What transcriptomic changes appear in muscle of CD163/pAPN double-knockout pigs?

Four RNA classes from one muscle tissue

The paper reports a whole-transcriptome sequencing dataset built from longissimus dorsi muscle, the main cut used to assess meat quality and lean yield in pigs. Each of six 11-month-old male Large White pigs — three double-knockout (DKO) and three wild-type (WT) — contributed one muscle sample taken from the 10th–11th rib region.

Two libraries were made per sample. An rRNA-depleted library captured mRNAs, lncRNAs and circRNAs, sequenced as 150-bp paired-end reads; a small RNA library captured miRNAs, sequenced as 50-bp single-end reads. Both ran on an Illumina HiSeq 3000. This design matters because the four RNA classes are measured with different library preparations and different abundance units, so their counts cannot be compared directly.

Detection counts differ by RNA class

After quantification, the authors report 18,658 mRNAs and 18,386 lncRNAs with FPKM above zero, 9,729 circRNAs with SRPBM above zero, and 419 miRNAs with TPM above zero in the muscle samples. FPKM, SRPBM and TPM are three different normalisation schemes: the first scales by transcript length and sequencing depth, the second counts backsplicing junction reads, and the third expresses miRNA reads as transcripts per million.

Those totals describe what was detected, not what changed. A large catalogue of expressed transcripts is the baseline against which the DKO–WT comparison is made, and the size of each catalogue differs by orders of magnitude between classes.

Sequencing quality supported downstream comparison

The six rRNA-depleted libraries yielded between roughly 163.5 million and 180.6 million raw reads per sample, and between 156.9 million and 177.6 million clean reads after filtering. Clean read rates exceeded 95.9% in every sample, and mapping rates to the Sscrofa11.1 reference genome were about 96.0–96.9% for the lncRNA/mRNA libraries.

Small RNA libraries were much shallower, with total reads between about 5.95 million and 8.56 million and genome mapping rates of 76.18–79.86%. The authors treat these metrics as evidence that the data are suitable for downstream transcriptomic analysis, which is the standard bar for a data descriptor rather than a functional claim.

Global profiles stay similar across genotypes

Correlation analysis of variance-stabilised expression found high overall similarity between DKO and WT groups: r = 0.920 for lncRNAs and r = 0.984 for mRNAs when group means were compared. This is the paper's clearest statement that the double knockout did not produce a broad, uniform reorganisation of the muscle transcriptome under the conditions tested.

Clustering and principal component analysis told a more mixed story. The three WT samples clustered together on lncRNA expression and separated from the three DKO samples, but in the mRNA analysis DKO2 clustered with the WT samples while DKO1 and DKO3 formed their own subcluster. The authors describe this as partial rather than complete separation and attribute it to biological or technical heterogeneity among individuals.

mRNAs and lncRNAs carry most of the signal

Differential expression used DESeq2 with an adjusted p-value below 0.05 and an absolute log2 fold change above 1. Under those thresholds the study identified 606 differentially expressed mRNAs (342 up, 264 down) and 281 differentially expressed known lncRNAs (72 up, 209 down). The lncRNA response was therefore skewed toward downregulation, while the mRNA response leaned toward upregulation.

The non-coding circular and micro RNA layers changed far less: 20 differentially expressed circRNAs (11 up, 9 down) and a single known miRNA. The authors highlight several strongly altered mRNAs, including upregulated PRSS36 and EGR1 and downregulated PFKFB3 and SLPI, and note that circRNA ssc_circ_0008362 showed very high statistical significance despite the small total number of changed circRNAs.

Expression changes are descriptive, not mechanistic

The paper is explicit that these mRNA changes should be read as descriptive transcriptomic observations rather than direct evidence of a defined functional pathway. EGR1, PFKFB3 and SLPI are described as associated with stress response, metabolic regulation or tissue microenvironment processes, but the dataset does not test whether those processes are altered in the animals.

A second reason for caution is tissue composition. Longissimus dorsi contains myofibers, adipocytes, fibroblasts, endothelial cells and resident immune cells, so a shift in the proportion of, for example, adipose or stromal cells could move global expression patterns without any genotype-specific change. The authors raise this as a possible explanation for WT2 deviating from the other WT replicates in some analyses.

Three replicates per group limit separation

Each genotype is represented by three biologically independent animals, all 11-month-old males of the same breed and housing condition. The authors state that this gives limited statistical power to fully separate genotype-associated effects from biological heterogeneity, which is consistent with the incomplete clustering seen for mRNAs.

The dataset is therefore framed as a resource for transcriptomic characterisation and hypothesis generation, not a definitive functional assessment of CD163/pAPN double knockout in muscle. The authors propose larger sample sizes, additional tissues, or pathogen-challenged spleen or muscle tissue as ways to validate the observations.

A reference for the non-coding layers

The most distinctive contribution here is not the mRNA list but the paired profiling of four RNA classes in the same muscle samples. Because the mRNA and lncRNA layers show hundreds of changes while circRNA and miRNA show 20 and 1, the dataset's practical value lies in letting future work ask whether the non-coding layers are genuinely buffered or simply underpowered at this replicate count.

For a reader weighing the safety question, the relevant comparison is between the high group-mean correlations (0.920 for lncRNAs, 0.984 for mRNAs) and the hundreds of transcripts crossing the differential expression thresholds. Both are true at once: the muscle transcriptome of these DKO pigs is broadly similar to wild type, yet a specific and enumerable set of mRNAs and lncRNAs differs — and that set, not the global correlation, is what follow-up studies would need to test.

Sources used in this answer