, even though the stable CRISPR/Cas9 or TALEN mutants do not carry any transgene and do not possess foreign DNA. However, the regulatory landscape in this regard is continuously evolving, and the legal status of CRISPR/Cas9 mutants is undergoing changes, reflecting the shifting perspectives and emerging legislative frameworks in various jurisdictions (e.g. the newest, dated July 5th 2023, proposal, for a REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL on plants obtained by certain new genomic techniques and their food and feed, and amending Regulation (EU) 2017/625).

In general, TILLING and GE approaches should complement each other in functional genetic studies or in pre-breeding programs. There are several examples of joined usage of these approaches to study gene function. For example, Wang and co-workers by using CRISPR/Cas9 and TILLING strategies demonstrated that TaGW2 homoeologs are negative regulators of grain size and weight, that contribute additively to the phenotype. The obtained TILLING and CRISPR/Cas9 mutants provide the opportunity to combine mutant alleles in various configurations, enabling fine-tuning of the phenotypic outcomes (Wang et al., 2018). However, it is also needed to be highlighted that for many plant species, including those agronomically important, there is no efficient protocol for transformation. Hence, GE technologies may be applied only to a limited number of species or to a limited number of cultivars of particular species (e.g. a protocol of barley transformation is efficient only for one cultivar - ‘Golden Promise’), whereas TILLING may be applied to any plant species and genotype.

The problem of background mutation was also reduced in a FIND-IT approach, because of using a mutagenized population with very low mutation density (Knudsen et al., 2022). This is a powerful barley platform generated with the use of NaN₃, possessing high capacity for the detection of premature stop codons, which was demonstrated by the identification of one hundred such changes in genes of interest. Mutations that cause the emergence of a premature stop codon have a high probability of disrupting the function of the encoded protein. However, these types of mutations are detected with low frequency as compared with other types of mutations (silent, missense) in previously reported plant TILLING populations (Kurowska et al., 2011; Szurman-Zubrzycka et al., 2018, Gao et al., 2020; Brisou et al., 2021).

We are currently in the so-called postgenomic era. Due to the high throughput methods of genome sequencing, there is more information about sequences than about gene function. Functional analysis of genes lags behind the acquisition of new sequences of whole genomes. It is well established that induced mutations have been a powerful tool in functional genomics and breeding for over 80 years of their use in plant research. Establishing a TILLING platform may be time-consuming, however for many species, these platforms have already been developed and the identification of TILLING mutants is really fast. Additionally, one can predict that as technologies of mutation detection are getting better and faster, so TILLING will too. As an example, FIND-IT, the new TILLING-based approach, provides ultrafast screening for variants of interest (within 10 days). The bottleneck is usually the characterization of identified mutants – determining the phenotypic consequence of the mutation, but this is a problem of each mutation technology, including gene editing by CRISPR/Cas9. In terms of functional genetics studies, it is always an added value to combine different approaches that complement each other, e. g. TILLING and CRISPR/Cas9 strategies that produce different types of mutants. It is also worth reminding that TILLING mutants are not GMOs so they can be directly used, without regulations, in pre-breeding programs around the world to improve crop performance.

MS-Z and IS conceive the manuscript. MS-Z and MK wrote the manuscript; MK generated tables; BT and IS revised and edited manuscript. All authors approved the submitted version.