High‐Intensity interval training reduces transcriptomic age: A randomized controlled trial

Abstract While the relationship between exercise and life span is well‐documented, little is known about the effects of specific exercise protocols on modern measures of biological age. Transcriptomic age (TA) predictors provide an opportunity to test the effects of high‐intensity interval training (HIIT) on biological age utilizing whole‐genome expression data. A single‐site, single‐blinded, randomized controlled clinical trial design was utilized. Thirty sedentary participants (aged 40–65) were assigned to either a HIIT group or a no‐exercise control group. After collecting baseline measures, HIIT participants performed three 10 × 1 HIIT sessions per week for 4 weeks. Each session lasted 23 min, and total exercise duration was 276 min over the course of the 1‐month exercise protocol. TA, PSS‐10 score, PSQI score, PHQ‐9 score, and various measures of body composition were all measured at baseline and again following the conclusion of exercise/control protocols. Transcriptomic age reduction of 3.59 years was observed in the exercise group while a 3.29‐years increase was observed in the control group. Also, PHQ‐9, PSQI, BMI, body fat mass, and visceral fat measures were all improved in the exercise group. A hypothesis‐generation gene expression analysis suggested exercise may modify autophagy, mTOR, AMPK, PI3K, neurotrophin signaling, insulin signaling, and other age‐related pathways. A low dose of HIIT can reduce an mRNA‐based measure of biological age in sedentary adults between the ages of 40 and 65 years old. Other changes in gene expression were relatively modest, which may indicate a focal effect of exercise on age‐related biological processes.


Human T-cell leukemia virus 1 infection (KEGG: 05166)
Human T-cell leukemia virus type 1 (HTLV-1) is a pathogenic retrovirus that is associated with adult T-cell leukemia/lymphoma (ATL). It is also strongly implicated in non-neoplastic chronic inflammatory diseases such as HTLV-1-associated myelopathy/tropical spastic paraparesis (HAM/TSP). Expression of Tax, a viral regulatory protein is critical to the pathogenesis. Tax is a transcriptional co-factor that interfere several signaling pathways related to antiapoptosis or cell proliferation. The modulation of the signaling by Tax involve its binding to transcription factors like CREB/ATF, NF-kappa B, and SRF.

RNA degradation (KEGG: 03018)
The correct processing, quality control and turnover of cellular RNA molecules are critical to many aspects in the expression of genetic information. In eukaryotes, two major pathways of mRNA decay exist and both pathways are initiated by poly(A) shortening of the mRNA. In the 5' to 3' pathway, this is followed by decapping which then permits the 5' to 3' exonucleolytic degradation of transcripts. In the 3' to 5' pathway, the exosome, a large multisubunit complex, plays a key role. The exosome exists in archaeal cells, too. In bacteria, endoribonuclease E, a key enzyme involved in RNA decay and processing, organizes a protein complex called degradosome. RNase E or R interacts with the phosphate-dependent exoribonuclease polynucleotide phosphorylase, DEAD-box helicases, and additional factors in the RNA-degrading complex.

Autophagy -animal (KEGG: 04140)
Autophagy (or macroautophagy) is a cellular catabolic pathway involving in protein degradation, organelle turnover, and non-selective breakdown of cytoplasmic components, which is evolutionarily conserved among eukaryotes and exquisitely regulated. This progress initiates with production of the autophagosome, a double-membrane intracellular structure of reticular origin that engulfs cytoplasmic contents and ultimately fuses with lysosomes for cargo degradation. Autophagy is regulated in response to extra-or intracellular stress and signals such as starvation, growth factor deprivation and ER stress. Constitutive level of autophagy plays an important role in cellular homeostasis and maintains quality control of essential cellular components. https://ipathwayguide.advaitabio.com/report/56822/contrast/73693/summary/print 11/40

Methods
For each Gene Ontology (GO) term (Ashburner et al., 2002;Gene Ontology Consortium, 2004), the number of differentially expressed (DE) genes annotated to the term is compared to the number of DE genes expected just by chance. iPathwayGuide uses an over-representation approach to compute the statistical significance of observing at least the given number of DE genes. The p-value is computed using the hypergeometric distribution as described for pORA in the Pathway Analysis section. This p-value is corrected for multiple comparisons using FDR and Bonferroni.
The classical enrichment method used above considers all GO terms to be independent. By definition, all genes annotated to a GO term are also annotated to its ancestors. Because of this, the enrichment approach counts each gene multiple times by propagating it through the GO hierarchy from the most specific term the gene is associated with, all the way to the root of the ontology. This introduces redundancy in the analysis and reports many general and non-informative terms as significant. To overcome this limitation, iPathwayGuide allows users to use two more sophisticated pruning methods: high-specificity pruning and smallest common denominator pruning. The high-specificity pruning method identifies the most specific GO terms that are significantly associated with the set of DE genes. Let us consider, BP1 = "induction of apoptosis by intracellular signals" and BP2 = "induction of apoptosis by extracellular signals," which are two of the children of BP3 = "induction of apoptosis." If enough DE genes are associated with BP1 and BP2, the high-specificity pruning will report them as significant. The smallest common denominator pruning method identifies the GO terms that best encapsulate the set of DE genes, at times consolidating significance of two or more specific terms into their common parent. In the example above, this pruning method might report BP3 as significant because it is the most specific biological term that would include all DE genes that make both BP1 and BP2 significant. https://ipathwayguide.advaitabio.com/report/56822/contrast/73693/summary/print 13/40   Catalysis of a biochemical reaction at physiological temperatures. In biologically catalyzed reactions, the reactants are known as substrates, and the catalysts are naturally occurring macromolecular substances known as enzymes. Enzymes possess specific binding sites for substrates, and are usually composed wholly or largely of protein, but RNA that has catalytic activity (ribozyme) is often also regarded as enzymatic. In this experiment, the algorithm identified 911 differentially expressed gene(s) out of ALL 5,574 gene(s

Methods
The prediction of active miRNAs (Friedman et al., 2009;Lewis et al., 2005) is based on enrichment of differentially downregulated target genes of the miRNAs. In general, miRNAs have an inhibitory effect on their targets. Therefore, for any given miRNA the method computes the ratio between the number of differentially downregulated targets and all differentially expressed targets, and compares it to the ratio of all downwardly expressed targets to all targets. Overall, iPathwayGuide calculates the probability of observing at least the number of differentially downregulated target genes for a given miRNA just by chance. This p-value is computed using the hypergeometric distribution as described for pORA in the Pathway Analysis section.

Methods
The prediction of upstream regulators is based on two types of information: i) the enrichment of differentially expressed genes from the experiment and ii) a network of regulatory interactions from our proprietary knowledge base (see the report information for details). The network is a directed graph in which the nodes represent genes, and the edges represent regulatory interactions between two genes. A signed edge in this graph consists of a source gene, a target gene, and a sign to indicate the type of signal: activation (+) or inhibition (-). To create the network, the analysis selects only those edges observed in the literature with at least a medium confidence (evidence score greater than or equal to 400). The analysis considers two hypotheses: HA. The upstream regulator is activated in the condition studied.
HI. The upstream regulator is inhibited in the condition studied.
The analysis divides the set of all the genes obtained from NCBI Gene database into several subsets based on the measurements in the experiment and the definitions shown in

Upstream regulators Z-score
For both research hypotheses, the analysis computes a Z-score for each upstream regulator z(u) by iterating over the genes in DT(u) and their incoming edges in(g). We can then compute the p-value corresponding to the z-score P as the one-tailed area under the probability density function for a normal distribution, N(0,1).

Upstream regulators predicted as activated
Here, the research hypothesis considers the upstream regulator as activated. For each upstream regulator u, the number of consistent DE genes downstream of u, DTA(u) is compared to the number of measured target genes expected to be both consistent and DE just by chance. iPathwayGuide uses an over-representation approach to compute the statistical significance of observing at least the given number of consistent DE genes. The p-value P is computed using the hypergeometric distribution , Draghici 2011).
After computing a p-value for both types of evidence, P and P , we need to combine these two probabilities into one global probability value, P that is used to rank the upstream regulators and test the research hypothesis that the upstream regulators are predicted as activated in the condition studied.
Since only a positive z-score indicates that the upstream regulator is predicted as activated, we only combine p-values for a positive z-score. Moreover, to avoid introducing false positives, only P for significant z-scores ( z ≥ 2 ) are combined. The analysis uses the standard Fisher's method to combine pvalues into one test statistic (Fisher 1925).

Upstream regulators predicted as inhibited
In parallel with upstream regulators predicted as activated, we use P and P to predict upstream regulators that are inhibited. Here, the research hypothesis states that the upstream regulators are inhibited in the conditions studied. For each upstream regulator u, the number of consistent DE genes downstream of u, DTI(u) is compared to the number of measured target genes expected to be both consistent and DE just by chance. Using the Fisher's method as above, the analysis combines P and P , where P is considered only for significant negative z-scores ( z ≤ -2).    DDX20  THOC5  GEMIN8  NUP37  MIS12  GEMIN7  NUP133  POLDIP3  AAAS  NDE1  NDC1  DYNLL1  SRSF6  RPS27  NUP43  B9D2  KNTC1  U2AF2  NUP88          6. Predicted Upstream Regulator Analysis -Chemicals, Drugs, Toxicants (CDTs)

Methods
The prediction of upstream Chemicals, Drugs, Toxicants (CDTs) is based on two types of information: i) the enrichment of differentially expressed genes from the experiment and ii) a network of interactions from the Advaita Knowledge Base (AKB v2201). The network is a directed graph in which the source node represents either a chemical substance or compound (e.g. zinc), a drug (e.g. aspirin), or a toxicant (e.g. tobacco smoke). The generic abbreviation CDT will be used henceforth to designate any of these. The edges represent known effects that these CDTs have on various genes. A signed edge in this graph consists of a source CDT, a target gene, and a sign to indicate the type of effect: activation (+) or inhibition (-). The analysis considers two hypotheses: HP. The upstream chemical, drug or toxicant is present (or overly abundant) in the condition studied.
HA. The upstream chemical, drug or toxicant is absent (or insufficient) in the condition studied.
The analysis divides the set of all the genes from AKB into several subsets based on the measurements in the experiment and the definitions shown in

Z-score
For both research hypotheses, the analysis computes a Z-score for each CDT z(u) by iterating over the genes in DT(u) and their incoming edges in(g). We can then compute the p-value corresponding to the z-score P as the one-tailed area under the probability density function for a normal distribution, N(0,1).

Upstream CDTs predicted as present (or overly abundant)
Here, the research hypothesis considers presence of the CDT. This hypothesis is useful when investigating whether the given phenotype has been impacted by the presence of a given chemical, drug or toxicant (e.g. tobacco smoke, dioxin, etc.). For each CDT u, the number of consistent DE genes downstream of u, DTA(u) is compared to the number of measured target genes expected to be both consistent and DE just by chance. iPathwayGuide uses an over-representation approach to compute the statistical significance of observing at least the given number of consistent DE genes. The p-value P is computed using the hypergeometric distribution , Draghici 2011).
After computing a p-value for both types of evidence, P and P , we combine these two probabilities into one global probability value, P that is used to rank the upstream regulators and test the research hypothesis that the upstream CDTs are predicted as present in the condition studied. The analysis uses the standard Fisher's method to combine p-values into one test statistic (Fisher 1925).

Upstream CDTs predicted as absent (or insufficient)
In parallel with upstream CDTs predicted as present, we use P and P to predict upstream CDTs that are absent. This hypothesis is relevant when investigating whether the given phenotype has been impacted by the lack of a given chemical that is necessary for the well-functioning of the organism or cell (e.g. a vitamin deficiency, iron deficiency, etc.). Here, the research hypothesis states that the upstream CDT are insufficient in the condition studied.

Methods
For each disease, the number of differentially expressed (DE) genes annotated to a disease term is compared to the number of DE genes expected just by chance. iPathwayGuide uses an over-representation approach to compute the statistical significance of observing at least the given number of DE genes. The p-value is computed using the hypergeometric distribution as described for pORA in the Pathway Analysis section. This p-value is corrected for multiple comparisons using FDR and Bonferroni.

Congenital disorders of glycosylation type I (H00118)
Congenital disorders of glycosylation (CDG) are a group of disorders caused by defects in various genes for N-glycan biosynthesis. CDG type I is defined by mutations in genes encoding enzymes which involves disrupted synthesis of the lipid linked oligosaccharide precursor and its transfer to polypeptide chain of protein, affecting N-glycan assembly in cytosol and endoplasmic reticulum. An increasing number of disorders have been discovered, with many subtypes identified. PMM2-CDG is the most common form, with over 800 patients diagnosed mostly in Europe. Almost all type present in infancy. These diseases demonstrate a broad range of clinical manifestation, associated with developmental delay, psychomotor retardation, hypotonia, seizures, hepatomegaly, microcephaly, and pericardial effusion. In this experiment, the algorithm identified 11 differentially expressed genes out of 29 genes associated with the disease. Mental retardation (MR) is a neurodevelopmental disorder characterized by low intelligence quotient (IQ) and deficits in adaptive behaviors. Although Xlinked MR has been extensively studied, and over 80 causal genes have been cloned, little is known about the genetic basis of autosomal recessive mental retardation (MRT). To date, several genes have been identified. These genes have a variety of functions and participate in multiple biochemical pathways. In addition, there are several known disease loci for which genes have not yet been identified. In this experiment, the algorithm identified 14 differentially expressed genes out of 50 genes associated with the disease.

Joubert syndrome (H00530)
Joubert syndrome and related disorders are a group of multiple congenital anomaly syndromes characterized by 'molar tooth sign', a specific midbrainhindbrain malformation seen in brain images. Joubert syndrome is associated with retinal dystrophy, nephronophthisis, liver fibrosis and polydactyly.
Most of the causative genes encode cilium-related proteins. In this experiment, the algorithm identified 11 differentially expressed genes out of 37 genes associated with the disease.

Pontocerebellar hypoplasia (H00897)
Pontocerebellar hypoplasia (PCH) is a group of inherited progressive neurodegenerative disorders with prenatal onset. Up to now ten different subtypes have been reported. All subtypes share common characteristics, including hypoplasia/atrophy of cerebellum and pons, progressive microcephaly, and variable cerebral involvement. Mutations in three tRNA splicing endonuclease subunit genes were found to be responsible for PCH2, PCH4 and PCH5. Mutations in the nuclear encoded mitochondrial arginyl-tRNA synthetase gene underlie PCH6. PCH1 is caused by homozygous mutation in the VRK1 gene. In this experiment, the algorithm identified 7 differentially expressed genes out of 15 genes associated with the disease. https://ipathwayguide.advaitabio.com/report/56822/contrast/73693/summary/print 38/40 Cytochrome c oxidase (COX) deficiency; Mitochondrial complex IV deficiency (MT-C4D) (H01368) Cytochrome c oxidase (COX) deficiency is a mitochondrial disease that is caused by the lack of the COX. Cytochrome c oxidase (COX) is the terminal enzyme of the mitochondrial respiratory chain (complex IV). Since COX is encoded by nuclear and mitochondrial genes, COX deficiency can be inherited in either an autosomal recessive or a maternal pattern. Patients can present with a number of different clinical phenotypes, including Leigh syndrome, Fatal infantile cardioencephalomyopathy, and Leber hereditary optic neuropathy. In this experiment, the algorithm identified 7 differentially expressed genes out of 15 genes associated with the disease.