IsoXpressor: A Tool to Assess Transcriptional Activity within Isochores

Abstract

Genomes are characterized by large regions of homogeneous base compositions known as isochores. The latter are divided into GC-poor and GC-rich classes linked to distinct functional and structural properties. Several studies have addressed how isochores shape function and structure. To aid in this important subject, we present IsoXpressor, a tool designed for the analysis of the functional property of transcription within isochores. IsoXpressor allows users to process RNA-Seq data in relation to the isochores, and it can be employed to investigate any biological question of interest for any species. The results presented herein as proof of concept are focused on the preimplantation process in Homo sapiens (human) and Macaca mulatta (rhesus monkey).

Introduction

Genomes are characterized by long DNA segments with a fairly homogeneous base composition known as isochores. The latter are divided into five classes: L1, L2, and H1 (GC poor) and H2 and H3 (GC rich). Several studies have shown distinct functional (e.g., replication timing, expression activity, methylation, and gene ontology) and structural (e.g., chromatin structure/architecture and gene density) properties between GC-poor and GC-rich isochores (Arhondakis et al. 2011, 2020; Bernardi 2015, 2018; Jabbari and Bernardi 2017; Jabbari et al. 2019). It is clear that a link between isochores and transcription represents a study of relevant evolutionary importance, associating the appearance of GC-rich isochores to natural selection. To aid in this important subject, we present IsoXpressor, a tool that allows for a rapid and accurate investigation of any biological process (e.g., ageing and development) or condition (e.g., cancer and genomic diseases) of any species in relation to its isochores.

Herein, preimplantation was selected as proof of concept. The utilization of compositional strategies to investigate biological processes or conditions offers different advantages. The first is to enhance our understanding on how isochores drive functionality by coordinating and shaping regulation mechanisms, where the latter is known to be correlated to GC levels. The second one is of practical relevance, because our approach identifies isochores, that is, specific genomic regions, that represent targets for a more detailed investigation regarding the genes they contain and their function (Arhondakis et al. 2020).

Materials and Methods

IsoXpressor is implemented in Python and is freely available at https://github.com/lorrainea/IsoXpressor under GNU GPLv3. A wiki page to facilitate its usage is also available at the same location.

In this section, the external tools used by the pipeline are discussed as well as how IsoXpressor uses their output to produce analytical results for each isochore.

External Tools

The IsoXpressor pipeline makes use of two external tools to be able to assess the transcriptional activity within the isochores.

REAL

REAL (Frousios et al. 2010) is an efficient short-read aligner for next-generation sequencing data. It takes as input two files, the first is a FASTA file containing a reference genome sequence, and the second is a FASTA or FASTQ file containing short reads. It then aligns the reads onto the reference genome.

There are several parameters used by IsoXpressor to run REAL. These are used to accustom the output of IsoXpressor. These include -s, the maximum number of errors allowed in the seed part of the read; -e, the total maximum number of errors; and -ℓ, the length of the seeds.

The output of REAL is a text file containing properties of the aligned reads. These include, but are not limited to, the length of the read $R_{\ell }$; the id of the reference sequence; that is, the chromosome to which the read was aligned to, which we denote by C_R; as well as the starting position R_s of where the read is aligned to the chromosome.

IsoSegmenter

The second external tool used by IsoXpressor is IsoSegmenter (Cozzi et al. 2015). It is a tool designed to segment a genomic sequence into isochores. It takes as input a single sequence in FASTA format and splits the sequence into windows of a user specified length. These windows are assigned to different isochore classes (L1, L2, H1, H2, and H3) depending on the GC level (percentage of G and C bases) identified in each window.

We make use of the window parameter -w for IsoSegmenter to be able to scan the chromosomes of the input genome.

IsoSegmenter outputs a list of isochores denoting properties such as their GC level, the starting position I_s of the isochore within chromosome C_I, as well as the ending position and the length of the isochore denoted by $I_{\ell }$.

IsoXpressor

Let us describe the main pipeline of the tool.

Input: A reference genome and a collection of RNA-Seq read data sets, which are partitioned into k processes or conditions.

Output: The number of reads that align within each isochore of the reference genome as well as the corresponding Transcripts Per Kilobase Million (TPM) score (Li and Dewey 2011) or Reads Per Kilobase Million (RPKM) score (Mortazavi et al. 2008) for each isochore class in each process or condition.

1. The reads are aligned onto the reference genome using the REAL tool.

2. IsoSegmenter extracts the isochores for each chromosome of the reference genome.

3. The count of the aligned reads within each isochore for each chromosome is computed using the output files from Steps 1 to 2.

4. Using the read counts from Step 3, the TPM and RPKM scores of each isochore for each process or condition are computed, representing the transcriptional activity. Note that these scores are adapted to the isochore lengths rather than gene lengths. Details on how TPM and RPKM are computed can be found in the corresponding section below.

Counting Reads

Given the set S of computed isochores and the set T of read sets, IsoXpressor computes the count $r_{i,j}$ of the number of reads identified within each isochore S_i from each read set T_j. For each isochore and read, such that C_I = C_R, more than half of the base pair positions $R_s,\dots ,R_s+R_{\ell }-1$ should overlap with $I_s,\dots ,I_s+I_{\ell }-1$. More formally, as long as one of the two following formulae hold, the count of the number of reads located within this specific isochore is increased.

$$R_s+R_{\ell }-1-I_s>I_s-R_s \text{and} R_s+R_{\ell }\le I_s+I_{\ell }$$

or

$$I_s+I_{\ell }-1-R_s>(R_s+R_{\ell }-1)-(I_s+I_{\ell }-1) \text{and} R_s\ge I_s.$$

TPM and RPKM Scores of Isochores

The TPM of an isochore is computed as follows:

1. Divide the read counts by the length of each isochore in kilobases to obtain the reads per kilobase (RPK).

2. Count up all the RPK values in a sample and divide this number by 1,000,000, to assess the “per million” scaling factor.

3. Divide the RPK values by the “per million” scaling factor to obtain the TPM of each isochore.

The RPKM of an isochore is computed as follows:

1. Count up the total reads in a sample and divide that number by 1,000,000 (“per million” scaling factor).

2. Divide the read counts by the “per million” scaling factor to obtain the reads per million.

3. Divide the reads per million values by the length of the isochore, in kilobases, to obtain the RPKM of each isochore.

IsoXpressor computes the TPM or RPKM expression for each isochore as described in detail below; It uses the same computation for sequencing depth (number of unique reads aligned to the reference genome [Sims et al. 2014]) and isochore length rather than the more commonly used gene length.

Given k conditions or processes, $c_0,\dots ,c_{k-1}$, where c_i is processed using read counts $r_{i,m},\dots ,r_{i,m+n-1}$, where $m,\dots ,m+n-1$ corresponds to the indices of read sets associated with condition c_i within the set of all read sets, the TPM and RPKM scores for each isochore and process or condition are formally computed as follows:

$$\begin{array}{c}{A}_j=\sum _{i=0}^{|S|-1}\frac{r_{i.j}\times {10}^3}{I_{{\ell }_i}}{\text{TPM}}_{i,c_i}=\sum _{j=m}^{m+n-1}\frac{r_{i.j}\times {10}^9}{I_{{\ell }_i}\times A_j\times n}\\ B_j=\sum _{i=0}^{|S|-1}{r}_{i.j}{\text{RPKM}}_{i,c_i}=\sum _{j=m}^{m+n-1}\frac{r_{i.j}\times {10}^9}{I_{{\ell }_i}\times B_j\times n}.\end{array}$$

Table 1 contains synthetic counts $r_{i,j}$ of each read set T_j which occurs in an isochore S_i where $|S|=3$ and $|T|=4$. Note that random values have been used in this example.

Table 1.

Read Counts for Three Isochores and Four Read Sets

Chromosome	Isochore Class	GC Level	Isochore Start	Isochore End	Isochore Size	Reads 1	Reads 2	Reads 3	Reads 4
7	H2	47.41	1	2,396	2,396	13	5	21	1
2	H2	50.94	673,437	873,436	200,000	725	1,007	808	3,128
9	H3	56.12	873,437	1,573,436	700,000	6,666	16,489	19,444	17,242

Table 2 contains the TPM results for table 1 given that k = 2 where c₀ (Condition 1) is computed using reads $r_{i,0},\dots ,r_{i,1}$ and c₁ (Condition 2) is computed using reads $r_{i,2},\dots ,r_{i,3}$.

Table 2.

TPM Scores for Table 1 Where k = 2

Chromosome	Isochore Class	GC Level	Isochore Start	Isochore End	Isochore Size	Condition 1	Condition 2
7	H2	47.41	1	2,396	2,396	180,072.03	113,115.77
2	H2	50.94	673,437	873,436	200,000	179,648.25	241,966.59
9	H3	56.12	873,437	1,573,436	700,000	640,279.71	644,917.64

Table 3 contains the RPKM results for table 1 given that k = 2 where c₀ (Condition 1) is computed using reads $r_{i,0},\dots ,r_{i,1}$ and c₁ (Condition 2) is computed using reads $r_{i,2},\dots ,r_{i,3}$.

Table 3.

RPKM Scores for Table 1 Where k = 2

Chromosome	Isochore Class	GC Level	Isochore Start	Isochore End	Isochore Size	Condition 1	Condition 2
7	H2	47.41	1	2,396	2,396	426.02	226.41
2	H2	50.94	673,437	873,436	200,000	388.65	483.52
9	H3	56.12	873,437	1,573,436	700,000	1,316.07	1,289.65

IsoXpressor also computes the average TPM and RPKM scores for each chromosome as well as each isochore class. These are output as CSV files along with the average TPM and RPKM scores for the read counts as well as an output of the raw read counts. It also outputs a visualization of the chromosome profiles for each chromosome as seen in the Supplementary Material online.

Experimental Results

The whole human genome (version GRCh38.p13) was used as input as well as RNA-Seq data from four preimplantation stages from the same species: Condition 1 (two-cell/GSM116020–GSM116022); Condition 2 (four-cell/GSM116023–GSM116026); Condition 3 (eight-cell/GSM116027–GSM116037); and Condition 4 (Morula/GSM116038–GSM116040) (Xue et al. 2013).

We also used the whole genome of rhesus monkey (version Mmul_10) as input as well as RNA-Seq data from five preimplantation stages from the same species: Condition 1 (two-Cell/GSM2310522); Condition 2 (four-cell/GSM2310523); Condition 3 (eight-cell/GSM2310524); Condition 4 (Morula/GSM2310525); and Condition 5 (Blastula/GSM2310526); each containing a single RNA-Seq read data set (Wang et al. 2017).

The default parameters of IsoXpressor were used in this experiment ($\mathrm{e}=5,\hspace{0.17em}\mathrm{s}=2,\hspace{0.17em}\ell =32$, and $\mathrm{w}=100,000$). The following results show the output produced by IsoXpressor with these inputs and parameters. More experimental results can be found in the Supplementary Material online.

Figure 1a shows the profile of GC-rich Chromosome 19 in human and figure 1b shows likewise of rhesus monkey. Similarly, figure 2a and b shows the transcriptional activity of the isochores of Chromosome 14 in human and rhesus monkey, respectively. As observed for Chromosome 19, coordinated transcriptional activity can be assessed for Chromosome 14. From the figures, it is evident that there is an extensive coordinated activation of blocks of adjacent isochores as preimplantation advances, rather than random activation. Regarding the isochore profiles (bottom) of Chromosome 19 and 14 of human and rhesus monkey as depicted in figures 1 and 2, respectively, these agree with those reported for human in Cozzi et al. (2015), as well as in previous studies (Costantini et al. 2009). In addition, the isochores’ distribution across rhesus monkey chromosomes are given for the first time. The results are consistent for all chromosomes (see the Supplementary Material online).

Fig. 1.

(a) TPM profile of each stage along the isochores of Chromosome 19 (top) of human; colored bars show the isochore profile (bottom). (b) TPM profile of each stage along the isochores of Chromosome 19 (top) of the rhesus monkey; colored bars show the isochore profile (bottom).

Fig. 2.

(a) TPM profile of each stage along the isochores of Chromosome 14 (top) of human; colored bars show the isochore profile (bottom). (b) TPM profile of each stage along the isochores of Chromosome 14 (top) of the rhesus monkey; colored bars show the isochore profile (bottom).

Summarizing, the expression of the chromosomes’ isochores appears to hold a nonrandom profile during the preimplantation process. Our findings add evidence of a sequential activation of adjacent blocks of isochores as preimplantation advances. The latter may reflect regulatory effects such as rearrangements of chromatin structure, supporting the presence of a link between compositional properties to structural, spatial, and functional properties at a chromosomal level (Sabbia et al. 2009; Arhondakis et al. 2011; Jabbari and Bernardi 2017; Bernardi 2018, 2019; Hansen et al. 2018; Lamolle et al. 2018; Jabbari et al. 2019).

Figure 3 depicts the percentages of the up- and downregulated isochores with a fold change above two. The fold change is computed as the $\hspace{0.17em}{\text{log}}_2$ ratio of the expression (TPM or RPKM) of each stage over that of the two-cell stage. In this work, upregulation of an isochore indicates an increase in its transcriptional activity as preimplantation advances, whereas downregulation a decrease (silencing). In other words, the upregulated isochores here are those that have a 2-fold higher expression (TPM or RPKM) in another condition compared with the two-cell condition (see position 58 MB in fig. 2b). The downregulated isochores here are those that have a 2-fold lower expression (TPM or RPKM) in another condition compared with the two-cell (see position 58 MB in fig. 2a). The up- and downregulated percentages are computed by dividing the number of up- or downregulated isochores of each class over the total number of isochores of the corresponding class.

Fig. 3.

(a) Percentages of the upregulated isochores in human above a 2-fold change. (b) Percentages of the downregulated isochores in human above a 2-fold change. (c) Percentages of the upregulated isochores in rhesus monkey above a 2-fold change. (d) Percentages of the downregulated isochores in rhesus monkey above a 2-fold change.

In human, we observe that after the four-cell stage there is an increase in the percentage of the GC-rich ones (H2 and H3; fig. 3a) and a decrease of the downregulated ones (fig. 3b). This can be translated as a tendency of GC-rich isochores to increase their expression as preimplantation advances. This is similarly observed for H1, although it is less evident than it is for H2 and H3. On the contrary, L1 and L2 display a constant profile as preimplantation advances. Hence, a lower percentage within upregulated (fig. 3a) compared with downregulated isochores (fig. 3b) is observed. The latter indicates that most of the L1 and L2 isochores tend to reduce their expression activity as preimplantation advances.

In rhesus monkey, profiles are similar to human, with the exception of the H1 class. Indeed, within the upregulated isochores (fig. 3c) after the first stage (two-cell to four-cell), the GC-rich ones (H2 and H3) increase, whereas the GC-poor ones (L1, L2, and H1) decrease (fig. 3c). It is worth mentioning that L1 and L2 classes reach a decrease of ≈50% (fig. 3c). The above indicates that the GC-rich isochores increase their expression (upregulated) as preimplantation progresses. Regarding the downregulated isochores (fig. 3d), during the early stage (between two- and four-cell) H2 and H3 classes exhibit a peak compared with the GC-poor classes. This points toward a possible silencing of most of the H2 and H3 isochores at very early stages. Interestingly, percentages of L1 and L2 increase ∼3 times in the late stages of preimplantation (Morula and Blastula), further supporting their silencing or downregulation in these stages.

The observed differences between human (fig. 3a and b) and rhesus monkey (fig. 3c and d) may partially reflect experimental differences, without excluding potential biological diversification between these species. To conclude, our observations within each species support a preferential activation (upregulation) of GC-rich isochores at late preimplantation stages, and a silencing (downregulation) of the GC-poor ones.

Conclusion

We introduced IsoXpressor, an efficient and accurate tool dedicated to the analysis of isochores’ transcriptional activity. Our tool offers the possibility to investigate any biological process or condition and can be applied to any species. To the best of our knowledge, no other automated tool is currently available which is capable of processing RNA-Seq data related to isochores.

Despite the solidity of the results presented, it would be relevant to use larger data sets in order to gain a more holistic perspective on the case study under investigation. However, we do believe that the findings presented herein offer novel insights on how genome organization, in terms of isochores, acts upon regulation during the preimplantation process in both human and rhesus monkey, for example, affecting chromatin architecture (Bernardi 2015, 2019; Du et al. 2017).

We anticipate that IsoXpressor will open new perspectives on investigating important biological questions, where the genome is considered as a functional and structural ensemble, with its most fundamental property, the base composition, shaping regulation.

Data Availability

The data underlying this article are available in the article and in its Supplementary Material online.

Supplementary Material

Supplementary data are available at Genome Biology and Evolution online.