The RIPper, a web-based tool for genome-wide quantification of Repeat-Induced Point (RIP) mutations

The RIPper

The RIPper includes an open source set of tools that allows for genome-wide quantification of RIP mutations in the genomes of Ascomycota. The RIPper accepts whole genome sequence data in FASTA format. This program can be used as an online tool (http://theripper.hawk.rocks) or it can function as a stand-alone executable file using Windows, Mac or Linux operating systems. The RIPper is also available on github (https://github.com/TheRIPper-Fungi/TheRIPper).

The RIPper is a web-based bioinformatics tool that was built using Asp.Net Core 2.1. It uses Entity Framework as its database together with Object Relational Mapping (ORM). Furthermore, the .Net bio library is used for loading and handling FASTA files. The server used for the online version of The RIPper uses a PostgreSQL database to handle multiple concurrent users. The standalone version uses SQLite. Additionally, The RIPper is compatible with all commonly used web browsers including Google Chrome, Mozilla Firefox and Microsoft Edge.

The design of The RIPper is based on the principles of measuring changes in RIP product and substrate dinucleotide frequencies using RIP indices (Selker et al., 2003; Lewis et al., 2009; Margolin et al., 1998). The frequency of dinucleotides that are recognized as suitable genetic targets of the RIP pathway are calculated using the RIP substrate index (Selker, 1990; Lewis et al., 2009), while the products produced are calculated using the RIP product index (Lewis et al., 2009; Margolin et al., 1998). The RIP composite index reflects changes in both RIP product and RIP substrate dinucleotide frequencies (Lewis et al., 2009). The RIPper calculates changes in the RIP index values and the average GC content using a sliding window approach (e.g., Table S1). The three indices are determined as indicated below, and windows with index values of x suggest that the region is affected by RIP. These RIP index values for x were previously published (Selker, 1990; Selker et al., 2003; Lewis et al., 2009; Margolin et al., 1998), but are additionally adjustable according to user preferences.

Product index value: $\frac{TpA}{ApT}$: 0.8 <  x ≥ 1.1

Substrate index value: $\frac{CpA+TpG}{ApC+GpT}$: 0.9  ≥ x

Composite index value: $\left[\frac{TpA}{ApT}\right]\text{-}\left[\frac{CpA+TpG}{ApC+GpT}\right]$: x > 0

The GC content of each window is calculated as follows:

GC content: $\frac{G+C}{G+C+A+T}$

Analyses and outputs of The RIPper

In a standard analysis, the genome-wide impact of RIP is quantified by comparing the total number of windows indicating RIP-positive index values against the total number of windows investigated for the entire genome sequence. For a window to be considered RIP-positive, all three RIP indices should indicate RIP activity. The window size for the RIP index analyses is user defined, but default parameters for genome-wide quantification is pre-set at 1,000 bp with a 500 bp step size (e.g., Table S1).

The RIPper can identify large genomic regions that are putatively extensively affected by RIP. The criteria for LRARs is adjustable according to user preferences, but default parameters constitute seven consecutive RIP-positive windows, spanning at least 4000 bp (Fig. 1). An index chain describes the total number of consecutive RIP-positive windows needed to define a LRAR. For more stringent analysis, the RIP index values considered for each window can be adjusted. The data generated using the ‘Calculate LRAR’ tool provides information on the genomic location, the total number of consecutive RIP-positive windows, average RIP index values across the length of the LRAR and the average GC content of each LRAR identified.

Another important feature of The RIPper is its ability to visualize changes in RIP index values across the length of a genomic sequence. Changes in RIP composite, product and substrate indices are provided for each window (Fig. 2). This tool further allows the user to browse through the query sequence based on a user defined range. Furthermore, changes in RIP index values, the total number of LRARs and LRAR statistics for each scaffold or chromosome within a genome assembly, can be considered individually.

The RIPper summarizes the statistics for a defined set of genomic sequences in a RIP profile. The RIP profile includes statistics on the average RIP product, substrate, and composite index values, as well as the average GC content for all LRARs identified. The RIP profile further provides information on the total size of the genome affected by RIP, its average GC content, and the total number of windows investigated.

All information generated using The RIPper is downloadable in Excel, CSV and PDF format. The parameters for determining RIP index values and GC content (calculated for each window and each step window) are all user-defined. The summary of RIP statistics is also downloadable as a RIP profile in Excel format. Fine-scale RIP analyses figures produced by The RIPper are downloadable in Portable Graphics format (PNG).

Potential for erroneously identifying RIP affected regions

GC content is one of the major factors that might cause incorrect identification of a sequence as being affected by RIP (i.e., false positives results). To investigate this issue, simulated nucleic acid sequences were generated and subjected to RIP analyses. Each simulated sequence was one million base pairs (Mbp) in size and was generated using the Random DNA Generator (http://www.faculty.ucr.edu/ mmaduro/random.htm) software. Nine sets (each containing 100 sequences) were generated, where each set contained sequences with an average GC content of respectively 10%, 20%, 30% through to 90%. All simulated sequences were subjected to full RIP analysis with The RIPper using a 1,000 base pair (bp) sliding window with a 500 bp step size. The RIP substrate index values evaluated were 0.75, 0.8, 0.85 and 0.9, while the RIP product index values evaluated were 1.1, 1.15, 1.2 and 1.25.

These ranges of RIP index values were also evaluated using the genome sequences of organisms known to be either RIP capable or incapable. These included Neurospora crassa strain OR74a, Trichoderma reesei strain QM6a and Leptosphaeria maculans strain JN3, which have all been experimentally verified to be RIP capable (Selker et al., 2003; Li et al., 2017; Fudal et al., 2009). To represent genomes in which RIP is not possible, we utilized those of the ascomycete yeast Candida albicans strain SC5314, the microsporidian fungus Encephalitozoon cuniculi strain GB-M1 and the bacterium Escherichia coli strain K-12. Experimental studies showed that C. albicans is not RIP capable (Clutterbuck, 2011), while Microsporidia diverged prior to the evolution of the RIP pathway (Horns et al., 2012) and bacteria cannot acquire cytosine and thymine transition mutations via RIP. All six genome sequences were obtained from the database of the National Centre for Biotechnology Information (NCBI; https://www.ncbi.nlm.nih.gov) using the accession numbers GCA_000182965.3, GCF_000091225.1, GCF_000005845.2, GCA_000230375.1, GCA_000182925.2 and GCA_002006585.1. As before, analyses were done using the range of values for the RIP substrate and product indices.

Validation of RIP analyses using The RIPper

The RIPper was used to determine the occurrence and frequency of RIP mutations in sequences that were experimentally and computationally shown to be affected by RIP.These included particular regions (ranging from 478 to 61,000 bp in size) in the genomes of L. maculans (Plissonneau et al., 2016), N. crassa (Margolin et al., 1998), Podospora anserina (Hamann, Feller & Osiewacz, 2000), Colletotrichum cereale (Crouch et al., 2008), Aspergillus fumigatus (Paris & Latgé, 2001), Aspergillus oryzae (Montiel, Lee & Archer, 2006), Magnaporthe grisea (Nakayashiki et al., 1999) and Chrysoporthe deuterocubensis (Kanzi et al., 2019). We also used the genomes of N. crassa strain OR74a and T. reesei strain QM6a, which are known to be RIP competent (Selker et al., 2003; Li et al., 2017). All these sequences were obtained from the NCBI database using accession numbers GCF_000182925.2/, GCA_002006585.1, KT804641.1, AF181821.1, AJ270953.1, DQ663509.1, AF202956.1, DQ327733.1, AB024423.1 and GCF_001513825.1.

The RIPper was used to determine whether this software was capable of identifying large regions affected by RIP in the sequences analyzed. Additionally, for the N. crassa (OR74a) genome assembly, we also generated genome-wide RIP statistics using the “RIP profile” tool. The genomic locations and RIP statistics of putative LRARs were determined using the “calculate LRAR” tool, while the proportion of individual N. crassa chromosomes affected by RIP were calculated using the “RIP sequence” tool. Also, for the T. reesei (QM6a) genome, chromosome-wide changes in RIP index values were compared to changes in GC content, as it was previously shown that centromeric regions are AT-rich due to RIP mutations (Li et al., 2017).

All chromosome- and genome-wide analyses used sliding windows of 1,000 bp and 500 bp steps. For particular genomic regions, fine-scale analyses were performed using 100 bp sliding windows with a step size of 50 bp. The genomic sequence containing the mating type region of C. deuterocubensis was analyzed using a 1,000 bp window and a 500 bp step size. The RIPper was used to generate graphs for visualizing changes in RIP index values and GC content across the length of the query sequences. This software was also used to generate RIP summary statistics.

Potential for erroneously identifying RIP affected regions

To investigate the influence of GC content composition on RIP statistics, simulated sequences consisting of different GC content ranges were subjected to RIP analysis using varying degrees of stringency of the RIP parameters. The overall results showed that less stringent RIP parameters (i.e., lower RIP product index values and higher RIP substrate values) more frequently led to the identification of false positives in the simulated datasets (Table S2 and Fig. S1). RIP analyses performed under the least stringent conditions (RIP product index value of 1.1 and RIP substrate index value of 0.9), identified RIP affected regions in 656 data sets, while those done with the most stringent parameters (RIP substrate index value of 0.75) allowed detection of RIP mutations in only two of the 900 simulated sequences (Fig. S1). Both of these sequences had an average GC content of 80%, where 6% of the sequences apparently consisted of RIP affected regions. However, no LRARs were identified in any of the simulated data sets (900 simulations).

Notably, at less stringent RIP substrate index values (0.8 to 0.9), more than 1% of the simulated data sets were suggested to be affected by RIP (Table S2). The occurrence and frequency of RIP mutations were more often recorded in data sets composed of high GC contents (Table S2; 50–90%) under less stringent RIP substrate parameters. Also, the frequency of erroneous identification of RIP mutations was much higher in data sets with >70% GC. This was particularly true when less stringent RIP substrate values (0.8 to 0.9) were used.

Analysis of the simulated data sets showed that stringent RIP substrate index parameters are needed to minimize the possibility of erroneous identification of RIP mutations. Our results showed that the most optimal parameter for limiting or excluding the occurrence of false positives is to use a RIP substrate index value of 0.75 (Fig. S1). Changes in RIP product value cut-offs generally had little effect on the RIP statistics, but fewer false positives were obtained using a RIP product cut-off value of 1.1.

We also investigated whether the stringency of RIP index parameters could potentially influence the outcome of analyses performed on the genome sequences of organisms known to be either RIP capable or incapable (Table S3). By making use of these two sets of genomes, it was thus possible to contrast the effects of parameter stringency in genomes where RIP is possible (positive controls; i.e., N. crassa, L. maculans and T. reesei) and genomes where RIP is impossible (negative controls; i.e., E. coli, C. albicans and E. cuniculi). Accordingly, under the most stringent RIP substrate parameter applied (0.75), no RIP positive windows were recorded in any of the negative controls (except for a single window in the genome of E. cuniculi). At this stringency level, numerous RIP positive windows and LRARs were detected in the genomes of the three positive controls. Considerable proportions of their genome sequences also constituted RIP mutations. For both the positive and negative controls, analysis at lower stringencies allowed for the identification of many more LRARs and windows suggestive of RIP. Also, as with the simulated data, RIP product parameters did not affect the RIP statistics much, but fewer windows suggestive of RIP were detected in the genomes of the negative controls. Therefore, based on these findings, windows indicating RIP mutations with product index values above 1.1 and substrate index values below 0.75 were implemented on all further analyses performed in this study.

Validation of RIP analyses using The RIPper

RIP activity has been experimentally and computationally verified in the N. crassa genome (Clutterbuck, 2011; Cambareri et al., 1989). Therefore, The RIPper was applied to the genome assembly of N. crassa for calculating genome-wide RIP statistics, to identify putative LRARs, and to determine the extent of RIP mutations in each linkage group (chromosome) of this fungus. Genome-wide RIP analyses showed that a considerable proportion of the N. crassa genome assembly (15.26%) constitutes RIP mutations (Table 1). A large number of RIP affected windows (12,545), and numerous LRARs (435) (Table 2 and Table S3) were recorded throughout this genome. Extensive RIP affected windows further showed a reduced GC content compared to that of the remainder of the genome (Table 1). The total percentage of each linkage group of N. crassa affected by RIP varied from 12.66% (linkage group I) to 17.65% (linkage group VII) (Fig. 3).

Analysis of the T. reesei genome with The RIPper showed a distinctive pattern of GC content and RIP mutations across each of the seven chromosomes of this fungus (Fig. 4, Table 3). The genomic regions corresponding to the centromeres of the T. reesei chromosomes were previously suggested to be prone to RIP mutations (Li et al., 2017). In agreement with these previous findings, reduced GC content, together with RIP index values indicating RIP activity, were observed across the length of the seven centromeric genomic regions of this fungus.

Query sequences for RIP analyses were retrieved based on previously published data (Margolin et al., 1998; Plissonneau et al., 2016; Hamann, Feller & Osiewacz, 2000; Crouch et al., 2008; Paris & Latgé, 2001; Montiel, Lee & Archer, 2006; Nakayashiki et al., 1999; Kanzi et al., 2019). Our results for these gene regions and TE sequences showed an increase in RIP product dinucleotide frequencies (RIP product index), coupled with reduced frequency in dinucleotides targeted by RIP (RIP substrate index) (Fig. 5). RIP was recorded in the AvrLm4-7 pseudogene of L. maculans, the 5S rRNA pseudogene of N. crassa, the DNA transposon of P. anserina, retrotransposon sequences of C. cereale, M. grisea, C. deuterocubensis, and within long terminal repeat sequences of A. fumigatus. Moreover, regional variation in the extent of RIP was observed in the M. grisea and C. deuterocubensis sequences. Some windows were RIP affected, while others remained unchanged by RIP. In comparison, the particular nucleic acid sequences of L. maculans, N. crassa, P. anserina, C. cereale, A. fumigatus, and A. oryzae were more extensively RIP affected across the length of the sequences investigated. The occurrence overall of less RIP in the M. grisea and C. deuterocubensis sequences were further reflected in the summary of RIP statistics (Table 4). Overall, extensively RIP affected sequences had RIP index values indicating strong RIP responses that are coupled with reduced GC content (Table 4). Conversely, the sequences with a reduced RIP response had higher GC content averages. These results were all in agreement with previous findings (Margolin et al., 1998; Plissonneau et al., 2016; Hamann, Feller & Osiewacz, 2000; Crouch et al., 2008; Paris & Latgé, 2001; Montiel, Lee & Archer, 2006; Nakayashiki et al., 1999; Kanzi et al., 2019), which suggests that the fine-scale RIP analyses implemented in The RIPper is valuable for identifying specific RIP affected regions.