Caffeine Therapy for Apnea of Prematurity: Role of the Circadian CLOCK Gene Polymorphism

Standard-dose caffeine citrate has been routinely prescribed for apnea of prematurity (AOP) management; however, some preterm infants respond well to the therapy while others do not. The AOP phenotype has been attributed solely to the immature control of the respiratory system consequent to preterm birth, but there are also important genetic influences. Based on our previous report, we tested the hypothesis that the human circadian locomotor output cycles kaput (CLOCK) gene polymorphisms play a role in the response to caffeine citrate therapy in preterm infants. We also studied the interactions of the circadian clock with aryl hydrocarbon receptor (AHR) signaling pathways in preterm babies who received caffeine citrate. This single-center study collected data from 112 preterm infants (<35 weeks gestational age) between July 2017 and July 2018, including apnea-free (n = 48) and apneic (n = 64) groups. Eighty-eight candidate single nucleotide polymorphisms (SNPs) were tested using the MassARRAY system. Association analysis was performed using the PLINK Whole Genome Data Analysis Toolset and SNPStats software. Linkage disequilibrium (LD) and haplotype analyses were performed using Hapview software. No significant intergroup differences in allele distributions or genotype frequencies of CYP1A2, CYP3A4, CYP3A5, and CYP3A7 were detected in our study on preterm babies. Two more SNPs in AHR were found to be associated with determining the response to caffeine citrate therapy in our pediatric patients. Of the 46 candidate SNPs in the CLOCK gene, 26 were found to be associated with determining the response to caffeine treatment in these babies. Interestingly, a significant association was retained for 18 SNPs in the CLOCK gene after false discovery rate correction. Moreover, strong LD formed in those variants in AHR, ADORA2A, and CLOCK genes was confirmed to be significantly associated with a better response to standard-dose caffeine therapy. In summary, CLOCK gene polymorphisms play a role in determining the response to caffeine therapy in premature neonates with AOP. However, whether the AHR and CLOCK signaling pathways crosstalk with each other during caffeine treatment remains largely unclear. Future clinical studies including more immature babies and basic research are needed to explore the mechanism by which circadian rhythms affect the response to caffeine therapy.


INTRODUCTION
The WHO defines preterm birth as birth before 37 weeks of gestation. Premature babies, especially those born very early, often have complicated medical problems. A preterm baby may have trouble breathing due to an immaturely developed respiratory control system. For example, some preterm babies may experience a cessation of breathing for 20 s or longer or a shorter pause accompanied by bradycardia (<100 beats per minute), cyanosis, or pallor, known as apnea of prematurity (AOP) (Eichenwald et al., 2016). Neonates with AOP experience frequent episodes of apnea, resulting in hypoxemia and bradycardia, all of which place the infant at risk for prolonged mechanical ventilation, retinopathy of prematurity (ROP), bronchopulmonary dysplasia (BPD), and long-term neurodevelopmental impairment (Puia-Dumitrescu et al., 2019). Frequent apnea may be one of the most troublesome problems in the neonatal intensive care unit (NICU).
Various mechanisms implicated in the pathogenesis of AOP have been identified, at least in the central nervous system and peripheral reflex pathways (Eichenwald et al., 2016). Ventilatory responses to hypoxia and inhibitory reflexes were exaggerated in preterm neonates. These unique vulnerabilities predispose the neonate to the development of apnea (Mathew, 2011;Martin and Wilson, 2012). Importantly, AOP may also have a critical genetic basis underlying this development-related disorder of respiratory control (Bloch-Salisbury et al., 2010). Methylxanthines have been the mainstay of pharmacological therapy for AOP for over 40 years, and caffeine is generally preferred over other methylxanthines because of its wider therapeutic index and lower incidence of serious complications (Morton and Smith, 2016). However, caffeine is not completely efficient, and in about half of treated infants their apnea frequency remains elevated; thus, aggressive interventions such as mechanical ventilation are needed (Laouafa et al., 2019;He et al., 2021).
The blockade of inhibitory adenosine A 1 receptors (A 1 -AR; encoded by the ADORA1 gene), with the resultant respiratory neural output, as well as the blockade of excitatory adenosine A 2A receptors (A 2A -AR; encoded by ADORA2A gene) located on γ-aminobutyric acidergic neurons has been recognized as the primary mechanism of action of caffeine therapy (Eichenwald et al., 2016). Caffeine also competitively inhibits phosphodiesterases (PDEs) and binds to intracellular calcium channel ryanodine receptors, leading to intracellular Ca 2+ release (Kumar and Lipshultz, 2019). Specific ADORA1 and ADORA2A polymorphisms have been associated with variability in response to caffeine therapy, as well as with a higher risk of AOP and BPD (Kumral et al., 2012). In our recent study, 88 single nucleotide polymorphisms (SNPs) in 19 genes encoding proteins involved in determining the disposition and pharmacological actions of caffeine were genotyped to evaluate the association between genetic mutations and the response to caffeine therapy in 112 preterm infants . The major findings of our previous study included the following: 1) No significant association between the plasma concentrations of caffeine and polymorphisms of caffeine-metabolism-related enzymes such as CYP1A2 and transcription factor aryl hydrocarbon receptor (AHR) was found; 2) polymorphisms of ADORA1, ADORA2A, ADORA3, and PDE4D, especially those of AHR and adenosine dehydrogenase (ADA) genes, play a critical role in determining interindividual variability to caffeine therapy. Interestingly, the influence of AHR polymorphisms on the response to caffeine therapy could not be explained by the mechanisms involving the AHR-CYP1A2 metabolic pathway . This AHR signaling pathway may play an important role in AOP development and the response to caffeine therapy.
In the respiratory control system, A 1 -ARs are found at high densities in the brainstem and anterior hypothalamus, while A 2A -ARs are widely distributed in the medulla. Increased expression of ADORA1 and ADORA2A was observed in both the brainstem and hypothalamus of caffeine-treated neonatal rats. Increased adenosinergic maturation in the central cardiorespiratory areas could partly explain the pharmacological effects of caffeine observed in premature infants (Gaytan and Pasaro, 2012). Adenosine is an inhibitory neuromodulator involved in sleepwake regulation. A 2A -ARs, but not A 1 -ARs, mediate the arousal effect of caffeine (Huang et al., 2005). Intriguingly, caffeine has been shown to influence circadian timing in humans by an A 1 -AR/cAMP-dependent mechanism (Burke et al., 2015). The basic mechanism of the clock is a cell-autonomous interlocked transcription-translation feedback loop sustained by transcriptional activators brain and muscle ARNT-Like 1 (BMAL1), also known as aryl hydrocarbon receptor nuclear translocator-like protein 1 (ARNTL1), Circadian Locomotor Output Cycles Kaput (CLOCK) or neuronal PAS domain protein (NPAS2), as well as period (PER) and cryptochrome (CRY) (Hardin and Panda, 2013). As the core molecular clock is present in nearly every cell in mammals (Mohawk et al., 2012), the circadian clock is likely involved in determining the tissue response to circulating factors, including drugs (Zhang et al., 2021). However, whether the human circadian clock affects the response to caffeine therapy in preterm infants is unknown.
Caffeine citrate was routinely prescribed for AOP management in the NICU. Some preterm infants respond well to caffeine, while others do not. We hypothesized that the human circadian clock is involved in the response to caffeine citrate therapy in preterm infants. The overarching goal of this study was to determine the association between circadian CLOCK gene polymorphisms in preterm babies and the clinical outcomes after caffeine treatment. In addition, this study touches on the crosstalk between the circadian clock and AHR signaling pathways based on our findings in preterm babies who received caffeine citrate treatment.

METHODS
A comprehensive description of the AOP, BPD, ROP, intraventricular hemorrhage (IVH), patent ductus arteriosus (PDA), and necrotizing enterocolitis (NEC), as well as the methods for determining the plasma caffeine concentration and collecting clinical data for a single-center, retrospective study in Chinese preterm infants has been published elsewhere . The caffeine citrate treatment protocol, clinical Frontiers in Pharmacology | www.frontiersin.org January 2022 | Volume 12 | Article 724145 outcomes, and grouping criteria have also been clearly described . Briefly, the apneic group included patients with apnea episodes that occurred even once despite the combined use of necessary nonpharmacological therapies, and the apnea-free group included those without apnea following the administration of caffeine and necessary respiratory support options. The study protocol was approved by the hospital ethics committee of the Children's Hospital of Nanjing Medical University (protocol number 201902082-1).

Study Subjects
Fifty-five female and 57 male preterm babies were included in this study, of whom 104 were born at <32 weeks gestational age. Among these preterm babies, the median postmenstrual age was 29.3 weeks. The median birth weight was 1,265 g. The mean total duration of caffeine exposure was approximately 34 days. The median invasive intubation duration for respiratory support therapy in the 61 infants was 2 days, and the repeat intubation rate was approximately 26%. The use of nasal continuous positive airway pressure or supplemental oxygen therapy continued for a median of 26 days.

MassArray SNP Detection
Because of the CLOCK pathways connected to the AHR pathway and the roles of caffeine in human circadian timing, we focused mainly on the following genes: AHR repressor (AHRR), aryl hydrocarbon receptor nuclear translocator (ARNT), BMAL1, and CLOCK. For CLOCK and BMAL1 genes, SNPs associated with neuropsychiatric diseases and sleep homeostasis were selected as candidates. In addition, CYP3A4, CYP3A5, CYP3A7, and CYP3A43, especially CYP3A7, this isoform of CYP3A, has been demonstrated to metabolize endogenous compounds that are known to be important in the growth and development of fetuses and neonates. Ultimately, 88 relative SNPs were genotyped and evaluated in the current study. The selected 88 SNPs in nine human genes were genotyped using the Agena MassARRAY platform 4.0 with iPLEX gold chemistry (Agena Bioscience, Inc., CA, United States). The main reagents contained Agena polymerase chain reaction (PCR) reagent, Agena shrimp alkaline phosphatase (SAP) reagent, and Agena iPLEX reagent. Genomic DNA was extracted from the peripheral blood cell sediment using a DNA extraction kit (BioTeke Corporation, Wuxi, China) according to the manufacturer's instructions and stored at −80°C before genotyping. The purity and concentration of the extracted genomic DNA samples were assessed by absorbance measurements at 230, 260, and 280 nm using a NanoDrop 2000 UV-Vis spectrophotometer (Thermo Fisher Scientific, MA, United States). The Agena Bioscience Assay Designer 4.0 was used to design PCR amplification and extension primers for the 88 SNP variants. The primers used are listed in Supplementary Table S1. PCR master mixtures were obtained using the Agena PCR reagent set, the PCR procedures were started, and then the mixtures were treated with SAP. After extending the reaction, the samples underwent resin purification processing and dispensing on the 384-well SpectroCHIP bioarray using the MassARRAY Nanodispenser RS1000 spotter (Agena Bioscience, Inc., CA, United States), and the masses of the primer extension products correlated with genotype were then determined using matrix-assisted laser desorption/ionization time-of flight mass spectrometry (MALDI-TOF MS) (Agena Bioscience, Inc., CA, United States). The spectral profiles generated by MALDI-TOF MS were analyzed using MassARRAY TYPER 4.0 software (Agena Bioscience, Inc., CA, United States). Please refer to the study of Cheung and colleagues for a specific process (Cheung et al., 2016).

Statistical Analysis
Statistical analyses were performed using SPSS version 25.0 software (IBM, Armonk, United States), and Figure 1 was drawn using GraphPad Prism 5 (GraphPad Software, CA, United States). The allele and genotype frequencies of various genes, including CYP1A2, CYP3A4, CYP3A5, CYP3A7, AHR, AHRR, ARNT, BMAL1, and CLOCK, were examined for deviation from the Hardy-Weinberg equilibrium (HWE) using the goodness-of-fit chi-square test. Distributions of genotype among neonates in the apneic and apnea-free groups were compared using the goodness-of-fit chi-square test. The Benjamini-Hochberg false discovery rate (FDR) was used for multiple hypothesis testing, and the adjusted p value (p FDR ) was obtained. Multivariable logistic regression analysis was established through variable screening to explore the effect of significant outcomes from the univariate analysis of apneic and apnea-free groups. Statistical significance was set at p < 0.05.
The association between these tested SNPs and the response to caffeine citrate therapy was estimated by calculating odds ratios (ORs) and 95% confidence intervals (CIs) from logistic regression under the allele and recessive [AA vs. (CA + CC)], dominant FIGURE 1 | Significances (plotted as−log10 of nominal p values, closed blue circles) resulting from the associations between the investigated SNPs and different response (apnea-free vs. apneic) to caffeine therapy in preterm infants (n 112). The x-axis represents the relative genomic position/ chromosome, and the y-axis represents the distribution of the nominal−log10(P) values in the two groups of infants. The 26 SNPs, which are associated with the response to caffeine treatment (p < 0.05; above stippled red line), map to the CLOCK gene located on chromosome 4.
Frontiers in Pharmacology | www.frontiersin.org January 2022 | Volume 12 | Article 724145      disequilibrium (LD) and haplotype analyses were evaluated using the Hapview 4.2 software (The Broad Institute, MA, United States), while the difference in distribution of haplotype frequencies between the apneic and apnea-free groups was also tested using the goodness-of-fit chi-square analysis.

Impact of Genetic Variability on the Clinical Response to caffeine Citrate Therapy
No significant intergroup differences in allele distributions or genotype frequencies of genes, that is, CYP1A2, CYP3A4, CYP3A5, and CYP3A7, encoding CYP450 proteins, were found in our study. Similar to our previous findings , two more SNPs (rs1476080 and rs2066853) ( Table 2) in AHR, but not AHRR or ARNT genes, were found to be associated with the response to caffeine citrate therapy in our pediatric patients.
Although the significance disappeared after FDR correction, the AHR polymorphism (rs1476080) still differed significantly (p 0.033) between the two groups after adjusting for clinically important variables, including birth weight, birth height, respiratory support time, repeat intubation and caffeine therapy duration (Table 3).
Most notably, in our 46 candidate SNPs in the CLOCK gene, excluding three unsuccessfully genotyped SNPs, 26 SNPs were found to be associated with the response to caffeine citrate treatment in these neonates ( Table 2). Of these, 18 SNPs still differed significantly (P FDR <0.05) between the two groups after correction for multiple hypothesis testing. Furthermore, multivariable logistic regression analysis showed that the significant influence of these SNPs in the CLOCK gene on the response to caffeine therapy still survived after adjustment (Table 3). Intriguingly, no association was observed between the 14 candidate SNPs in the BMAL1 gene, which encodes two essential components (i.e., BMAL1 and CLOCK) of the circadian clock together with the CLOCK gene, and the clinical response to caffeine treatment.
In addition, some SNPs, especially for variants in ADORA genes, were found to be associated with the incidence rate of BPD and severe neurological injury (SNI) in these preterm infants. It is worth noting that significant associations between genetic polymorphisms of circadian rhythm-related genes, BPD, and SNI incidence, was also observed. The detailed results are shown in Supplementary Tables S2, S3.

Haplotype and LD Analysis
Haplotype analysis was performed to investigate whether various genetic variants were in LD. Fifteen SNPs were excluded because of failed genotyping, and 14 SNPs that were unable to meet the requirement for LD analysis were excluded, which are marked as d and b in Table 1, respectively. Along with our previous research , Figure 1 shows the distribution of the 147 candidate SNPs in more than 40 human genes evaluated in the present study. SNPs significantly associated (p < 0.05; above stippled line) with response to caffeine citrate therapy were ADORA3, CLOCK, AHR, and ADORA2A, which are located on chromosomes 1, 4, 7, and 22, respectively. We then  Figure 2. We observed two small blocks with strong LD between each pair of rs4410790 and rs6968865, and among variants rs1476080, rs6960165, rs2158041, rs7811989, and rs2066853 in the AHR gene, located on chromosome 7. Similarly, block 1 (rs5751862, rs5760410, rs2298383, rs3761422, rs5996696, and rs2267076), as well as block 2 (rs2236624, rs2236625, rs5751876, rs34923252, rs5760423, rs5760425, and rs4822492) with strong LD were found in the ADORA2A gene on chromosome 22. Most notably, 40 genetic CLOCK polymorphisms formed one small block (seven SNPs) and one big block (33 SNPs) on chromosome 4. More importantly, the results of haplotype analysis showed significant differences in the haplotypes, that is, TTAAAG (Block 1) in CLOCK (p 0.0436), CT and TA (Block 1) (p 0.00834), GACGG (Block 2) in AHR (p 0.00243), GACCCC (Block 1) (p 0.0401), and CTTATTC (Block 2) (p 0.0284) in ADORA2A. Frequency distributions existed between these haplotypes and the response to caffeine citrate treatment ( Table 4).

DISCUSSION
The clear benefits and safety of caffeine citrate based on a standard-dose regimen have led to its wide and early use in very low birth weight infants. However, caffeine is not completely efficient, and in approximately 50% of treated infants, the apnea frequency remains elevated (Laouafa et al., 2019;He et al., 2021). In fact, studies are warranted to better understand why some preterm neonates persist with apnea after caffeine therapy, even after increasing the caffeine dose. The AOP phenotype has been attributed solely to immature respiratory system control consequent to preterm birth, but there may also be important genetic influences (Erickson et al., 2021). Therefore, the influence of genetics on the efficacy of caffeine in preterm infants should be better explored (Long et al., 2021).
In the present study, genetic polymorphisms of CYP3A genes, including 3A4, 3A5, 3A7, and 3A43, were also assessed ( Table 1). Especially for CYP3A7, this isoform of CYP3A has been demonstrated to metabolize endogenous compounds that are known to be important in the growth and development of the FIGURE 2 | Linkage disequilibrium (LD) structure (triangle plots) for all investigated single nucleotide polymorphisms (SNPs) (rs numbers below gene names). Red diamonds indicate strong historical LD, r 2 -based haplotype boundaries, among SNPs.
Frontiers in Pharmacology | www.frontiersin.org January 2022 | Volume 12 | Article 724145 fetus and neonate. In addition, CYP3A7 is an important component in the development and protection of the fetal liver and plays a role in certain disease status (Li and Lampe, 2019). However, because of MAF values < 0.001 or genotyping failure ( Table 1), most of the selected SNPs were excluded from further association analysis.
A major finding of this study was that AHR genetic variations (rs1476080 and rs2066853), but not AHRR or ARNT genes, were found to be associated with the response to caffeine therapy (Tables 2, 3). Our previous study showed that two particular polymorphisms of the AHR gene (rs6968865 and rs4410790) were significantly associated with the response to caffeine treatment between the two groups of preterm infants . Furthermore, these variants in the AHR gene, located on chromosome 7, formed strong LD, with increased ORs for haplotypes CT (p 0.00834), TA (p 0.0122), and GACGG (p 0.00243), which in part determined a better response to caffeine treatment ( Figure 2; Table 4). Based on the plasma concentration data of caffeine, we excluded that the AHR-CYP1A2 metabolic pathway was responsible for the variable response to caffeine therapy . However, AHR is necessary to protect fetal human pulmonary microvascular endothelial cells against hyperoxic injury (Zhang et al., 2015) and plays a critical role in the maintenance of lung health (Guerrina et al., 2018). Therefore, the findings of our study suggest the potential role of AHR signaling in preterm infants who experience AOP episodes. The AHR signaling pathway may act alone or in combination with adenosine receptors and circadian CLOCK.
Another important finding in the present study was the significant association between circadian CLOCK polymorphisms, but not BMAL1 variants (also known as ARNTL1; Table 1), and the response to caffeine citrate therapy in these preterm babies (Table 2). Notably, after adjustment for clinically important variables, such as birth weight, birth height, respiratory support therapy, repeat intubation, and caffeine therapy duration, the significance was retained (Table 3). Moreover, 40 candidate CLOCK SNPs formed one small block and one big block with strong LD (Figure 2). Especially for the small block, the haplotype TTAAAG was found to be associated with a better response to caffeine therapy (OR 3.1, p 0.0436; Table 4).
Caffeine has been reported to affect the phase of the human circadian clock and primarily affect human cellular circadian clocks via an A 1 -R/cAMP-dependent mechanism (Burke et al., 2015). Recent studies have revealed that caffeine and adenosine alter clock gene expression and circadian rhythms in vitro and in vivo via the Ca 2+ -ERK-AP-1 pathway (Jagannath et al., 2021). The present study is the first to show that genetic polymorphisms of CLOCK, encoding one of the two core components of the circadian rhythm, were significantly associated with the response to caffeine therapy. These findings suggest that the circadian rhythm may play critical roles in the response to caffeine citrate therapy in babies experiencing AOP episodes. Interestingly, caffeine has been reported to increase the light responsiveness of the mouse circadian pacemaker (van Diepen et al., 2014) and also affects the human circadian clock in vivo and in vitro (Burke et al., 2015). Therefore, synchronizing caffeine with the circadian rhythm may be useful for optimizing its treatment efficacy. In the meantime, investigations into essential mechanisms may provide therapies to reset or amplify circadian signals. However, very few studies have been performed to evaluate the proper timing of methylxanthine dosing (chronotherapy) as a means to maximize its efficacy and possibly reduce its adverse effects. Further research in this field is warranted to provide new insights and clinical advantages. One of the major strengths of the present study was our ability to assess how genetics affect the response to caffeine based on 147 candidate SNPs in more than 40 human genes ( Table 1), encoding various proteins related to the disposition of caffeine and/or pharmacological mechanisms of caffeine's actions, as well as circadian control functions. Along with our previous report , we observed significant associations between ADORA1, ADORA2A, ADORA3, PDE4D, ADA, AHR, and CLOCK polymorphisms and the response to caffeine treatment (Tables 2, 3). Notably, the haplotypes of AHR, CLOCK, and ADORA2A were found to be associated with a better response to caffeine therapy ( Table 4). A 2A -AR regulates phrenic nerve activity, and blockage with caffeine may be one mechanism responsible for the efficacy of xanthine against AOPs. Additional A 1 -AR in the brainstem may regulate hypoxic ventilator drive. Such central stimulation results in increased respiratory drive, increased sensitivity to hypercarbia, decreased hypoxic suppression of respiration, and increased diaphragmatic contractility (Morton and Smith, 2016). Caffeine competitively inhibits PDEs, a group of enzymes that degrade cAMP (Burke et al., 2015). In addition, intracellular adenosines undergo metabolism to inosine by ADA, a deamination that occurs preferentially under pathological conditions featuring raised adenosine levels (Borea et al., 2018). Collectively, caffeine may play a critical role in AOP treatment by directly and/or indirectly agonizing the adenosine-AR-cAMP pathway, rather than the AHR-FIGURE 3 | A model of caffeine, AR signaling pathway, AHR signaling pathways, and human circadian clock. (A) Caffeine alters intracellular cAMP levels through binding to adenosine receptors (A 1 , A 2A , A 2B and A 3 ) and inhibiting phosphodiesterases. In addition, caffeine mobilizes intracellular Ca 2+ from the endoplasmic reticulum through activation of the RyR channels. Increased cytosolic cAMP/Ca 2+ signaling culminates CREB activation acting in tandem with rhythmic transcriptional activation by CLOCK/BMAL1. CLOCK/BMAL1 binds to E-box elements in the PER promoter, where they act to stimulate PER transcription. (B) Inactive AHR complexes with HSP90, AIP, p23 and SRC in the cytosol. Upon agonist binding, AHR and some components of the chaperone complex translocate to the nucleus, where AHR forms a dimer with ARNT binding to the XRE to control gene expression (such as AHRR, CYP1A1, CYP1A2 and CYP1B1). (C) Activated AHR can physically interact with BMAL1, which exhibits high homology with ARNT, thereby reducing CLOCK-BMAL1 interactions and repressing PER transcription. In addition, activated AHR phosphorylates and activates JNK, which represses CRE-mediated transcriptional activity to suppress PER transcription (left). Rhythmic transcription of AHR is driven by CLOCK: BMAL1 at E-Box promoter elements. PER and CRY inhibit AHR transcript levels (right). Abbreviation: cAMP, cyclic adenosine monophosphate; RyR, ryanodine receptor; PDEs, phosphodiesterases; AHR, aryl hydrocarbon receptor; AHRR, AHR repressor; ARNT, AHR nuclear translocator; XRE, xenobiotic response element; HSP90, 90 kDa heat shock protein; AIP, AHR-interacting protein. CLOCK, circadian locomotor output cycles kaput; BMAL1, brain and muscle ARNT-Like 1; CREB, cAMP response element-binding; CBP, CREB binding protein; PER, periods; CRY, crypto-chromes.
Frontiers in Pharmacology | www.frontiersin.org January 2022 | Volume 12 | Article 724145 and apneic groups. However, genetic candidate variants in AHR and CLOCK genes were found to be associated with variable responses to caffeine treatment in these preterm babies. Moreover, the variants in AHR, ADORA2A, and CLOCK genes formed strong LD with increased ORs, which were associated with a significantly better response to standard-dose caffeine therapy. Future and larger studies, as well as basic research, are required to understand how these SNPs affect the response to caffeine therapy. Our findings also indicate that circadian rhythm may play an essential role in the response to caffeine therapy in babies experiencing AOP episodes.

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT
• Caffeine citrate therapy is a preferable choice for preterm infants with AOP in the NICU. • The current standard-dose caffeine therapy leads to variable clinical outcomes. It remains unclear why some preterm infants respond well to this therapy but others do not. • Questions regarding dose selection, routine TDM of caffeine, and influence of genetic variants are still unanswered.

WHAT THIS STUDY ADDS
• No significant intergroup differences in allele distributions or genotype frequencies of CYP1A2, CYP3A4, CYP3A5, and CYP3A7 were found in our study on preterm babies. • To the best of our knowledge, the present study is the first to report that CLOCK gene polymorphisms are involved in determining the response to caffeine therapy in premature neonates with AOP. • Strong linkage disequilibrium was observed in AHR, ADORA2A, and CLOCK variants. These genetic variants were significantly associated with a better response to standard-dose caffeine therapy. However, whether the AR, AHR, and CLOCK signaling pathways crosstalk with each other during caffeine treatment remains largely unknown.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by The hospital ethics committee at the Children's Hospital of Nanjing Medical University. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.

AUTHOR CONTRIBUTIONS
J-YL, H-LG, FC, JX, and RC: Principal investigators for the study, data analysis, primary authors of the paper. XH, Y-HH, YX, YL, and LL: Performed the data collection and analysis. X-SD: Assisted in the design and performance of the study and the writing of the paper. All authors have read and approved the final manuscript.