Introduction

The need for characterizing environmental influences on human physiology in a holistic manner, similar to what genomics has done for our genetic code, has been recognized for over two decades. In 1995, Anthony, Eaton, and Henderson proposed the broad study of the role of the environment in mental health [1]. Later in 2001, a special supplement of the British Journal of Psychiatry [2] was devoted to the “envirome” concept, which is defined as “the total set of environmental factors, both present and past, that affect the state, and in particular the disease state, of an organism” [3]. In 2005, the epidemiologist Christopher Wild first proposed the concept of the human exposome, which encompasses the entire set of environmental exposures from conception to death [4]. The exposome comprises two components: the external exposome, which ranges from specific exposures that impact individuals (e.g., toxicants/chemicals, heavy metals, radiation, diet, infections, smoking, physical activity, psychosocial stress) to general exposures that impact entire populations (e.g., climate, air pollution, social capital); and the internal exposome, which reflects all external exposures that enter into the body and the biological responses to these exposures [5, 6]. The exposome paradigm represents a shift toward discovery based, in-depth exposure assessment by emphasizing the importance of multiple exposures that occur in mixtures, unique windows of susceptibility to the health effects of exposures over the lifecourse, and interindividual variability in exposure responses that may contribute to differences in health outcomes [7].

High-throughput, cutting-edge technologies are necessary to study the exposome on a broad scale. One such technology for exposome profiling is untargeted metabolomics, which detects thousands of small molecule metabolites in biological specimens of endogenous and exogenous origin without specifying these metabolites a priori [8]. By enabling the discovery of both novel chemical mixtures and endogenous metabolic responses to exposures, untargeted metabolomics is a key platform for exposome research, especially for the characterization of the internal exposome.

Untargeted metabolomics, like other omics technologies, relies on biological samples. In studies in adults, urine and blood collection are standard. In contrast, exposome studies in children require less-invasive sampling methods that can be repeated across developmental windows. Here, we describe emerging matrices collected using non- or minimally invasive biosamplers that can be used for untargeted metabolomics in children’s health research, including dried capillary blood, interstitial fluid, saliva, teeth, and hair. We also highlight applications of these biosamples in population-based exposure research, particularly children’s studies when available. With this review, we provide a foundation for the development of untargeted metabolomics methods for future children’s exposome research. Next steps require that these matrices are optimized, tested, and validated for metabolomics studies.

Minimally Invasive Sampling for Children’s Exposome Research

Life-stage transitions and physiological development occur more rapidly in children compared with adults, which necessitates biomonitoring at abbreviated timescales. In adults, timescales of up to 10 years may not be critical (e.g., there is not much difference between 45 and 55 years old); in contrast, windows of exposure sensitivity are much smaller for children. Data from the Amsterdam Longitudinal Aging Study, which examined four age classes (0–1, 2–5, 6–10, and 11–14 years old) exposed to the Dutch famine, found that severe undernutrition during ages 11–14 in females was most significantly associated with developing diabetes mellitus and/or peripheral arterial diseases at ages 60–76 [9]. Cohort mortality data in France, England and Wales, and Sweden revealed that stressors experienced by males during ages 10–14 are more strongly associated with a decrease in lifespan compared with those experienced during infancy and ages 1–4, 5–9, and 15–19 [10]. Other examples that used direct biological measurements of exposure include an examination of the interplay among the early-life microbiome, allergy, and wheeze in a cohort study of 244 children in Australia. Certain microbiome profiles in the first 2 years of life, in tandem with early allergic sensitization, were associated with chronic wheeze at 5 years of age; however, in the absence of allergic sensitization, the microbiome profiles were only associated with transient early wheeze, which resolved by the fourth year of life [11]. Finally, in a study in Mexico City examining longitudinal exposures to metal mixtures and children’s mental health, metal markers in deciduous teeth identified two specific windows of susceptibility for mixtures of manganese (Mn), zinc (Zn), and lead (Pb) that were associated with increased anxiety in children aged 8–11. The first window (0–8 months) appeared driven by Mn, while the second window (8–12 months) was driven by the metal mixture and dominated by Pb [12]. The rapid development of the brain during the first 3 years of life highlights the high temporal resolution of exposure measurements needed to adequately examine associations with neurological outcomes. Together, these examples suggest that, in order to identify critical windows of exposure and delineate disease mechanisms, pediatric exposure studies require in-depth, longitudinal follow-up with direct biological measurements during short timescales from birth through adolescence [13].

Longitudinal studies of biological samples in adults are facilitated by careful selection of the biological matrix most appropriate for the exposure of interest and for the study population. Primary factors of determination include the toxicokinetics and concentrations of the analyte(s) of interest and relevance to the health outcome, but the logistics and costs of specimen collection and storage are also considered. For example, urine is most appropriate for measuring acute, non-persistent chemical exposures with rapid half-lives, while serum or plasma is the gold standard for persistent exposures with higher blood-level concentrations. Repeated measures of these samples are readily collected in adults, with the benefits of established handling protocols, decades of epidemiological findings, and published biomarker concentrations for comparisons across populations, in support of continued use of these matrices.

While these adages also apply to children’s health research, collection of blood and urine samples in children is challenging, especially as the exposure window shifts to earlier life. Venous blood collection can be painful and invasive and is thereby not generally acceptable to parents. Young children may be fearful of providing urine samples in a cup, and sampling in infants and toddlers who use diapers requires a collection device that is susceptible to contamination [14]. Collecting neonatal samples through cord blood requires timely collection during delivery, and while amniotic fluid can contribute valuable information on cumulative fetal exposure during pregnancy, sampling poses inherent risks to the health of the fetus. Therefore, the use of non- or minimally invasive methods for at-home biological collection can facilitate exposome studies throughout childhood development from the fetal stage onward (see Fig. 1). A comparison of matrix volumes and key advantages of minimally invasive technologies that will be discussed in this review, including dried blood spots (DBS), dried plasma spots (DPS), volumetric absorptive microsampling (VAMS), interstitial fluid (ISF), touch-activated phlebotomy (TAP), saliva, teeth, and hair are summarized in Table 1.

Fig. 1
figure 1

Schematic of appropriate ages for selected at-home biosampling devices. Age ranges are prenatal (before birth), newborn (birth–6 months), infant (6–12 months), toddler (12–36 months), preschool (3–5 years), school age (5–10 years), adolescents (10–19 years), and adults (> 19 years). Solid arrows reflect concurrent exposure measurements. Dashed arrows represent retrospective exposure measurements. aat-home interstitial fluid collection is still in the prototype phase. Similar at-home technologies using microneedle arrays but for blood collection (e.g., TAP) are commercially available

Full size image
Table 1 Comparison between urine and plasma and alternative matrices for biosampling in pediatric populations
Full size table

Dried Capillary Blood

Collection of small volumes of blood from microsampling (e.g., < 100 μL) [15] is a less-invasive means of blood collection than venous blood (usually ~ 10 mL) and, when dried, is more stable at higher temperatures than liquid blood [16]. In fact, DBS collection, which involves collecting a blood droplet via lancet from a finger- or heel-prick onto a paper card and allowing it to dry, has been utilized since 1963 in newborn screening programs [17] and is now being applied in exposome research [18, 19•]. Beyond the classic Guthrie card, newer technologies offer similar advantages of a traditional dried sample—namely, the absence of a required cold chain and easy and economical storage—while also overcoming the limitations of unknown sample volume and the effect of hematocrit. The blood volumes from a droplet collected onto the Guthrie paper are typically unknown, but are estimated to range from 2 to 72 μL in a single spot from a newborn heel prick card [20]. Therefore, metabolite measurements in whole spots cannot be directly compared across subjects. An alternative is to use a set diameter punch-out of a spot for each subject. However, hematocrit, which is the volume percentage of red blood cells in whole blood and directly proportional to the viscosity of the blood, affects the flux and diffusion of the blood on the filter paper, which introduces a source of unwanted variation in analyte measurements of punches [21]. At a high hematocrit value, the distribution of a blood sample through the paper can be poor, resulting in a smaller blood spot when compared with a blood sample with a low hematocrit. Accordingly, a punch-out from a DBS with high levels of hematocrit will generally contain more blood (and presumably higher mean metabolite levels) compared with a punch-out from a low hematocrit sample. Comparisons of metabolite levels measured from uniform punches may, therefore, be biased by differences in hematocrit levels [19•]. This phenomenon is of particular concern for infant and children’s blood, which can range in hematocrit from 33 to 65% depending on age, sex, and health status [22,23,24]. New microsampler technologies seek to reduce this hematocrit bias and are making headway as acceptable replacements to traditional sampling in nonclinical pharmaceutical research [15], suggesting their potential future use in other research fields. Furthermore, since all of these technologies utilize capillary blood obtained from a lancet prick, as is conducted in newborn screening programs worldwide, they are appropriate for collecting blood from children as young as neonates. While extensive reviews can be found elsewhere [25••, 26, 27•], we provide a summary of these technologies and their applications.

Blood microsamplers new to the commercial market include several technologies that require a classic capillary prick and produce a droplet on a card similar to DBS. Plasma microsampling offers an alternative to the laborious process of aliquoting and centrifugation to obtain venous plasma samples on microscale volumes. The Noviplex™ card uses a combination of spreading and filtering to remove red blood cells from a whole-blood droplet (60 μL) via lancet and then deposits a quantitative volume of plasma onto a precut disc(s), generating dried plasma spots [28]. Complete sample collection occurs in approximately 3 min; after drying for 15 min, the sample can be stored or extracted similar to a whole-blood DBS. Gravimetric analysis suggests little variability in the plasma volume deposited on the disc for hematocrit levels ranging from 20 to 71% when 50 μL whole blood is applied, and an acceptable variability between discs when vitamin D is measured in the plasma [28]. However, a study evaluating the use of Noviplex cards for steroid measurements suggests that inter-card variability may be larger and absorption can be significant, depending on the analyte of interest [29].

Other technologies use microfluidic properties to obtain a quantitative volume of blood from a lancet-derived blood droplet. These technologies include the HemaPEN® (Trajan Scientific and Medical, Australia), which uses capillaries to draw and then dispense a fixed blood volume of 2.7 μL onto pre-punched DBS paper [30] the HemaXis DB (DBS System SA, Gland, Switzerland), which uses a microfluidic chip to draw a fixed amount of blood from a droplet and deposit it onto a standard DBS card [31] and the Captainer-B (KTH, Stockholm, Sweden), which draws 13.5 μL of whole blood from a droplet placed at the inlet and empties the channel onto perforated discs. To date, only Captainer-B has been utilized for exposure measurements; an alcohol exposure biomarker, phosphatidylethanol, was quantified in whole blood from an adult hospital population, yielding comparable results to blood sampling via classic DBS collection and venipuncture analysis [32].

Another technology, volumetric absorptive microsampling (VAMS™) (Neoteryx, CA, USA), uses a dried blood sample, reducing the need for cold-chain requirements, but involves a single-use applicator with a hydrophilic tip that absorbs a fixed volume of blood. This applicator can be dried and stored similar to the DBS technology, but sample analysis is performed on the whole applicator tip, eliminating the need for manual punching of the DBS. Several studies have investigated the reproducibility and accuracy of quantitative environmental chemicals and endogenous metabolites using VAMS. Examples include measurement of praziquantel [33], perfluorinated compounds [34], and caffeine [35], as well as 36 amino acids and organic acids in whole blood [36]. The former two studies included biospecimen collection from children, highlighting their applicability as blood collection devices appropriate for pediatric populations.

Preliminary studies also suggest that VAMS is appropriate for untargeted metabolomics analysis. Untargeted GC-MS analysis on whole blood collected through VAMS from 10 adult females with breast cancer and 10 healthy controls indicated nine metabolites significantly (p < 0.05) associated with breast cancer status [37]. Untargeted LC-MS analysis profiled thousands of polar features in adult donor blood, which were stable when the VAMS capillaries were stored at − 80 °C for up to 6 months [38]. Overall, blood sampling using VAMS is acceptable for children’s collection and compatible with targeted and untargeted metabolomics analysis, and pilot studies provide proof-of-principle for their application in epidemiologic studies.

A liquid blood alternative, “TAP” (touch-activated phlebotomy) technology, is a self-contained device that, when attached to the skin with a hydrogen adhesive, uses an array of solid microneedles and vacuum to withdraw whole blood and perform anticoagulant mixing [39]. While still in the prototype phase, targeted profiling of metabolites from 5 adult donors suggests that median biological variability for all 45 metabolites was comparable between traditional blood draws and TAP. In particular, quantitation of 39 metabolites overlapped between the methods [40]. This comparability of measurements across other studies using venous blood makes TAP appealing for its convenience, i.e., avoiding the logistics and costs of a phlebotomist, but the sample is maintained in liquid form and thus requires a cold chain. In addition, its size (4.7 × 3.5 cm2) limits potential pediatric applications to only school-aged children and adolescents.

Interstitial Fluid

Interstitial fluid (ISF), which comprises 80% of the extracellular fluid (ECF) in the human body, is a relatively unexplored matrix for analyte detection. ISF fills the spaces between cells and tissues and transports nutrients, waste, and signaling molecules between cells and capillaries [41]. While ISF is believed to have a similar chemical composition to plasma, it offers two distinct advantages for metabolomics/exposomics analysis. First, unlike plasma, wherein analytes reflect an integrated measurement from systemic circulation, analytes in ISF can also reflect the microenvironment at the sampling site. This characteristic can enable the identification of disease- or tissue-specific biomarkers: for example, several studies have examined the composition of tumor tissue ISF for the discovery of novel cancer biomarkers using proteomics [42,43,44,45]. Second, ISF contains markedly lower concentrations of proteins than plasma [46]. Given the potential for proteins to interfere with metabolite detection—and the variable recovery of metabolites following different sample preparation strategies for protein removal [47]—exposomics analysis utilizing ISF rather than plasma may enhance the detection of low-abundance metabolites and reduce non-biological variance in metabolite measurements.

Historically, the investigation of ISF for analyte detection in population-based and clinical studies has been hampered by a lack of methods for rapid, less-invasive ISF sampling. However, emerging technologies for minimally invasive ISF collection are likely to enable ISF sampling in pediatric studies. Microneedle patches, originally developed as a means for drug delivery, sample dermal ISF through an array of small needles that penetrate the skin’s surface [48]. Similar to TAP technology for capillary blood collection, microneedle patches are simple to administer and cause little-to-no pain due to the small volume of ISF collected (up to 10 μL) [49], which may provide an alternative to venipuncture for ECF sampling for individuals with fear of needles, including children. For routine ISF sampling in healthy populations, collection of ISF from skin is ideal due to its accessibility [49]. An established method for dermal ISF collection is suction blister sampling, which involves the application of a vacuum to the skin at an elevated temperature for 1 h to create a fluid-filled blister [50]. While this technique is well established for collection of large volumes of ISF, the length of the procedure and associated discomfort has limited its use in human studies and clinical practice. Microdialysis and open-flow microperfusion, two technologies considered minimally invasive and that allow for continuous sampling, collect ISF through probes implanted at specific sites [51] but can cause local tissue trauma and require trained personnel to administer.

Metabolomics studies of ISF, while few in number, nevertheless suggest the utility of ISF for exposomics. The measurement of ISF metabolites provides a proxy for plasma analytes in some cases and novel information in others. In 10 human subjects, an LC-MS untargeted metabolomics analysis of dermal ISF sampled using suction blisters identified numerous amino acids, lipids, nucleotides, and exogenous compounds, including dietary compounds, environmental toxicants, and pesticides [52••]. While ISF and plasma were found to have distinct profiles, correlation analyses indicated that dermal ISF can be a reliable surrogate for plasma for the detection of many analytes, including endogenous metabolites and xenobiotics such as caffeine [52••]. In 18 healthy volunteers, a LC-MS lipidomics analysis found similar lipid profiles between suction blister ISF and plasma [53]. A recent GC-MS metabolomics study of skeletal muscle ISF, sampled with a microdialysis catheter in 5 human volunteers, detected 414 metabolites across a range of classes [54]. In a murine model of pancreatic ductal adenocarcinoma, LC-MS analysis of 118 metabolites found a unique metabolomic composition of tumor ISF compared with plasma, showcasing the utility of ISF metabolomics for disease biomarker discovery [55]. With rapid developments in minimally invasive ISF sampling technologies, ISF metabolomics holds great promise for pediatric exposomics studies.

Saliva

Saliva is an attractive matrix, particularly for vulnerable populations like neonates, toddlers, and children, due to its non-invasive collection. Saliva analyte concentrations are linked to blood via passive and active transportation, diffusion, and/or ultrafiltration through salivary glands and gingival crevicular fluid [56]. These analytes reflect body concentrations of exogenous compounds obtained through therapy, drug dependency, environmental exposures, or recreational uses, as well as endogenous compounds related to psychosocial factors, neurological status, and nutritional and metabolic influences [57], providing a snapshot of the internal exposome. Saliva is even the diagnosis matrix of choice for both Cushing’s syndrome and Addison’s disease [58]. In addition, saliva collection does not require highly trained personnel, is safer to handle than blood, and allows for repeated sampling without the risk of anemia, which is increased for repeated phlebotomy in at-risk infants and children [59].

Saliva collection can be performed via passive or stimulated collection. Passive collection requires the donor to hold the flask at their lips with their head tilted, and slowly collect the saliva while restricting movement in their mouth for approximately 5 min [60], limiting its use to older children and adolescents. Alternatively, stimulated collection methods use chewing actions to generate saliva. For example, the Salivette® uses a cotton swab that can be placed in the mouth and absorbs saliva as the infant, young child, or adolescent chews on it. In less than 2 min, the cotton swab is spit out; subsequently, centrifugation is used to draw out the clear liquid saliva that can be used for analysis [61]. While some studies suggest differences in metabolomics profiles between the two means of collection in adults [62], a single targeted study of 45 metabolites in children found no difference in profiles between active or passive saliva collection [63•]. However, some salivary metabolites are under circadian control and have high diurnal variation [64, 65]. To reduce non-biological variability in saliva analyte measurements, established protocols must be maintained throughout sample collection, including using the same collection device, obtaining the sample at the same time of day for all subjects, and requiring at least 1 h of fasting before collection to reduce contamination by food stuffs [66, 67].

While not as established as urine or venous blood collection for exposomics analysis, saliva is a valuable biological matrix in targeted and untargeted metabolomics studies, including in pediatric populations. Several chronic diseases have been associated with salivary metabolite levels [67], such as metabolic syndrome and obesity [61], type 1 diabetes [68], and congenital Zika syndrome in infants [69]. Furthermore, saliva metabolites have been linked with exposures such as traffic-related air pollution [70] and pesticides [71]. In the latter study, adult male pesticide applicators showed alterations in 16 saliva metabolites linked with energy metabolism and amino acid and polyamine metabolism, while only 13 urine metabolites measured in the same population (seven of which overlapped with saliva) were significantly associated with the exposure group. This finding illustrates the additional value of the non-invasive sampling matrix compared with a traditional biological matrix for biomarker discovery.

Teeth

The analysis of biomarkers in shed deciduous teeth is a unique, non-invasive approach for assessing retrospective exposures at high temporal resolution. In the developing tooth, deposition of both enamel and dentine commences during early gestation (14 to 19 weeks); this process occurs in a rhythmic manner, forming incremental lines similar to growth rings in a tree throughout life [72]. During this development and continued growth, circulating organics and metals in the body are also deposited, providing a reservoir of time-stamped exposures. At birth, the neonatal line forms which functions as a histological landmark demarcating pre- and postnatal parts of teeth [73]. Based on timing estimated from the incremental and neonatal lines, regions of the tooth corresponding to distinct developmental windows can be sampled for exposure analysis, including the prenatal period [74].

Assessing biomarkers in teeth offers several advantages compared with traditional biomatrices, particularly the ability to retrospectively assess cumulative exposures and exposure timing as well as the non-invasive nature of the sampling. However, unlike other biomatrices that can be collected at point-of-care in a clinical setting or in accordance with a scheduled cohort visit, teeth are obtained when they are naturally shed between the ages of 5 and 13 years. Therefore, depending on the health outcome of interest, studies may have to wait until participants shed teeth before they can be obtained for analysis, or conversely, studies must recruit participants with stored naturally shed teeth to donate, which may be stored under variable conditions. Despite these limitations, the use of teeth to retrospectively measure prenatal and early life exposures holds promise for pediatric exposome analysis.

The utility of teeth to assess temporal exposures to heavy metals, such as lead (Pb), is well documented. In early studies, whole teeth were digested to determine toxicant concentrations [75, 76]. In later work, researchers proposed the use of detailed intra-tooth spatial measurements to assess perinatal Pb exposure. For example, Gulson and Wilson [77] used measurements of stable Pb isotopes in enamel to estimate prenatal exposure. More recently, detailed studies using laser- and nuclear-beam based sampling methods have provided weekly temporal information on prenatal metal exposure during the second and third trimesters. Detailed validation of metal concentrations in teeth against blood and urine metals collected during pregnancy and early childhood have also been undertaken [78,79,80].

Advances in tooth biomarker technology indicate that similar analyses of organic compounds in teeth are feasible using mass spectrometry. Past studies have analyzed pulverized teeth for the presence and quantification of various organic chemicals, contaminants, and metabolites such as drugs [81,82,83], plastics additives [84], metabolites of alcohol [85], tobacco [86,87,88], and pesticides [89, 90]. More recently, there have been advances in undertaking omic-scale analyses while retaining the temporal information from different tooth fractions. This has allowed thousands of prenatal chemical signatures to be obtained at trimester-by-trimester temporal resolution [91••]. These studies show the potential of using teeth biomarkers to substantially improve exposome analysis, particularly for case-control studies of low frequency diseases and disorders, in which prospective designs are not economically feasible.

Hair

Hair grows at approximately 1 cm/month, working as part of the integumentary system to remove toxicants from the body. Circulating exogenous and endogenous metabolites in the blood are transferred to the hair shaft through the hair follicle as it grows [92]. Thus, hair serves as a long-term inert storage matrix for measuring toxicant exposure [93] from different routes of entry. When segmented, hair can provide information on the time of exposure across several months, depending on the length of the hair and if knowledge of date of collection is present [94]. In addition, fetal hair begins growing from approximately 20 weeks gestation [95] and can be obtained up to several months after birth, providing a means to retrospectively assess prenatal exposures during the late second trimester onward [96]. Hair is easy to collect: the sample is usually cut close to the scalp, taped to indicate the scalp end, and stored in aluminum foil or paper in a zip-close bag [94, 97]. Less than 5 mg is needed, equivalent to 2–3 strands [94, 98]. From a collection standpoint, this makes hair one of the easiest biospecimens to obtain from infants and children to assess exposure during critical developmental timepoints.

Hair analysis has evolved from its original focus on drugs of abuse, pharmaceuticals, and metals to its current use for targeted environmental chemical analysis. Several recent applications of hair analysis in children have enabled quantification of cumulative exposure. For example, organophosphate pesticide (OP) exposure was monitored in children of an agricultural community [97]. OP overall detection frequency was higher in hair than in spot urine, possibly due to the cumulative nature of hair that includes a wider exposure window and a potentially more reliable and sensitive measurement [97]. Similarly, urban air pollution [polyaromatic hydrocarbons (PAHs) and their metabolites] and passive tobacco smoke (cotinine and nicotine) exposure were monitored in hair samples of 25 children aged 2–11 years from either an urban area (Paris, France) or rural area (Yeu Island, France) [99]. Cotinine and nicotine as well as 10 out of the 15 PAHs were detected in 100% of the subjects, showing the sensitivity of the matrix. Interestingly, younger children had higher levels of select PAH metabolites suggestive of metabolic or physiologic differences related to age or closer physical proximity to the higher air pollution levels observed closer to ground [99]. In another recent study, hair from children (ages 2–12) and hair from adults showed concentration differences in three endocrine disrupting chemical (EDC) biomarkers possibly reflecting different exposure routes (food vs. dermal) [100]. These studies provide evidence of the valuable exposure information obtained from hair as well as illustrate the feasibility of obtaining these samples for population-based children’s research.

More recently, hair has been used to measure endogenous metabolites. Amino acids, lipids, and steroids, for example, have all been detected in targeted hair analyses, which have led to recent exploration of metabolomics of hair as a means for biomarker discovery in a range of clinical applications [101,102,103]. In particular, maternal hair during pregnancy has been explored as a non-invasive measure to investigate biomarkers of pregnancy complications and birth outcomes [94, 98, 104, 105]. While there are currently few metabolomics studies using hair, these proof-of-concept studies suggest that hair analysis in children’s studies will provide valuable insights into the fetal and early-life exposome.

Instrumentation for Measurements

The most popular analytical methodologies for highly multiplexed metabolite measurements include mass spectrometry (MS) or nuclear magnetic resonance spectroscopy (NMR). While NMR has the advantage of quantitative measurements with minimal sample preparation, mass spectrometry can be easily adapted to quantitatively measure 10s to 100 s of endogenous metabolites or environmental chemicals such as phthalates, phenols, or pesticides in a single analytical run using labeled internal standards for absolute quantification with tandem MS (MS/MS). While the complex matrix, small volumes, and low concentrations in these emerging biosamples and biomarker techniques cause analytical challenges for metabolite analysis, recent advances in high-resolution (HR) MS now enable sensitivities of ng/mL concentrations and span 5–6 orders of magnitude, allowing for the expansion of quantitative chemical assays to the semi-quantitative analysis of 100s to 1000s of metabolites measured in untargeted methodologies [106, 107]. With further advances in sample preparation techniques, these untargeted technologies can be paired with non- and minimally invasive biosampling as discussed here to robustly investigate the myriad of exposures and multifactorial etiologies associated with exposome research in children. A summary of the exposure studies utilizing these analytical platforms for analysis of non- and minimally invasive samples highlighted in this review can be found in Table 2.

Table 2 Selected examples of metabolomics or exposure studies using dried blood microsamplers, interstitial fluid, saliva, teeth, or hair
Full size table

Conclusions

Alternative biomatrices will greatly expand the characterization of the exposome. As outlined in Table 1, emerging technologies for biosampling from dried capillary blood, ISF, saliva, teeth, and hair provide unique advantages over traditional urine and plasma collection that can help guide future longitudinal exposome studies in pediatric populations. Key criteria include (1) the invasiveness of the procedure and acceptability to the study participants, (2) storage stability over the short and long term, (3) the ability of the matrix to provide information on the timing of exposure (including past), (4) the costs of administering the technology, (5) diurnal sample variability, (6) for an alternative biomatrix, whether it provides comparable or new information compared with plasma or urine, and (7) the possibility for at-home sample collection. While these technologies hold promise for exposome studies, applications in metabolomics are still limited. Therefore, optimization and validation studies of metabolomics workflows in these novel matrices are imperative prior to the widespread adoption of these analyses in large studies. Nevertheless, untargeted metabolomics investigations of the alternative biomatrices described in this review are poised to substantially advance the understanding of the impacts of early environmental exposures on health over the lifecourse.