KR102657185B1 - Method for detecting cellular senescence using cellular senescence biomarkers GAS2L3 and WDR76 from isolated cells - Google Patents

Abstract

It relates to a method for screening markers of cellular aging using machine learning, cellular aging biomarkers, and a method for screening senolytic agents using the same. A method for screening cellular aging markers using machine learning according to one aspect According to this, by establishing an analysis pipeline that performs meta-analysis on aging cell RNA-seq data and securing a sufficient number of samples, statistically more significant genes can be selected, and through machine learning methodology, aging can be identified among various variables. It has the effect of identifying candidates for cell-specific markers. In addition, the marker of cellular senescence derived by the above method is a more meaningful genetic signature that can specify cellular senescence in various types of cells than existing studies, and can be usefully used to specify or detect senescent cells. In addition, it has an effect that can be useful in selecting senolytic agents.

Description

Method for detecting cellular senescence using cellular senescence biomarkers GAS2L3 and WDR76 from isolated cells {Method for detecting cellular senescence using cellular senescence biomarkers GAS2L3 and WDR76 from isolated cells}

It relates to a method of screening markers of cellular aging using machine learning, cellular aging biomarkers, and a method of screening senolytic agents using the same.

Cellular senescence refers to permanent cell cycle arrest of cells caused by internal/external stimuli. Cells that have undergone cellular senescence are known to secrete a senescence associated secretory phenotype (SASP), causing various inflammatory responses, and as a result, controlling age-related diseases, including cancer. The senolysis technique, which aims to treat various diseases through selective removal of senescent cells, has recently been in the spotlight. For example, research targeting aging cells was designated as one of the top 10 technologies selected by MIT in 2020.

In various existing studies, it was confirmed that the lifespan of mice treated with senolytic drugs increased and age-related diseases decreased. However, due to the heterogeneous nature of senescent cells, there is a lack of markers that can specifically select senescent cells, and senescent cell markers to date are also expressed in cells other than senescent cells. Recently, a small number of drugs, including senolytic drugs developed at the Mayo Clinic in the United States, have been clinically tested, but most of them do not show clear effects or cause side effects, making no progress.

Meanwhile, cellular aging is known to have very diverse characteristics depending on the type of cell in which aging occurs and the type of derivative that induces aging. However, not much individual research has been conducted on this. Therefore, there is a need for an approach to discover markers by identifying the molecular characteristics of each type of senescent cell.

Cellular aging is caused by various inducers, such as replicative senescence caused by telomere attrition, oncogene induced senescence induced by the expression of oncogenes, and therapy induced senescence caused by drugs or irradiation during the treatment of a disease. According to existing studies, cellular senescence is known to have stronger characteristics depending on the type of cell than the type of inducer type, but the upper path of substances secreted by senescent cells depends on the type of inducer by which cellular senescence is induced. There are reports that the expression of cell aging is different, and existing studies related to cellular aging have the disadvantage of not being able to select relatively significant genes due to limitations in the number of samples per experiment. Therefore, a method of analysis that complements these problems is needed.

One aspect includes acquiring a plurality of different experimental data including RNA sequencing data of a plurality of different causes of senescent cell populations and control cells; The transcriptome of the obtained plurality of different experimental data is quantified through an analysis pipeline to reduce the batch effect derived from different experiments to obtain batch-corrected data. steps; Among the batch-corrected data, selecting differentially expressed genes compared to control cells for each of a plurality of senescent cell groups of different causes in a training set; Selecting genes that are commonly present in a plurality of senescent cell populations of different causes among the selected genes; And a mechanical learning method comprising selecting a marker of cellular aging among the selected genes using a regression model capable of supervised learning using the expression value of the selected gene as an independent variable. A method for selecting markers of cellular aging using the present invention is provided.

Another aspect provides a computer-readable recording medium recording a program for executing the method on a computer.

Another aspect provides a device for detecting markers of cellular senescence performing the method.

Another aspect provides markers of cellular senescence derived by the above method.

Another aspect provides a cellular aging prediction model generated using markers of cellular aging derived by the above method.

Another aspect is the expression of any one or more genes or proteins selected from the group consisting of RRM2B, DUSP6, GSAP, AKAP6, C2orf92, AC008012.1, CLDN11, HMGN2, CHAF1B, KLHL13, MXD3, AUTS2, NAV2, NCAPD2, and PXMP2, or A composition for detecting cellular senescence comprising an agent capable of measuring the level of activity is provided.

Another aspect is one or more genes or proteins selected from the group consisting of EZH2, TMPO, AMOT, GAS2L3, SYNE2, HMGB2, CDCA7L, WDR76, H2BC8, MMP15, ACSL5, CD36, PLD1, CYP26B1, GSAP, H2BC12, and H2AC6. A composition for detecting cellular senescence comprising an agent capable of measuring expression or activity levels is provided.

Another aspect provides a composition for detecting cellular senescence, comprising an agent capable of measuring the expression or activity level of any one or more genes or proteins selected from the group consisting of GAS2L3, AMOT, and WDR76.

Another aspect provides a kit for detecting cellular aging comprising the composition.

Another aspect provides a method for detecting cellular senescence comprising measuring the expression or activity level of the gene or protein from an isolated biological sample.

Another aspect includes contacting the gene or protein with a test substance; and selecting a test substance that has changed the expression or activity level of the gene or protein as a senolytic drug compared to an untreated control group.

One aspect is to provide a method for screening for markers of cellular senescence.

As used herein, the term “cell” is meant to include individual cells, but can also mean the specific tissue or organ from which it originates.

The cells include endothelial cells, smooth muscle cells, macrophages, fibroblasts, retinal pigment epithelial cells, other epithelial cells (e.g., lung epithelial cells or renal epithelial cells), immune cells (e.g., macrophages), and chondrocytes. , or stem cells (eg, mesenchymal stem cells), or nerve cells (eg, neurons).

The term “cellular senescence” refers to epigenetic markers that emerge from the cell cycle, are consistent with aging, or are senescent cellular markers (e.g., senescence-associated beta-galactosidase, or inflammatory cytokines). It may refer to cells that express . Cellular senescence may be partial or complete.

The term “epigenome” or “epigenetics” refers to modifications and structural changes within a cell that control the expression of nucleic acids (e.g., engineered nucleic acids) or genomic information in the cell. Changes to the epigenome occur and induce during the processes of embryonic development, disease progression, and aging.

The term “epigenetic clock” can refer to an age estimator or an innate biological process. In some embodiments, the age estimator is an epigenetic age estimator. For example, an epigenetic age estimator may be a set of CpG dinucleotides that, when used in combination with a mathematical algorithm, can be used to estimate the age of a DNA source, including a cell, organ, or tissue. In some embodiments, the age estimator is a DNA methylation-based (DNAm) age estimator. In some embodiments, the DNAm age estimator can be calculated as an age correlation using the Pearson correlation coefficient r between DNA methylation-based (DNAm) age (also known as estimated age) and chronological age. In some embodiments, the DNA methylation-based (DNAm) age estimator may be a single-tissue DNA methylation-based age estimator. In some embodiments, the DNA methylation-based age estimator may be a multi-tissue DNA methylation-based age estimator.

Senescent cells may exhibit any one or more of the seven characteristics below. (1) Aging growth arrest is essentially permanent and cannot be reversed by known physiological stimuli. (2) Senescent cells increase in size and are more than two-fold enlarged compared to the size of their non-senescent counterparts. (3) Senescent cells express senescence-associated β-galactosidase (SA-β), which partly reflects an increase in lysosomal mass. (4) Most senescent cells express p16INK4a, which does not universally It is not expressed by resting or terminally differentiated cells. (5) Cells that age due to continuous DDR signaling have DNA segments with chromatin alterations reinforcing senescence. ), these foci contain activated DDR proteins and are distinguishable from transient damage foci, which are called dysfunctional telomeres or telomere dysfunction-induced foci (TIFs). (6) Senescent cells can express and secrete molecules associated with aging, which in certain cases is observed in the presence of persistent DDR signaling and, in certain cases, is persistent for its expression. (7) The nuclei of senescent cells lack structural proteins such as Lamin B1 or chromatin binding proteins such as histones and HMGB1, as described in Freund et al. Mol Biol. Cell 23:2066-75 (2012); Ivanov et al., J. Cell Biol. /jcb.201212110, page 1-15] (published online July 1, 2013); Funayama et al., J. Cell Biol 175:869-80 (2006).

Senescent cells and senescent cell-related molecules can be detected by techniques and methods described in the art. For example, the presence of senescent cells in tissue can be analyzed by histochemistry or immunohistochemistry to detect the senescence marker, SA-beta galactosidase (SA-β) (see, e.g., Dimri et al. ., Proc. Acad. Sci. USA 92: 9363-9367 (1995)), using any of a number of immunochemical methods practiced in the art. The expression of p16 mRNA in a cell can be measured by a variety of techniques practiced in the art, including quantitative PCR, and the presence and level of senescent cell-related polypeptides (e.g., polypeptides of SASP). Can be measured using automated and high-throughput assays, such as the automated Luminex array assay described in the art (e.g., Coppe et al., PLoS Biol 6: 2853-68 (2008) ] reference).

As used herein, the term "marker" refers to a substance that can be identified by distinguishing normal cell groups from senescent cell groups, and is a polypeptide, protein or nucleic acid, gene, lipid, glycolipid, or sugar that shows an increase or decrease in the senescent cell group of the present invention. Includes all organic biomolecules such as proteins or sugars.

When described with reference to FIG. 1, the method includes acquiring a plurality of different experimental data including RNA sequencing data of a group of senescent cells and control cells having a plurality of different causes of aging (S10);

The transcriptome of the obtained plurality of different experimental data is quantified through an analysis pipeline to reduce the batch effect derived from different experiments to obtain batch-corrected data. Step (S20);

Among the batch-corrected data, selecting differentially expressed genes compared to control cells for each group of senescent cells with a plurality of different causes of aging in the training set (S30);

Among the selected genes, selecting genes commonly present in a group of senescent cells having a plurality of different causes of aging (S40); and

It may include a step (S50) of selecting markers of cellular aging among the selected genes using a regression model capable of supervised learning using the expression value of the selected gene as an independent variable. .

In one embodiment, the plurality of different experimental data may include a senescent cell group and a control cell group (proliferating cell group). The senescent cell group may be a plurality of senescent cell groups having at least two different causes of aging. The cause of aging may be any one selected from the group consisting of replicative senescence, oncogene induced senescence, and therapy induced senescence. For example, the experimental data may include a replicative senescent cell group, an oncogene-induced senescent cell group, and a control cell group. Additionally, the experimental data may include a replicative senescent cell group, an oncogene-induced senescent cell group, a treatment-induced senescent cell group, and a control cell group. Each cell population in the experimental data may be derived from the same laboratory or different laboratories.

In one embodiment, the experimental data may include data measured experimentally from biological samples, obtained from known literature, or stored in a database. The database may include National Center for Biotechnology Information (NCBI), Gene Expression Omnibus (GEO), European Bioinformatics Institute databases, or European Nucleotide Archive. The RNA sequencing data may be in FASTQ format.

In one embodiment, the step of classifying the obtained plurality of different experimental data into a training set and a test set may be further included. The classifying step may be performed by acquiring the experimental data, or may be performed by acquiring batch-corrected data. The experimental data of the training set and the test set are stored in a predetermined ratio (e.g., about 90:10, about 80:20, about 75:25, about 60:40, about 50:50, about 40:60, about 25 :75, about 20:80, or about 10:90). Additionally, the training set and the test set may include both a control group and a group of senescent cells having a plurality of different causes of aging. For example, the training set or test set may include a replicative senescent cell group, an oncogene-induced senescent cell group, and a control cell group. Additionally, the training set or test set may include a replicative senescent cell group, an oncogene-induced senescent cell group, a treatment-induced senescent cell group, and a control cell group. Each cell population within the training set or test set may be derived from the same laboratory or from different laboratories.

In one embodiment, the step of obtaining the batch-corrected data includes configuring the RNA sequencing data as Fastq data and inputting the Fastq data; Obtaining clean reads by quality controlling and pre-processing the input Fastq data; Aligning the clean reads to the corresponding genome or transcript; and/or assembling a transcript of the genome or transcript corresponding to the clean read; It may include the step of quantifying the assembled transcript.

In one embodiment, the quality control may be performed using a FASTQ program, NGSQC program, or RNA-SeQC program. The step of obtaining clean reads through preprocessing may be performed using Trimmomatic, PRINSEQ, or Soapnuke. The alignment step may be performed using Salmon, Ensemble, Tophat2, HISAT2, STAR, BWA, or Bowtie. The step of assembling the transcript may be performed using Tximeta, Cufflinks, StringTie, Trinity, SOAPdenovoTrans, or Trans -It may be performed using AByS. The quantification step may be performed using limma FeatureCount, HTSeq-Count, Cufflinks, eXpress, RSEM, DEXSeq, Kallisto, Sailfish, or Salmon.

In one embodiment, the step of removing genes with significantly low expression levels in all cell populations for the training set among the batch-corrected data from the aging marker candidate genes may be further included. The genes with significantly low expression levels may be removed through the edgeR package.

In one embodiment, the step of selecting differentially expressed genes is when the significance p value (probability value) is less than 0.05 or the expression level of genes whose expression level has changed compared to the control group in the senescent cell group is 0.05 or less. Selecting genes with a change amount of about 2 times or more or about 1/2 times or less; Alternatively, it may further include verifying the differentially expressed genes by performing GO (gene ontology) analysis. The analysis of the differentially expressed genes was conducted using the R package edgeR (v3.10) and may be performed by constructing an analysis design according to the aging status information and batch information of the sample after TMM normalization.

In one embodiment, the regression model may include a machine learning model, for example, a supervised-learning model. In one embodiment, the regression model may include a LASSO model, a linear regression model, and/or other supervised-learning model. The independent variable (feature point) of the regression model may be the expression value (TMM normalized Log ₂ CPM (count per million)) of the selected gene, and the dependent variable is 0 when the data state is proliferating control, aging In case of sentient status, it may be assigned as 1.

Additionally, the regression model may include leave one out cross validation (LOOCV). The LOOCV creates a total of N models (as many as the number of samples), excludes only one sample when creating each model, and calculates the test set performance with the excluded sample to calculate the N performance. This may include an averaging method.

In one embodiment, the method may further include the step of testing markers of the selected senescent cells.

The testing step includes (i) using at least a portion of the plurality of different experimental data as a test set to determine whether markers of senescent cells selected in the same way as the training set are expressed in the test set. steps; and/or (ii) applying the selected marker of cellular aging to a training set to generate an aging prediction model, and using this to perform a test on the test set.

In one embodiment, step (ii) generates an aging prediction model by applying the acquired plurality of different experimental data and markers of cellular aging obtained through different paths to the training set, and uses this to generate an additional aging prediction model. It may include performing a test on a test set and comparing the test result with the test result of step (ii).

In one embodiment, the aging prediction model may include a machine learning model, for example, a supervised-learning model. In one embodiment, the aging prediction model may be generated through a support vector machine.

The aging prediction model is one or more selected from the group consisting of the selected genes RRM2B, DUSP6, GSAP, AKAP6, C2orf92, AC008012.1, CLDN11, HMGN2, CHAF1B, KLHL13, MXD3, AUTS2, NAV2, NCAPD2, and PXMP2. The gene may be generated through a machine learning model, for example, a support vector machine.

The aging prediction model is any selected from the group consisting of selected genes EZH2, TMPO, AMOT, GAS2L3, SYNE2, HMGB2, CDCA7L, WDR76, H2BC8, MMP15, ACSL5, CD36, PLD1, CYP26B1, GSAP, H2BC12, and H2AC6. One or more genes may be generated through a machine learning model, for example, a support vector machine.

The aging prediction model may be generated using one or more genes selected from the group consisting of GAS2L3, AMOT, and WDR76, which are selected genes, through a machine learning model, for example, a support vector machine.

Another aspect of the present specification is to provide the generated cellular aging prediction model.

In another embodiment, the method may further include selecting and verifying genes at each step. Verification of the selected genes may be performed through GO (Gene ontology) analysis, principal component analysis, or clustering.

Another aspect is to provide a computer-readable recording medium on which a program for executing the method on a computer is recorded.

Another aspect is to provide a device for detecting markers of cellular aging.

The device includes an acquisition unit that acquires a plurality of different experimental data including RNA sequencing data of a group of senescent cells and control cells having a plurality of different causes of aging;

A batch-corrected data acquisition unit that quantifies the transcriptome of the obtained plurality of different experimental data through an analysis pipeline to reduce the batch effect derived from different experiments. ;

A first selection unit that selects differentially expressed genes compared to control cells for each group of senescent cells with a plurality of different causes of aging in a training set among the batch-corrected data;

a second selection unit that selects genes commonly present in a group of senescent cells with a plurality of different causes of aging among the selected genes; and

It may include a third selection unit that selects markers of cellular aging for the selected genes using a regression model capable of supervised learning.

Another aspect is to provide a composition for detecting cellular aging.

The composition expresses or activates one or more genes or proteins selected from the group consisting of RRM2B, DUSP6, GSAP, AKAP6, C2orf92, AC008012.1, CLDN11, HMGN2, CHAF1B, KLHL13, MXD3, AUTS2, NAV2, NCAPD2, and PXMP2 It may include an agent whose level can be measured.

The composition expresses one or more genes or proteins selected from the group consisting of EZH2, TMPO, AMOT, GAS2L3, SYNE2, HMGB2, CDCA7L, WDR76, H2BC8, MMP15, ACSL5, CD36, PLD1, CYP26B1, GSAP, H2BC12, and H2AC6. Alternatively, it may include an agent whose activity level can be measured.

The composition may contain an agent capable of measuring the expression or activity level of one or more genes or proteins selected from the group consisting of GAS2L3, AMOT, and WDR76.

In one embodiment, measuring the expression level of the gene is a process of confirming the presence and expression level of mRNA of the genes in a biological sample and may include measuring the amount of mRNA. Analysis methods for this include reverse transcription polymerase reaction (RT-PCR), competitive reverse transcription polymerase reaction (Competitive RT-PCR), real-time reverse transcription polymerase reaction (Real-time RT-PCR), and RNase protection assay (RPA). assay), Northern blotting, DNA chip, etc.

Agents capable of measuring the expression level of the gene may include primers, probes, or antisense oligonucleotides. Since the nucleic acid information of the genes is known in GeneBank, etc., a person skilled in the art can design primers or probes that specifically amplify specific regions of these genes based on the sequences.

The term “primer” used herein includes all combinations of primer pairs consisting of forward and reverse primers that recognize the target gene sequence, and more specifically, a primer pair that provides analysis results with specificity and sensitivity.

The term "probe" used herein refers to a substance that can specifically bind to a target substance to be detected in a sample, and refers to a substance that can specifically confirm the presence of the target substance in the sample through said binding. do. The type of probe molecule is not limited as it is a material commonly used in the art, but preferably may be PNA (peptide nucleic acid), LNA (locked nucleic acid), peptide, polypeptide, protein, RNA or DNA. More specifically, the probe is a biomaterial that is derived from or similar to living organisms or includes those produced in vitro, such as enzymes, proteins, antibodies, microorganisms, animal and plant cells and organs, nerve cells, DNA, and RNA. DNA may include cDNA, genomic DNA, and oligonucleotides, RNA may include genomic RNA, mRNA, and oligonucleotides, and examples of proteins may include antibodies, antigens, enzymes, peptides, etc.

As used herein, the term "antisense oligonucleotide" is DNA or RNA or a derivative thereof containing a nucleic acid sequence complementary to the sequence of a specific gene (e.g., mRNA of a specific gene), and binds to the complementary sequence in the mRNA. It acts to inhibit the translation of mRNA into protein. Antisense oligonucleotide sequence refers to a DNA or RNA sequence that is complementary to the mRNA of the genes and is capable of binding to the mRNA. This may inhibit translation, translocation into the cytoplasm, maturation, or any other essential activity for overall biological functions of the gene mRNA. The length of the antisense oligonucleotide may be 6 to 100 bases, preferably 8 to 60 bases, and more preferably 10 to 40 bases.

In one embodiment, measuring the expression or activity level of the protein may include the process of confirming the presence and expression (or activity) level of the protein in a biological sample. Analysis methods for this include protein chip analysis, immunoassay, ligand binding assay, MALDI-TOF (Matrix Desorption/Ionization Time of Flight Mass Spectrometry) analysis, and SELDI-TOF (Sulface Enhanced Laser Desorption/Ionization Time of Flight Mass Spectrometry). Analysis, radioimmunoassay, radioimmunodiffusion method, Ouchteroni immunodiffusion method, rocket immunoelectrophoresis, tissue immunostaining, complement fixation assay, two-dimensional electrophoresis analysis, liquid chromatography-Mass Spectrometry (LC) -MS), LC-MS/MS (liquid chromatography-Mass Spectrometry/ Mass Spectrometry), Western blot, and ELISA (enzyme linked immunosorbent assay).

Agents capable of measuring the activity level of the protein may include antibodies, aptamers, avidity multimers, or peptidomimerics.

As used herein, the term “antibody” may refer to a specific protein molecule directed against an antigenic site. For the purposes of the present invention, an antibody refers to an antibody that specifically binds to the estrogen receptor, MUCIN5AC, or aromatase protein, and includes polyclonal antibodies, monoclonal antibodies, and recombinant antibodies. Antibodies can be easily produced using techniques well known in the art. Antibodies herein also include intact forms with two full-length light chains and two full-length heavy chains, as well as functional fragments of the antibody molecule. Functional fragments of antibody molecules refer to fragments that possess at least an antigen-binding function, and include Fab, F(ab'), F(ab') 2, and Fv.

Another aspect is the expression of any one or more genes or proteins selected from the group consisting of RRM2B, DUSP6, GSAP, AKAP6, C2orf92, AC008012.1, CLDN11, HMGN2, CHAF1B, KLHL13, MXD3, AUTS2, NAV2, NCAPD2, and PXMP2 Alternatively, a kit for detecting cellular senescence comprising an agent capable of measuring the activity level is provided.

Another aspect is one or more genes or proteins selected from the group consisting of EZH2, TMPO, AMOT, GAS2L3, SYNE2, HMGB2, CDCA7L, WDR76, H2BC8, MMP15, ACSL5, CD36, PLD1, CYP26B1, GSAP, H2BC12, and H2AC6. To provide a kit for detecting cellular senescence that includes an agent capable of measuring the expression or activity level of.

Another aspect is to provide a kit for detecting cellular senescence, which includes an agent capable of measuring the expression or activity level of any one or more genes or proteins selected from the group consisting of GAS2L3, AMOT, and WDR76.

In one embodiment, the kit is a reverse transcription polymerase chain reaction (RT-PCR) kit, a DNA chip kit, a microarray kit, an enzyme-linked immunosorbent assay (ELISA) kit, a protein chip kit, a rapid kit, or an MRM. It may be a (multiple reaction monitoring) kit.

In addition, the kit may further include one or more other component compositions, solutions, or devices suitable for the analysis method. For example, the kit may be a kit containing essential elements required to perform a reverse transcription polymerase reaction. The reverse transcription polymerase reaction kit includes each primer pair specific for the marker gene. The primer is a nucleotide having a sequence specific to the nucleic acid sequence of each gene, and is about 7 bp to 50 bp in length, more preferably about 10 bp to 30 bp in length. It may also include primers specific to the nucleic acid sequence of the control gene. In addition, the reverse transcription polymerase reaction kit consists of test tubes or other suitable containers, reaction buffer (pH and magnesium concentration vary), deoxynucleotides (dNTPs), enzymes such as Taq-polymerase and reverse transcriptase, DNAse, and RNAse inhibitor DEPC- It may include water (DEPC-water), sterilized water, etc. Additionally, for example, a DNA chip kit may include a substrate to which a cDNA or oligonucleotide corresponding to a gene or a fragment thereof is attached, and reagents, agents, enzymes, etc. for producing a fluorescent probe. The substrate may also include cDNA or oligonucleotides corresponding to control genes or fragments thereof. Additionally, for example, the kit may be a diagnostic kit containing essential elements necessary to perform LISA. ELISA kits contain specific antibodies against these proteins. Antibodies have high specificity and affinity for each marker protein and almost no cross-reactivity to other proteins, and are either monoclonal antibodies, polyclonal antibodies, or recombinant antibodies. Additionally, ELISA kits may include antibodies specific for control proteins. Other ELISA kits include reagents capable of detecting bound antibodies, such as labeled secondary antibodies, chromophores, enzymes (e.g. conjugated with antibodies) and their substrates or those capable of binding to antibodies. It may contain other substances, etc. Additionally, for example, the kit may be a rapid kit containing the essential elements needed to perform a rapid test that can provide analysis results. Rapid kits contain specific antibodies against proteins. Antibodies have high specificity and affinity for each marker protein and almost no cross-reactivity to other proteins, and are either monoclonal antibodies, polyclonal antibodies, or recombinant antibodies. Additionally, the rapid kit may include antibodies specific for control proteins. In addition, the rapid kit includes reagents that can detect bound antibodies, such as nitrocellulose membranes with specific antibodies and secondary antibodies immobilized on them, membranes bound to antibody-bound beads, absorption pads, and sample pads. It may include substances, etc. Additionally, for example, the kit may be a multiple reaction monitoring (MRM) kit in MS/MS mode that includes the essential elements necessary to perform mass spectrometry. While SIM (Selected Ion Monitoring) is a method that uses ions created by a single collision in the source part of a mass spectrometer, MRM selects a specific ion from among the broken ions once more and uses another continuously connected MS source. This is a method of passing it through one more time, colliding with it, and then using the ions obtained from these. Through the above MRM (multiple reaction monitoring) analysis methods, the protein expression level of the normal control group can be compared with the protein expression level of the senescent cell group, and also, by determining whether there is a significant increase or decrease in the expression level of the marker gene to protein, the senescent cell group You can judge whether or not.

Another aspect is one or more genes selected from the group consisting of RRM2B, DUSP6, GSAP, AKAP6, C2orf92, AC008012.1, CLDN11, HMGN2, CHAF1B, KLHL13, MXD3, AUTS2, NAV2, NCAPD2, and PXMP2 from an isolated biological sample. Alternatively, a method for detecting cellular senescence comprising measuring the expression or activity level of a protein is provided.

In one embodiment, the method determines the expression or activity level of one or more genes or proteins selected from the group consisting of RRM2B, DUSP6, GSAP, AKAP6, C2orf92, and AC008012.1 compared to the expression of the genes or proteins in a normal control sample. or if it is higher than the activity level, determining cell senescence; or the expression or activity level of any one or more genes or proteins selected from the group consisting of LDN11, HMGN2, CHAF1B, KLHL13, MXD3, AUTS2, NAV2, NCAPD2, and PXMP2 is lower than the expression or activity level of the genes or proteins in the normal control sample. In this case, it may include a step of determining cell aging.

Another aspect is one selected from the group consisting of EZH2, TMPO, AMOT, GAS2L3, SYNE2, HMGB2, CDCA7L, WDR76, H2BC8, MMP15, ACSL5, CD36, PLD1, CYP26B1, GSAP, H2BC12, and H2AC6 from an isolated biological sample. To provide a method for detecting cellular aging comprising measuring the expression or activity level of the above genes or proteins.

In one embodiment, the method determines the expression or activity level of one or more genes or proteins selected from the group consisting of H2BC8, MMP15, ACSL5, CD36, PLD1, CYP26B1, GSAP, H2BC12, and H2AC6 compared to the genes in the normal control sample. Or, if the expression or activity level of the protein is higher than that, determining cell senescence; or when the expression or activity level of any one or more genes or proteins selected from the group consisting of EZH2, TMPO, AMOT, GAS2L3, SYNE2, HMGB2, CDCA7L, and WDR76 is lower than the expression or activity level of the genes or proteins in the normal control sample, It may include a step of determining cellular aging.

Another aspect is to provide a method for detecting cellular senescence, comprising measuring the expression or activity level of one or more genes or proteins selected from the group consisting of GAS2L3, AMOT, and WDR76 from an isolated biological sample.

In one embodiment, the method is used to cause cellular senescence when the expression or activity level of any one or more genes or proteins selected from the group consisting of GAS2L3, AMOT, and WDR76 is lower than the expression or activity level of the genes or proteins of the normal control sample. It may include a judgment step.

In one embodiment, the method for measuring the expression level of the gene may include reverse transcription polymerase reaction, competitive reverse transcription polymerase reaction, real-time reverse transcription polymerase reaction, RNase protection assay, Northern blotting, or DNA chip.

In one embodiment, the method for measuring the activity level of the protein includes Western blotting, ELISA, radioimmunoassay, radioimmunodiffusion, Ouchterlony immunodiffusion, rocket immunoelectrophoresis, and immunohistochemical staining. , immunoprecipitation assay, complement fixation assay, FACS, or protein chip.

Another aspect is a test agent and any one or more genes or proteins selected from the group consisting of RRM2B, DUSP6, GSAP, AKAP6, C2orf92, AC008012.1, CLDN11, HMGN2, CHAF1B, KLHL13, MXD3, AUTS2, NAV2, NCAPD2, and PXMP2 contacting; and

The present invention provides a method for screening a senolytic drug, which includes selecting a test substance that changes the expression or activity level of the gene or protein as a senolytic drug compared to an untreated control group.

In one embodiment, the method uses a test substance in which the expression or activity level of one or more genes or proteins selected from the group consisting of RRM2B, DUSP6, GSAP, AKAP6, C2orf92, and AC008012.1 is reduced compared to the untreated control group. Screening with a senolytic drug; Or, the test substance that increases the expression or activity level of one or more genes or proteins selected from the group consisting of LDN11, HMGN2, CHAF1B, KLHL13, MXD3, AUTS2, NAV2, NCAPD2, and PXMP2 compared to the untreated control group is a senolytic drug. It may include a selection step.

Another aspect is one or more genes or proteins selected from the group consisting of EZH2, TMPO, AMOT, GAS2L3, SYNE2, HMGB2, CDCA7L, WDR76, H2BC8, MMP15, ACSL5, CD36, PLD1, CYP26B1, GSAP, H2BC12, and H2AC6 contacting the test substance; and

The present invention provides a method for screening a senolytic drug, which includes selecting a test substance that changes the expression or activity level of the gene or protein as a senolytic drug compared to an untreated control group.

In one embodiment, the method is performed in a subject in which the expression or activity level of one or more genes or proteins selected from the group consisting of EZH2, TMPO, AMOT, GAS2L3, SYNE2, HMGB2, CDCA7L, and WDR76 is reduced compared to the untreated control group. Screening a substance as a senolytic drug; Or, the test substance that increases the expression or activity level of any one or more genes or proteins selected from the group consisting of H2BC8, MMP15, ACSL5, CD36, PLD1, CYP26B1, GSAP, H2BC12, and H2AC6 compared to the untreated control group is a senolytic drug. It may include a selection step.

Another aspect includes contacting a test substance with one or more genes or proteins selected from the group consisting of GAS2L3, AMOT, and WDR76; and

The present invention provides a method for screening a senolytic drug, which includes selecting a test substance that changes the expression or activity level of the gene or protein as a senolytic drug compared to an untreated control group.

In one embodiment, the method includes the step of selecting a test substance with an increased expression or activity level of one or more genes or proteins selected from the group consisting of GAS2L3, AMOT, and WDR76 as a senolytic drug compared to the untreated control group. It may include

The test substance, for example, a drug candidate, a test compound, or a test composition may include a small molecule compound, an antibody, an antisense nucleotide, a short interfering RNA, a short hairpin RNA, a nucleic acid, a protein, a peptide, Other extracts or natural products may be included.

The contact may be performed in vitro. For example, transforming a cell with a vector expressing or knocking out the genes or proteins; It may include treating the transformed cells with a drug candidate.

The senolytic drug may refer to an agent that can be used to selectively (preferentially or to a greater extent) kill or destroy senescent cells in a clinically significant or biologically significant manner. The senolytic agent may be used in senescent cells (e.g., senescent adipocytes, senescent endothelial cells, senescent fibroblasts, senescent neurons, senescent epithelial cells, senescent mesenchymal cells, senescent smooth muscle cells, senescent macrophages, or senescent chondrocytes). Agents capable of selectively killing more than one type may be included.

According to a method of selecting markers of cellular aging using machine learning according to one aspect, a statistically more significant gene is identified by securing a sufficient number of samples by building an analysis pipeline that performs a meta-analysis on senescent cell RNA-seq data. It has the effect of being able to select candidates for senescent cell characteristic markers among various variables through machine learning methodology.

The marker of cellular senescence derived by the above method is a more meaningful genetic signature that can specify cellular senescence in various types of cells than existing studies, and can be usefully used to specify or detect senescent cells. , it has an effect that can be useful in selecting senolytic agents.

Figure 1 is a flowchart showing a method for detecting markers of cellular aging according to one embodiment.
Figure 2 is a schematic diagram for selecting biomarkers of cellular aging according to an embodiment.
Figure 3 is a diagram showing the results of PCA analysis for about 14,000 genes remaining after removing genes with low expression levels for the entire sample.
Figure 4 is a diagram showing the results of clustering analysis for about 14,000 genes remaining after removing genes with low expression levels for the entire sample.
Figure 5 is a Venn diagram showing the number of genes commonly present across all derivatives among differentially expressed genes for each inducer type; a: Gene with increased expression, b: Gene with decreased expression.
Figure 6 is a diagram showing the GO analysis results of selected commonly expressed genes; a: Gene with increased expression, b: Gene with decreased expression.
Figure 7 is a diagram showing the results of PCA analysis for 363 selected commonly expressed genes.
Figure 8 is a diagram showing the results of clustering analysis for 363 selected commonly expressed genes.
Figure 9 is a barplot showing the coefficient values of each of the 15 genes selected by LASSO.
Figure 10 is a diagram showing the results of PCA analysis for 15 genes selected by LASSO.
Figure 11 is a diagram showing the results of clustering analysis for 15 genes selected by LASSO.
Figure 12 is a diagram showing the expression boxplot of 15 genes selected by LASSO in the test set data (n = 84); a: Distribution within the test set of 6 genes with positive coefficients, b: Distribution within the test set of 9 genes with negative coefficients.
Figure 13 is a graph showing ROC curve results for performance evaluation of two cell aging classification models; a: Assay data of one embodiment, b: Additional various senescent cell data.
Figure 14 is a diagram showing the results of PCA analysis for about 14,000 genes remaining after removing genes with low expression levels for all reselected samples (n=147).
Figure 15 is a Venn diagram showing the number of genes commonly present across all derivatives among differentially expressed genes for each inducer type; a: Gene with increased expression, b: Gene with decreased expression.
Figure 16 is a diagram showing the GO analysis results of selected commonly expressed genes; a: Gene with increased expression, b: Gene with decreased expression.
Figure 17 is a barplot showing the coefficient values of each of the 17 genes selected by LASSO.
Figure 18 is a diagram showing the results of PCA analysis for 17 genes selected by LASSO.
Figure 19 is a diagram showing the results of clustering analysis for 17 genes selected by LASSO.
Figure 20 is a diagram showing the expression boxplot of 17 genes selected by LASSO in the test set data (n = 68); a: Distribution within the test set of 8 genes with positive coefficients, b: Distribution within the test set of 9 genes with negative coefficients.
Figure 21 is a graph showing ROC curve results for performance evaluation of five cell aging classification models; a: Assay data of one embodiment, b: Additional various senescent cell data.
Figure 22 shows the results of subculturing HUVEC cells and aging them, and then confirming aging through SA β-Gal staining.
Figure 23 shows the results of confirming the decrease in expression levels of GAS2L3, AMOT, and WDR76 in senescence-induced cells through RT-PCR and electrophoresis.
Figure 24 shows the results of quantitatively confirming the decrease in expression levels of GAS2L3, AMOT, and WDR76 in senescence-induced cells through qPCR.

Hereinafter, the present invention will be described in more detail through examples. However, these examples are for illustrative purposes only and the scope of the present invention is not limited to these examples.

Example 1. Identification of biomarkers of cellular aging

Figure 2 shows a schematic diagram for selecting biomarkers of cellular aging according to one embodiment.

Specifically, it is as follows.

1.1 Data set preparation and analysis

To identify markers related to senescent cells, RNA-seq data was used from the ENA (European Nucleotide Archive) public database (http://www.ebi.ac.uk/ena/browser/home).

Data from 13 different experiments containing RNA-seq expression data sets of senescent cells (GSE63577, GSE72407, GSE113060, GSE70668, GSE132370, GSE130727, GSE98440, GSE109700, GSE130306, GSE94395, GSE168994, , GSE75643) were selected, and various Fibroblast (BJ, HFF, IMR-90, WI-38, LF1, TIG3) data generated by inducer type and the corresponding control data were downloaded in Fastq format. The data consisted of a total of 165 samples, 73 in the control group and 92 in the experimental group, and the characteristics of each data were organized by derivatives in which senescent cells were induced (Replicative senescence, Oncogene induced senescence (OIS), Therapy induced senescence (TIS)). In addition, in order to verify the ultimately selected genes through later analysis, a training set and a test set were assigned to each GSE. By referring to data-related papers, training data and test data for model testing were allocated in a ratio of approximately 5:5. Each data consists of Fastq data generated using the Illumina Hiseq platform, and data characteristic information is shown in Table 1 below.

sample information GSE number number of samples
(Control group: senescent cell group) Types of cellular aging
(Inducer types) origin of cells category
(Tr(training set)/Te(testing set) GSE63577 23 (12:11) Replicative BJ, HFF, IMR90, WI38 Tr GSE72407 22 (8:14) OIS, TIS IMR90 Tr GSE113060 10 (5　: 5) OIS IMR90 Tr GSE70668 6 (3:3) OIS IMR90 Tr GSE132370 6 (3:3) TIS IMR90 Tr GSE130727 26 (12:14) Replicative, TIS, OIS IMR90, WI38 Tr/Te GSE98440 8 (4:4) Replicative IMR90 Te GSE109700 9 (3:6) Replicative LF1 Te GSE130306 6 (3:3) Replicative WI38 Te GSE94395 6 (3:3) TIS IMR90 Te GSE168994 4 (2:2) TIS IMR90 Te GSE72404 24 (6:18) OIS IMR90 Te GSE75643 15 (9:6) OIS TIG3 Te Total 165 (73:92) - -

In addition, the division information of the training set and test set is shown in Table 2 below.

Sample Category control group R.S. OIS TIS Total training set 37 13 17 14 81 black three 36 15 24 9 84 Total 73 28 41 23 165

The ratio of the training set and the test set was split 50:50, and the specific derivative data in the training set and validation set was kept in an approximately constant ratio to prevent biased distribution.

1.2 Data preprocessing and meta-analysis

The Fastq files of Example 1.1 were quantified through the same pipeline, and the SALMON pipeline was created and analyzed to minimize batch effects derived from different experiments. As shown in Figure 2, the SALMON pipeline includes QC (fastQC, v0.11.9), Trimming (Trimmomatics, v0.39), Mapping (SALMON, v1.10), Raw count quantification (tximeta, v1.4.5), and Batch. It consists of the correction (limma, v3.42.2) step (Tximeta and limma were analyzed within the R program v3.6.3).

1.3 Exploratory data analysis

For the expression data for each GSE, the batch effect was confirmed using the removeBatchEffect function of limma (v3.42.2) after TMM normalization & Log ₂ CPM (count per million) normalization. Afterwards, expression data were analyzed using batch-corrected, TMM normalized log ₂ CPM values, genes with low expression levels across the samples were removed using the edgeR package, and only about 14,000 genes were used for analysis. Principal component analysis (PCA) analysis was performed on the approximately 14,000 genes, and the PCA and clustering results are shown in Figures 3 and 4, respectively, through Spearman correlation.

Figure 3 is a diagram showing the results of PCA analysis for about 14,000 genes remaining after removing genes with low expression levels for the entire sample.

Figure 4 is a diagram showing the results of clustering analysis for about 14,000 genes remaining after removing genes with low expression levels for the entire sample.

1.4 Differentially expressed genes analysis (DEG analysis)

For each sample group classified by derivative among the training data, differential expression gene analysis was performed to select genes whose expression level increased or decreased in the senescent cell group compared to the normal cell group. The analysis was conducted using the R package edgeR (v3.10), and after TMM normalization, the analysis design was constructed according to the aging state information and batch information of the sample. To solve the multiple comparison problem that occurs when testing multiple hypotheses, adjusted p-values were calculated. Among genes whose expression levels were changed, the adjusted p-values satisfied the standard of being less than 0.05 and at the same time, the expression values Only genes with a fold change of more than 2-fold (up-regulated) or 1/2-fold or less (down-regulated) were selected. Looking at each inducer type, a total of 1,444 genes (up-regulated genes: 877, down-regulated genes: 567) in RS and a total of 1,647 genes (up-regulated genes: 602, down-regulated genes) in OIS. genes: 1,045), a total of 1,405 genes (up-regulated genes: 802, down-regulated genes: 603) in TIS were selected as differentially expressed genes, and these are shown in Table 3 below.

Differential expression gene analysis information Inducer type Comparison
(Control: Sensitive) Up-regulated genes Down-regulated genes Total DEGs R.S. 13:13 877 567 1,444 OIS 16:17 602 1,045 1,647 TIS 14:14 802 603 1,405

In order to select the characteristic genes of cellular aging that are expressed regardless of the type of inducer among differentially expressed genes (DEGs) by inducer type, the DEG results for each inducer were divided into up-regulated genes and down-regulated genes, and a Venn diagram was drawn. The intersection of genes was selected by indicating . The Venn diagram is shown in Figure 5. Figure 5 is a Venn diagram showing the number of genes commonly present across all derivatives among differentially expressed genes for each inducer type; a: Gene with increased expression, b: Gene with decreased expression.

As shown in Figure 5, a total of 363 genes, including 85 up-regulated genes and 278 down-regulated genes, were selected, indicating that these genes are expressed in senescent cells regardless of the type of derivative.

Afterwards, to confirm that the selected commonly expressed genes are a set of genes reflecting cell aging characteristics, GO (gene ontology) analysis was performed using gProfiler (https://biit.cs.ut.ee/gprofiler/). By conducting this study, biological pathways related to 363 genes were explored. GO analysis was performed for each gene according to the direction of increase or decrease in expression, and only GO results with an adjusted p-value of 0.05 or less were selected, and the results are shown in Figure 6.

Figure 6 is a diagram showing the GO analysis results of selected commonly expressed genes; a: Gene with increased expression, b: Gene with decreased expression.

As shown in Figure 6, the 85 genes whose expression was commonly increased in senescent cells were known characteristics of cellular senescence, such as DNA-damage response, SASP (Senescence associated secretory phenotype), Inhibition of DNA recombination at telomere, Immune response, and various cellular genes. It was confirmed that characteristics related to the senescence pathway appeared significantly. In addition, the 278 genes whose expression is commonly reduced in aging cells are also related to biological pathways related to the cell cycle, DNA replication, DNA repair, telomere maintenance, chromosome organization, and segregation, which correspond to the already known characteristics of cellular aging. was confirmed. Through this, it was confirmed that a total of 363 intersection gene sets well reflect the characteristics of cellular aging.

Next, PCA analysis and clustering analysis were performed in the same manner as in 1.3 using the 363 commonly expressed genes selected above, and the results are shown in Figures 7 and 8, respectively.

Figure 7 is a diagram showing the results of PCA analysis for 363 selected commonly expressed genes.

Figure 8 is a diagram showing the results of clustering analysis for 363 selected commonly expressed genes.

As shown in Figure 7, the PCA results of 363 selected commonly expressed genes explain the classification of data by PC1 better than the PCA results of about 14,000 genes in Figure 3, but clearly distinguish between senescent cells and control data. I found out that I couldn't do it.

As shown in Figure 8, the clustering results of 363 selected commonly expressed genes are compared with the clustering results of about 14,000 genes in Figure 4. Although the expression patterns of the genes are differentiated between senescent cells and the control group, the data for senescent cells and the control group are clearly distinct. It was found that it was not possible to distinguish clearly.

1.5 Variable selection using LASSO (least absolute shrinkage and selection operator)

To select the minimum number of genes that best represent the characteristics of cellular aging among the 363 selected genes, a variable selection process for genes was performed through LASSO regression.

Specifically, LASSO regression was performed through R package glmnet 4.0-2, and TMM normalized Log ₂ CPM (count per million) values of a total of 81 training data sets were used to control (=0) or control depending on the state of the sample. This is the result selected by assigning a Senescent (=1) value and then adjusting the parameters for the binary classification model. After setting the optimal lambda (λ value 0.006) through leave-one-out-cross-validation, the model was fit and finally 15 genes were selected. The genes in question are ENSG00000048392 (RRM2B), ENSG00000139318 (DUSP6), ENSG00000186088 (GSAP), ENSG00000151320 (AKAP6), ENSG00000228486 (C2orf92), and ENSG00000285901 (AC0). 08012.1), ENSG00000013297 (CLDN11), ENSG00000198830 (HMGN2), ENSG00000159259 (CHAF1B), ENSG00000003096(KLHL13), ENSG00000213347(MXD3), ENSG00000158321(AUTS2), ENSG00000166833(NAV2), ENSG00000010292(NCAPD2), ENSG00000176894(PXMP2) And the direction of expression of the corresponding gene in cellular aging is determined through the coefficient of each gene ( +/-) and influence. A total of 6 genes (RRM2B, DUSP6, GSAP, AKAP6, C2orf92, AC008012.1) have positive coefficients, indicating increased expression in senescent cells, and the remaining 9 genes (CLDN11, HMGN2, CHAF1B, KLHL13, MXD3, AUTS2) , NAV2, NCAPD2, and PXMP2) have negative coefficients, indicating reduced expression in senescent cells.

The results of the variable selection process for genes through the LASSO regression are shown in Table 4 and Figure 9 below.

Cellular aging characteristic genes selected through LASSO regression Ensembl gene id SYMBOL Coefficient Description ENSG00000048392 RRM2B 1.85 Ribonucleotide Reductase Regulatory TP53 Inducible Subunit M2B ENSG00000139318 DUSP6 0.45 Dual Specificity Phosphatase 6 ENSG00000186088 GSAP 0.45 Gamma-Secretase Activating Protein ENSG00000151320 AKAP6 0.28 A-Kinase Anchoring Protein 6 ENSG00000228486 C2orf92 0.16 Chromosome 2 Open Reading Frame 92 ENSG00000285901 AC008012.1 0.13 Novel Transcript ENSG00000013297 CLDN11 -0.02 Claudine 11 ENSG00000198830 HMGN2 -0.03 High Mobility Group Nucleosomal Binding Domain 2 ENSG00000159259 CHAF1B -0.14 Chromatin Assembly Factor 1 Subunit B ENSG00000003096 KLHL13 -0.26 Kelch Like Family Member 13 ENSG00000213347 MXD3 -0.32 MAX Dimerization Protein 3 ENSG00000158321 AUTS2 -0.41 Activator Of Transcription And Developmental Regulator AUTS2 ENSG00000166833 NAV2 -0.51 Neuron Navigator 2 ENSG00000010292 NCAPD2 -0.94 Non-SMC Condensin I Complex Subunit D2 ENSG00000176894 PXMP2 -1.19 Peroxisomal Membrane Protein 2

Figure 9 is a barplot showing the coefficient values of each of the 15 genes selected by LASSO.

Afterwards, PCA and clustering analysis were performed in the same manner as in 1.3 above using 15 genes selected by LASSO for the 81 training set data, and the results are shown in Figures 10 and 11, respectively.

Figure 10 is a diagram showing the results of PCA analysis for 15 genes selected by LASSO.

Figure 11 is a diagram showing the results of clustering analysis for 15 genes selected by LASSO.

As shown in Figure 10, the PCA results of 15 genes selected by LASSO are compared with the PCA results of Figures 3 and 7, and it can be seen that senescent cells and control data are clearly distinguished by PC1, and the It was found that explanations for data classification also increased.

As shown in Figure 11, the clustering results of 15 genes selected by LASSO show that the expression patterns of genes are clearly distinguished between senescent cells and the control group compared to the clustering results in Figures 4 and 8, and the senescent cell and control data was found to be clearly distinguished.

The above results mean that the 15 genes selected by LASSO well reflect the characteristics of senescent cells within the training set data.

Example 2. Assay of selected senescent cell marker candidate genes

2.1 Confirmation of gene expression differences between the control group and the cell aging group in the test data

It was confirmed whether the expression patterns of the 15 senescent cell marker candidate genes selected in Example 1 showed significant expression differences even within 84 test data that were not involved in the selection process of the 15 genes.

Specifically, for the 84 test data, the batch effect was removed through TMM normalization and the limma removeBatchEffect function in the same process as the training data, and then _the control group (n = 36) and the senescent cell group (n = 48), it was confirmed whether there was a significant difference in the expression of 15 genes, and the results are shown in Figure 12.

Figure 12 is a diagram showing the expression boxplot of 15 genes selected by LASSO in the assay set data (n=84); a: Distribution within the test set of 6 genes with positive coefficients, b: Distribution within the test set of 9 genes with negative coefficients.

2.2 Creation of cellular aging prediction model

The effectiveness of the selected genes was verified by applying 15 selected genes to the training data set to create two models that predict the state of senescent cells.

Specifically, the first model was produced using the RBF kernel of a support vector machine, and R package e1071 (version 1.7-3) was used. The data used to learn the model were the expression values of 15 genes for the 81 training data used for learning, and the optimal classification model was selected and produced through LOOCV (leave one out cross-validation). For the second model, an SVM model was created in the same manner as the first model by applying the expression values of 55 cellular aging genes selected from other cell aging-related research literature (Hernandez-Segura et al. 2017) to the training data set. The second model was created to compare the senescent cell classification performance with the first model.

2.3 Cell aging prediction model performance evaluation

A classification test was performed on 84 senescent fibroblast test data (36 control, 48 senescent cells) using the two models created in 2.2 above, and the results are shown in Figure 13A. Additionally, using the two models created in 2.2 above, data on 42 different senescent cells (control: 18, senescent cells: 24, 42 different types other than fibroblasts (HUVEC, HAEC, MSC, LS8817, Ovcar3) ) was subjected to a classification test, and the results are shown in Figure 13B. The Log ₂ CPM values of the test data were tested after scaling based on the mean and standard deviation of the expression values of the training data.

Figure 13 is a graph showing ROC curve results for performance evaluation of two cell aging classification models; a: Assay data of one embodiment, b: Additional various senescent cell data.

As shown in Figure 13, it was confirmed that the 15 gene model selected by LASSO showed better performance in both results.

a: LASSO 15 gene model AUC: 98.9% > Hernandez et al, 55 genes model AUC: 97.6%

b: LASSO 15 genes model AUC: 86.6% > Hernandez et al, 55 genes model AUC: 75.9%

Therefore, based on the results of Example 2, it was confirmed that the expression pattern of the 15 genes selected according to one embodiment showed a significant difference between the normal cell and senescent cell groups, identical to the coefficient sign pattern of LASSO. In addition, it was confirmed that the senescent cell classification model constructed with 15 genes better represents the characteristics of cellular aging than the senescent cell classification model constructed with 55 cellular aging-related genes previously known in research. In addition, a previous study presented characteristic genes considering various cells, including senescent fibroblasts (Hernandez-Segura et al. 2017), and although the method according to one embodiment selected genes using only senescent fibroblasts as training data, Nevertheless, it was confirmed that better classification performance was achieved for various types of cells than in previous studies.

The above results mean that the method according to one embodiment can select statistically more significant markers in various types of cells, including fibroblasts, and the 15 genes derived by the method according to one embodiment are senescent cells. This means that it can be usefully used to specify or detect.

Example 3. Additional Biomarkers of Cellular Aging

The same method as Example 1 (FIG. 2) was used, but after removing 18 samples with different aging characteristics among the samples specified in Table 1 in 1.1 Data Set Preparation and Analysis, 17 genes whose expression changes in senescent cells were analyzed. Through additional screening, more sophisticated genes were obtained.

3.1 Data set preparation and analysis

Of the 165 fibroblast samples in Example 1, 18 samples were additionally removed to obtain a total of 147 fibroblast samples. Information on the removed samples is as follows.

In GSE72404, 12 notch-induced senescence samples out of 24 previously used samples were removed. This is because it is known that notch-induced senescence shows different characteristics from primary senescence (RS, OIS and TIS). In GSE1307272, 4 control groups and 2 OIS samples were removed from the existing 26 samples. These samples were removed because they could be considered outliers in the PCA results. Data characteristic information is shown in Table 5 below.

Reselected Sample Information GSE number number of samples
(Control group: senescent cell group) Types of cellular aging
(Inducer types) origin of cells category
(Tr(training set)/Te(testing set) GSE63577 23 (12:11) Replicative BJ, HFF, IMR90, WI38 Tr GSE72407 22 (8:14) OIS, TIS IMR90 Tr GSE113060 10 (5　: 5) OIS IMR90 Tr GSE70668 6 (3:3) OIS IMR90 Tr GSE132370 6 (3:3) TIS IMR90 Tr GSE130727 20 (8:12) Replicative, TIS, OIS IMR90, WI38 Tr/Te GSE98440 8 (4:4) Replicative IMR90 Te GSE109700 9 (3:6) Replicative LF1 Te GSE130306 6 (3:3) Replicative WI38 Te GSE94395 6 (3:3) TIS IMR90 Te GSE168994 4 (2:2) TIS IMR90 Te GSE72404 12 (6:6) OIS IMR90 Te GSE75643 15 (9:6) OIS TIG3 Te Total 147 (69:78) - -

In addition, the division information of the training set and test set is shown in Table 6 below.

Partition information for training and testing sets Sample Category cell type control group R.S. OIS TIS Total training set Fibroblasts 37 13 15 14 79 black set 1 Fibroblasts 32 15 12 9 68 black set 2 Other cell types 15 7 - 11 33 Total - 84 35 27 34 180

3.2 Data preprocessing and meta-analysis

Data preprocessing and meta-analysis were performed in the same manner as in Example 1.2 above.

3.3 Exploratory data analysis

The exploratory data analysis was performed in the same manner as in Example 1.3 above. PCA (Principal component analysis) analysis was performed on about 14,000 genes, and the PCA results are shown in Figure 14 through Spearman correlation.

Figure 14 is a diagram showing the results of PCA analysis for about 14,000 genes remaining after removing genes with low expression levels for all reselected samples.

3.4 Differentially expressed genes analysis (DEG analysis)

Differentially expressed genes for each inducer type were selected in the same manner as the differentially expressed gene analysis process in Example 1.4 above. Looking at each inducer type, a total of 1,463 genes (up-regulated genes: 881, down-regulated genes: 582) in RS and a total of 2,039 genes (up-regulated genes: 791, down-regulated genes) in OIS. genes: 1,248), a total of 1,903 genes (up-regulated genes: 1,238, down-regulated genes: 665) in TIS were selected as differentially expressed genes, and these are shown in Table 7 below.

Differential expression gene analysis information Inducer type Comparison
(Control: Sensitive) Up-regulated genes Down-regulated genes Total DEGs R.S. 13:13 881 582 1,463 OIS 14:15 791 1,248 2,039 TIS 14:14 1,238 665 1,903

In order to select the characteristic genes of cellular aging that are expressed regardless of the type of inducer among differentially expressed genes (DEGs) by inducer type, the DEG results for each inducer were divided into up-regulated genes and down-regulated genes, and a Venn diagram was drawn. The intersection of genes was selected by indicating . The Venn diagram is shown in Figure 15.

Figure 15 is a Venn diagram showing the number of genes commonly present across all derivatives among differentially expressed genes for each inducer type; a: Gene with increased expression, b: Gene with decreased expression.

As shown in Figure 15, a total of 465 genes, including 181 up-regulated genes and 284 down-regulated genes, were selected, indicating that these genes are expressed in senescent cells regardless of the type of derivative.

Afterwards, to confirm that the selected commonly expressed genes are a set of genes that reflect cell aging characteristics, GO (gene ontology) analysis was performed using gProfiler (https://biit.cs.ut.ee/gprofiler/). By conducting this study, biological pathways related to 465 genes were explored. GO analysis was performed for each gene according to the direction of increase or decrease in expression, and only GO results with an adjusted p-value of 0.05 or less were selected, and the results are shown in Figure 16.

Figure 16 is a diagram showing the GO analysis results of selected commonly expressed genes; a: Gene with increased expression, b: Gene with decreased expression.

As shown in Figure 16, the 181 genes whose expression was commonly increased in senescent cells significantly exhibited known characteristics of cellular senescence such as SASP (Senescence associated secretory phenotype), Chromatin organization, Immune response, and various Cellular senescence pathway-related characteristics. I was able to confirm. In addition, in the case of 284 genes whose expression was commonly reduced in aging cells, it was confirmed that they were related to biological pathways related to the cell cycle, DNA replication, DNA repair, telomere maintenance, and organization, which correspond to the already known characteristics of cellular aging. . Through this, it was confirmed that a total of 465 intersection gene sets well reflect the characteristics of cellular aging.

3.5 Variable selection using LASSO (least absolute shrinkage and selection operator)

To select the minimal genes that best represent the characteristics of cellular aging among the 465 selected genes, the same method as the variable selection process using LASSO (least absolute shrinkage and selection operator) in Example 1.5 above was used.

Specifically, LASSO regression was performed using the R package glmnet 4.0-2, and TMM normalized Log ₂ CPM (count per million) values of a total of 79 training data sets were used to control (=0) or control depending on the state of the sample. This is the result selected by assigning a Senescent (=1) value and then adjusting the parameters for the binary classification model. After setting the optimal lambda (λ) through leave-one-out-cross-validation, the model was fit and finally 17 genes were selected. The genes in question are ENSG00000106462 (EZH2), ENSG00000120802 (TMPO), ENSG00000126016 (AMOT), ENSG00000139354 (GAS2L3), ENSG00000054654 (SYNE2), ENSG00000164104 (HMGB2) , ENSG00000164649 (CDCA7L), ENSG00000092470 (WDR76), ENSG00000273802 (H2BC8), ENSG00000102996 ( MMP15), ENSG00000197142 (ACSL5), ENSG00000135218 (CD36), ENSG00000075651 (PLD1), ENSG00000003137 (CYP26B1), ENSG00000186088 (GSAP), ENSG00000197903 (H 2BC12), ENSG00000180573 (H2AC6), and cell aging is determined through the coefficient of each gene. Indicates the expression direction (+/-) and influence of the gene in question. A total of 9 genes (H2BC8, MMP15, ACSL5, CD36, PLD1, CYP26B1, GSAP, H2BC12, H2AC6) have positive coefficients, indicating increased expression in senescent cells, and the remaining 8 genes (EZH2, TMPO, AMOT, GAS2L3) , SYNE2, HMGB2, CDCA7L, WDR76) have negative coefficients, so their expression in senescent cells is reduced.

The results of the variable selection process for genes through the LASSO regression are shown in Table 8 and Figure 17.

Cellular aging characteristic genes selected through LASSO regression Ensembl gene id
SYMBOL Coefficient Description ENSG00000106462 EZH2 -0.26 Enhancer Of Zeste 2 Polycomb Repressive Complex 2 Subunit ENSG00000120802 TMPO -0.26 Thymopoietin ENSG00000126016 AMOT -0.17 Angiomotin ENSG00000139354 GAS2L3 -0.09 Growth Arrest Specific 2 Like 3 ENSG00000054654 SYNE2 -0.09 Spectrin Repeat Containing Nuclear Envelope Protein 2 ENSG00000164104 HMGB2 -0.08 High Mobility Group Box 2 ENSG00000164649 CDCA7L -0.05 Cell Division Cycle Associated 7 Like ENSG00000092470 WDR76 -0.001 WD Repeat Domain 76 ENSG00000273802 H2BC8 0.005 H2B Clustered Histone 8 ENSG00000102996 MMP15 0.04 Matrix Metallopeptidase 15 ENSG00000197142 ACSL5 0.06 Acyl-CoA Synthetase Long Chain Family Member 5 ENSG00000135218 CD36 0.09 CD36 Molecule ENSG00000075651 PLD1 0.11 Phospholipase D1 ENSG00000003137 CYP26B1 0.14 Cytochrome P450 Family 26 Subfamily B Member 1 ENSG00000186088 GSAP 0.22 Gamma-Secretase Activating Protein ENSG00000197903 H2BC12 0.24 H2B Clustered Histone 12 ENSG00000180573 H2AC6 0.78 H2A Clustered Histone 6

Figure 17 is a barplot showing the coefficient values of each of the 17 genes selected by LASSO.

Afterwards, PCA analysis was performed in the same manner as in 3.3 above using 17 genes selected by LASSO for the 79 training set data, and the clustering results were confirmed and shown in Figures 18 and 19, respectively.

Figure 18 is a diagram showing the results of PCA analysis for 17 genes selected by LASSO.

Figure 19 is a diagram showing the results of clustering analysis for 17 genes selected by LASSO.

As shown in Figure 18, the PCA results of 17 genes selected by LASSO are compared with the PCA results in Figure 14, and it can be seen that senescent cells and control data are clearly distinguished by PC1, and the data classification of PC1 It was found that the number of explanations also increased.

As shown in Figure 19, the clustering results of 17 genes selected by LASSO showed that the expression patterns of genes were clearly distinguished between senescent cells and the control group, and clearly distinguished the senescent cell and control data.

The above results mean that the 17 genes selected by LASSO well reflect the characteristics of senescent cells within the training set data.

Example 4. Assay of selected senescent cell marker candidate genes

4.1 Confirmation of gene expression differences between the control group and the cell aging group in the test data

It was confirmed whether the expression patterns of the 17 senescent cell marker candidate genes selected in Example 3 showed significant expression differences even within 68 test data that were not involved in the selection process of the 17 genes.

Specifically, for the 68 test data, the batch effect was removed through TMM normalization and the limma removeBatchEffect function in the _same process as the training data, and then the control group (n = 32) and the senescent cell group (n = 36), it was confirmed whether there was a significant difference in the expression of 17 genes, and the results are shown in Figure 20.

Figure 20 is a diagram showing the expression boxplot of 17 genes selected by LASSO in the assay set data (n=68); a: Distribution within the test set of 8 genes with positive coefficients, b: Distribution within the test set of 9 genes with negative coefficients.

4.2 Creation of cellular aging prediction model

The effectiveness of the selected genes was verified by applying 17 selected genes to the training data set to create 5 models that predict the state of senescent cells.

Specifically, the first model was produced using the RBF kernel of a support vector machine, and R package e1071 (version 1.7-3) was used. The data used to learn the model were the expression values of 17 genes for the 79 training data used for learning, and the optimal classification model was selected and produced through LOOCV (leave one out cross-validation). Models 2, 3, 4, and 5 are described in other cell aging-related research literature [Hernandez-Segura et al. 2017, Kiss et al. 2020, Park et al. 2021, Casella et al. 2019], the expression values of various cellular aging genes selected were applied to the training data set to create an SVM model in the same way as the first model. The second, third, fourth, and fifth models were created to compare the senescent cell classification performance with the first model.

4.3 Cell aging prediction model performance evaluation

A classification test was performed on 68 senescent fibroblast test data (32 control, 36 senescent cells) using the five models created in 4.2 above, and the results are shown in Figure 21a. In addition, using the two models created in 4.2 above, data on 33 different senescent cells (control: 15, senescent cells: 18, 33 different types other than fibroblasts (HUVEC, HAEC, MSC, LS8817, Ovcar3) ), a classification test was performed, and the results are shown in Figure 21b. The Log ₂ CPM values of the test data were tested after scaling based on the mean and standard deviation of the expression values of the training data.

Figure 21 is a graph showing ROC curve results for performance evaluation of five cell aging classification models; a: Assay data of one embodiment, b: Additional various senescent cell data.

As shown in Figure 21, it was confirmed that the 17 gene model selected by LASSO showed better performance in both results.

a: LASSO 17 gene model AUC: 100% = Casella et al, model AUC: 100% > Hernandez et al, model AUC: 99.7% > Kiss et al, model AUC: 99.5% > Park et al, model AUC: 98%

b: LASSO 17 gene model AUC: 98.9% > Casella et al, model AUC: 98.1% > Kiss et al, model AUC: 86.3% > Hernandez et al, model AUC: 82.6% > Park et al, model AUC: 70.7% .

Therefore, based on the results of Example 4, it was confirmed that the expression pattern of the 17 genes selected according to one embodiment showed a significant difference between the normal cell and senescent cell groups, identical to the coefficient sign pattern of LASSO. In addition, it was confirmed that the senescent cell classification model constructed with 17 genes better represents the characteristics of cellular aging than the senescent cell classification model constructed with various cellular aging-related genes known in existing research.

Although the method according to one embodiment selected genes using only senescent fibroblasts as training data, it was confirmed that it showed better classification performance than existing studies for various types of cells.

The above results mean that the method according to one embodiment can select statistically more significant markers in various types of cells, including fibroblasts, and the 17 genes derived by the method according to one embodiment are senescent cells. This means that it can be usefully used to specify or detect.

Example 5. Confirmation of expression changes of selected biomarkers in senescence-induced cells

Among the 17 genes selected in Example 3, we examined whether the GAS2L3, AMOT, and WDR76 genes showed changes in actual expression levels in cells that induced senescence.

5.1 Induction and confirmation of cell aging

Human umbilical vein endothelial cells (HUVEC cells, LONZA, HUVEC, C2519A) were used and cultured in EGM2 (LONZA, EGM-2 Endothelial Cell Growth Medium-2 BulletKit, CC-3162) medium. Senescence was induced by varying the number of subcultures of cells. After culturing HUVEC cells at passage 3 (P3) and passage 7 (P7), SA β-Gal (senescence-associated beta-galactosidase, Cell Signaling, Senesence β-Galactosidase Staining Kit) was used to determine the degree of aging of these cells according to passage. , #9860) staining was performed. The staining results are shown in Figure 22.

As shown in Figure 22, P7 cells had a darker and more blue color than P3 cells, indicating that more aging had progressed.

5.2 Confirmation of expression changes in GAS2L3, AMOT and WDR76

RNA from P3 and P7 cells was extracted using Trizol (Invitrogen, TRIzol Reagent, 15596018), and cDNA (complementary DeoxyriboNucleic Acid) was synthesized using a cDNA synthesis kit (LaboPass, cDNA synthesis kit, SMRTK002). Afterwards, only the GAS2L3, AMOT, and WDR76 genes were specifically amplified using the primers and PCR premixture (SolGent, Solg 2X Multiplex PCR Smart mix, SMP01-M25P) shown in Table 9 below, and GAPDH was used to quantify the cDNA of cells. . A PCR device (Aplied Biosystems, MiniAmp Plus Thermal Cycler, A37835) was used in this process.

After performing PCR, the expression levels of GAS2L3, AMOT, and WDR76 were confirmed through electrophoresis, and the results are shown in Figure 23.

Additionally, in order to quantitatively compare expression levels, qPCR was performed using the cDNA obtained above. Specifically, the primers in Table 9 below were used, and for qPCR, a qPCR (quantitative PCR) reagent using SYBR green (Applied Biosystems, SYBR Green PCR Master Mix, 2109533) was used, and real time PCR (Applied Biosystems, StepOne Real) was used. -time PCR System, 4376357) was used. The results of qPCR are shown in Figure 24.

gene Primer direction 5'to 3' sequence number AMOT Forward AAA CGT GAG CAG CTA GAG CAC SEQ ID NO: 1 Reverse TGT GTC CCT CTG AGC AGC CA SEQ ID NO: 2 GAS2L3 Forward ATC TGG GCT CTA TGT CAG TCC G SEQ ID NO: 3 Reverse TTG GCA AAA GTG TGG ACT CGG SEQ ID NO: 4 WDR76 Forward CTG CTG CAA GAC TCC GTG AA SEQ ID NO: 5 Reverse GCC CAG GAG GTA ACA AAG GAG SEQ ID NO: 6 GAPDH Forward ATC CCA TCA CCA TCT TCC SEQ ID NO: 7 Reverse CCA TCA CGC CAC GAT TTC SEQ ID NO: 8

As shown in Figure 23, the bands corresponding to GAS2L3, AMOT, and WDR76 appeared darker in P3 than in P7, confirming that aged P7 cells showed lower expression of GAS2L3, AMOT, and WDR76 than P3 cells.

In addition, as shown in Figure 24, qPCR results also confirmed that GAS2L3, AMOT, and WDR76 gene expression levels were decreased in P7 cells.

From the above results, it was found that the expression of GAS2L3, AMOT, and WDR76 was decreased in actual senescent cells, and that the GAS2L3, AMOT, and WDR76 genes could be used as biomarkers of senescent cells.

Claims (3)

A method for detecting cellular senescence comprising measuring the expression or activity level of genes or proteins of GAS2L3 and WDR76 from isolated endothelial cells. delete The method of claim 1, wherein when the expression or activity level of the gene or protein of GAS2L3 and WDR76 is lower than the expression or activity level of the gene or protein of the normal control sample, detecting cellular senescence comprising determining cell senescence. method.