An overview of methods of fine and ultrafine particle collection for physicochemical characterisation and toxicity assessments

Particulate matter (PM) is a crucial health risk factor for respiratory and cardiovascular diseases. The smaller size fractions, ≤2.5μm (PM2.5; fine particles) and ≤0.1μm (PM0.1; ultrafine particles), show the highest bioactivity but acquiring sufficient mass for in vitro and in vivo toxicological studies is challenging. We review the suitability of available instrumentation to collect the PM mass required for these assessments. Five different microenvironments representing the diverse exposure conditions in urban environments are considered in order to establish the typical PM concentrations present. The highest concentrations of PM2.5 and PM0.1 were found near traffic (i.e. roadsides and traffic intersections), followed by indoor environments, parks and behind roadside vegetation. We identify key factors to consider when selecting sampling instrumentation. These include PM concentration on-site (low concentrations increase sampling Jo ur na l P re -p ro of Journal Pre-proof


Introduction
Particulate matter (PM) is of great concern due to its association with a variety of health impacts, such as respiratory and cardiovascular diseases (Dockery and Pope 1994;Kappos et al. 2004;Heal et al. 2012). Numerous studies have demonstrated the association of airborne PM mass and respiratory admission in hospitals (Bell et al., 2014;Tecer et al., 2008). Other works have shown a strong association between the chemical composition and biological response, such as the influence of transition metals such as iron and manganese and their potential to cause oxidative stress by producing reactive oxygen species Charrier and Anastasio 2015;Ghio and Devlin 2001;Huang et al., 2003;Schlesinger et al., 2006). PM with an J o u r n a l P r e -p r o o f Journal Pre-proof aerodynamic diameter of ≤10µm (PM 10 ) can be inhaled by humans and deposited into the respiratory tract. Fine particles ≤2.5µm (PM 2.5 ) -a subset of PM 10 -show a much higher fraction of deposition in the alveoli (ICRP, 1994). Another subset, ultrafine particles (≤100 nm in diameter, also referred to as UFP or PM 0.1 ), contribute negligibly to the particle mass but significantly to particle number concentrations (PNC) and are thus measured by their counts (Kumar et al. 2010(Kumar et al. , 2014. UFPs are so small in their size that they can penetrate deep into the lungs and translocate to the other parts of the body (HEI 2013). It is hypothesised that UFPs may have a greater potential for adverse health impacts compared with larger-sized particles (Kumar et al. 2014;WHO, 2013). Current ambient air quality regulations do not cover UFPs. They are also identified as one of the main contributing factors, after the fine PM fractions, responsible for adverse health effects observed at typical outdoor levels (Stone et al., 2017). UFPs, including the organic material attached to them, can cause acute respiratory and cardiovascular effects (Ahmed et al. 2019;Li et al. 2016), due to their large surface area, ability to generate reactive oxygen species (ROS), surface reactivity and their chemical composition of semi-volatile organic compounds. These effects induce potential cytotoxicity and inflammation in cells and need to be studied for their epidemiological effects on respiratory cells. Apart from UFPs, their larger counterparts such as PM 2.5 can also elicit varied biological responses, such as inflammatory responses causing COPD (chronic obstructive pulmonary disease) and other pulmonary infections. Furthermore, UFPs cause reactive oxidative stress on cells, effects on sensory receptors and neurons, and various other biological responses (Kreyling et al., 2006; Kodvanti ventilation, insulation and other indoor activities (Massey et al., 2016;Habil et al., 2015;Tan et al., 2013;Wayne and Roberts, 1998) affect the level of personal exposure in an indoor microenvironment. Hence, we consider indoor microenvironment as part of this review.
Furthermore, microenvironments close to traffic emissions, such as traffic intersections and roadside areas are often covered, since these microenvironments are exposed to elevated levels of PM 2.5 and UFPs (O¨zkaynak et al., 2008;Fujita et al., 2007). The PM concentrations in trafficrelated microenvironments were found to be higher, due to the contribution from the exhaust (tailpipe) and non-exhaust (e.g., tyre and brake wear, re-suspension of dust) emissions of vehicles; (Pant and Harrison, 2013;Kumar et al., 2013). The Non-exhaust emissions mainly generate coarse particles  ) as opposed to the exhaust emissions which contribute largely to fine particles (PM 2.5 ) (Kam et al., 2012;Thorpe et al., 2007). Conversley, emissions arising from the tailpipe of on-road vehicles are the major contributors to UFP emissions in cities (Zhu et al., 2002;Fujitani et al., 2012;Kerminen et al., 2007;Wang et al., 2008;Weijers et al., 2004).
Parks have also been considered since urban vegetation reduces PM concentrations (Paoletti et al., 2011;Brack, 2002;Tiwari and Kumar, 2020), increases lung function (Sinharay et al. 2018) and benefits mental health and wellbeing during any physical activity (Kumar et al., 2019). This microenvironment is important for various community and physical activities such as running, walking and exercise (Cohen et al., 2007;Rung et al., 2005;Ulrich et al., 1991) and people of all ages spend considerable amounts of time in parks with the confidence of less air pollution (Lin et al., 2014). A recent review published on selecting appropriate green infrastructure (GI) to mitigate air pollution (Barwise and Kumar, 2020) suggests a predominant role of GIs in reducing roadside exposure to pollution. The magnitude of air pollution reduction is demonstrated by field investigations and computer simulations using airflow models (Tiwari and Kumar 2020;Tiwari et al., 2019). Hence, PM concentration levels, size distribution and chemical composition in these microenvironments are of growing interest. We consider the roadside microenvironments J o u r n a l P r e -p r o o f Journal Pre-proof adjacent to GI in this review since these potentially mitigate the impacts of air pollution exposure (Kumar et al., 2019;Abhijith and Kumar 2019;Abhijith et al., 2017;Ottosen and Kumar, 2020).

3.1
Fine particles Table S1 shows a summary of PM 2.5 concentrations measured in various studies in different microenvironments. The indoor environment represents different atmospheres, such as homes, office spaces, schools, local shops, malls and food courts (Tan et al., 2020;Kelly and Fussell 2019). In general, the source of indoor particulates is associated with anthropogenic activities, intended usage of the space, type of activity being carried out and their air-exchange rate by natural or mechanical ventilation via HVAC systems. Among these, indoor PM concentrations are governed primarily by factors such as cooking activities (especially involving charcoal) (Lim et al., 2012), smoking (Gurung et al., 2016;Lai et al., 2004), office laser printers (Tang et al., 2012), resuspension of particles following human use of the space (Nazaroff, 2004) and poor ventilation conditions (Matic et al., 2017;Chatzidiakou et al., 2015). The levels of PM concentration might increase or decrease according to the intensity of these activities. A detailed literature review suggests that indoor concentrations are mostly measured using cascade impactors and, in some studies, personal monitors, continuous monitoring devices such as Harvard compact cascade impactor, Sioutas personal samplers, GRIMM and any kind of wireless sensor system (Table S1) and passive sampling technique (Canha et al., 2014). The average level of indoor PM 2.5 concentrations, based on the studies summarised in Table S1, were found to be 18.1±6.9 μgm -3 (Table S1). Contrarily, the concentrations of PM at traffic-related microenvironments, such as roadside and traffic intersections, depends on the density and the type of vehicles, time of sampling (i.e., peak versus off-peak hours) (Piotrowicz and Polednik, 2019), vehicle idling times and traffic signals (Wang et al., 2008). The major contributors of PM from traffic-related microenvironments are primarily from passenger cars (Iijima et al., 2007), heavy-duty vehicles (HDVs), light-duty vehicles (LDVs) (Robert et al., 2007a(Robert et al., , 2007b and buses J o u r n a l P r e -p r o o f (Rivas et al., 2017). These traffic-related concentrations are measured by a range of instruments such as high-volume samplers (Cui et al., 2016;Lawrence et al., 2013), fine particle samplers (Jain et al., 2020) or mini volume samplers (Tobler et al., 2020). Based on the studies summarised in Table S1, their concentration levels can be 27.5±19.8 μgm -3 and 33.1±5.3 μgm -3 for roadside and traffic intersections, respectively ( Figure 1). Furthermore, park microenvironments located at a moderate distance from the urban activities are also impacted from local sources such as vehicular combustion, resuspension and various other cooking activities in an urban area (Harrison et al., 2001) and they were found to possess PM concentration levels of 16±4 μgm -3 ( Figure 1). These park microenvironments were found to be using dichotomous sequential samplers, TEOM based samplers and PM 2.5 cyclone samplers (Table S1). Finally, the GI (foliated tree line and roadside hedges) microenvironments, which are an emerging architecture to reduce air pollution, involved continuous monitoring devices, such as GRIMM aerosol mass spectrometers, Dust Trak monitors and other continuous monitoring devices (Table S1). These GIs were found to reduce air pollutant exposure by up to 63% and respiratory deposited dose by 36% (Abhijith and Kumar 2019; Al-dabbous and Kumar, 2014).
Based on the studies summarised in Table S1, typical PM 2.5 concentrations around the GIs microenvironments were found to range from 18±17 μgm -3 ( Figure 1). In general, the PM 2.5 concentrations are measured primarily by using PM samplers which are equipped with a filtering medium, which serves both for gravimetric and other related chemical analysis. Also, the aforementioned concentration values are average values and they may vary according to various factors with respect to different microenvironments. However, this detailed collection of average concentrations build an understanding of the PM 2.5 levels in the respective microenvironments to select the appropriate sampler (discussed in Section 4) and provide guidance for how long is needed to collect PM for toxicological studies (Section 7).

Ultrafine particles
J o u r n a l P r e -p r o o f UFPs originate mainly from combustion activities (Costabile et al., 2019;Kumar et al., 2010;Morawska et al., 2013;Settimo et al., 2020) and are measured by PNC due to their negligible mass compared with larger diameter particles (Heal et al., 2012;Kumar et al., 2014;Slezakova et al., 2019). Table S2 presents a review of UFP concentrations in the different microenvironments. The factors affecting the typical concentrations are nearly identical for UFPs in most microenvironments (Section 3). The indoor concentration of UFPs have been measured using instruments such as the condensation particle counters (CPCs; particle size range 0.006-3 µm). Based on the studies summarised in Table S2, their concentrations were found to be 1.6±1.4 ×10 4 cm -3 ( Figure 1). The traffic-related emissions for roadside and traffic intersections use scanning mobility particle sizers (SMPS; 0.005-3 µm) and condensation particle counters (CPCs; 0.006-3 µm) for UFPs measurements (Table S2) with concentration levels of 2.4±0.9 ×10 4 cm -3 and 2.6±0.7 ×10 4 cm -3 ( Figure 1) for sites in and around London (Table S2). Finally, the park microenvironments were found to possess concentration levels of 6.8±1.3 ×10 3 cm -3 ( Figure 1) which are measured mostly by using instruments such as SMPS (Table S2). The concentration levels prevailing in the roadside microenvironments with GIs use fast mobility particle sizer (FMPS;), differential mobility spectrometers (DMS; 0.005-2.5 µm), GRIMM (0.25-32 μm) and PTrak model instruments (0.02-1 µm) (Table S2) for UFP measurements and their mean concentrations levels are 1.1±0.5 ×10 4 cm -3 (behind hedges) ( Figure 1). The UFP measurements were mostly carried out using instruments such as the SMPS in these different microenvironments. Toxicological study assays require reasonable quantities of particle mass (i.e., in milligrams; Section 6), which is remarkably lower for UFPs compared with PM 2.5 . Thus The PM collection strategies can vary according to the type of microenvironments. For instance, high volume samplers with heavier pumps are challenging to use indoors (due to the unsuitability of the pump noise in an office space or home environment). Similarly, the use of mini-volume samplers with lower pump capacities can have disadvantages in traffic-related or polluted outdoor environments due to the higher PM concentrations, potentially requiring more frequent filtering medium changes. Hence, the following section discusses the available methods and instrumentation for the collection of PM 2.5 and UFPs mass.

Fine particles
The mass of PM 2.5 particles for physicochemical and toxicity assessments is usually collected by three different types of air samplers, as summarised as in Table 2. They fall under the following categories: (i) high volume samplers, which mostly have a single impactor corresponding to the particle size of interest (e.g., PM 2.5 samplers, fine particulate samplers, and chemical speciation samplers); (ii) mini volume samplers (with two impactors of particle sizes PM 10 and PM 2.5 ); and (iii) Cascade impactors, which have multiple impactors to cover varied particle sizes (e.g., Harvard compact cascade impactors, HCCI). These samplers, in general, collect the PM 2.5 particles in quartz (Islam et al., 2019a;Hong et al., 2017) and PTFE filters (Jan et al., 2020), whereas the HCCIs use a polyurethane foam (PUF) impaction substrates and Teflon filters which can be subjected to chemical characterisation (Demokritou et al., 2004).
There are challenges in using impaction-based approaches for size-fractionated PM sampling.
Conventional impaction-based samplers use impaction substrates coated with oil and other adhesives to minimise particle bouncing. However, it is not possible to use such PM samplers to collect mg-level PM for toxicological studies due to the minimal loading capacity of hard impaction substrates (Lee et al., 2005). Demokritou et al. 2002 developed a series of impactionbased samplers using a "dry" porous substrate (PUF) which enables the collection of huge quantities of particles with no particle bounce. Such PUF-based impaction samplers have varied J o u r n a l P r e -p r o o f flow rates from 5 to 1000 LPM (Chang et al., 2013;Sarnat et al., 2003;Demokritou et al., 2002).
Another challenge associated with impaction-based PM samplers is the efficient extraction from filters of particles for toxicological studies without altering their physico-chemical properties. Pal et al. (2015) developed a method which enables the extraction of PM from substrates without any such alteration.
Among these different types of samplers, high volume samplers are the most commonly used samplers in PM 2.5 collection for toxicological studies ( Table 2). These are categorised as highvolume samplers due to their high flow rates, ranging from 16.7 LPM to 30 LPM that is usually denoted as 16.7 to 30 LPM, respectively (Jan et al., 2020;Deng et al., 2013;Fuentes-mattei et al., 2010;Oh et al., 2011;Billet et al., 2007;Deng et al., 2006;Dagher et al., 2005;Demokritou et al., 2004;Hsiao et al., 2000). These samplers are highly efficient in collecting large quantities of particles in a short period of time with similar power requirements as domestic appliances.
The high volume samplers do not employ multiple stages of impactors, as cascade impactors do, rather, they possess a single impactor with the particle size of interest (PM 2.5 impactor) which retains particles of ≤2.5μm diameter on a filter holder equipped with either Teflon or quartz filters (Islam et al., 2019a). These types of impactors are generally used in outdoor environments, usually for collecting particles from traffic-related (such as roadside, traffic intersections) microenvironments (  (Pal et al., 2015). Additionally, these samplers can be used both in indoor and outdoor environments due to their dry vacuum pump which generates less noise, while sampling a high-volume of air (900 LPM). Apart from the high-volume samplers and J o u r n a l P r e -p r o o f cascade impactors, certain studies use mini volume or low volume samplers with a flow rate of 5 LPM. These types of samplers are used in sites with very high concentrations of particles, where a study needs to collect particles continuously for a limited period (for months' time) during any form of traditional activity such as a winter heating period (Niu et al., 2017). Also, these mini volume samplers are well suited for an indoor environment, since they produce less noise in an office space and the sampling can be carried out without creating undue disruption to the office environment. However, due to the low flow rate, they tend to increase the sampling duration excessively.
Hence, for the collection of PM 2.5 particles, the selection of samplers plays a key role in managing the PM collection duration, with due consideration given to noise levels and concentration levels. For instance, a high volume sampler will be disruptive to use in an indoor environment to reduce the sampling time, whereas a mini volume sampler is ineffective in an outdoor environment with lower concentration levels. Thus, an informed selection of sampler is required, which satisfies criteria such as lower noise when used indoors and high particle collection efficiency with an adequate flow rate to collect the required amount of particles for physico-chemical and toxicological studies.

Ultrafine particles
Toxicological studies assaying UFPs could require a large amount of particle mass, usually up to milligram (mg) quantities (Demokritou et al., 2002). In general, UFPs for chemical characterisation and toxicological studies are collected using cascade impactors, with varied flow rates ( Table 3). The samplers involving the collection of UFPs generally operate at a flow rate of  (Borgie et al., 2016). The HCCI plays a prime role in UFP collection for toxicological studies since this sampler, unlike the high-volume samplers, collects particles at all the impaction stages (Setyawati et al., 2020;Pennanen et al., 2007;Demokritou et al., 2002). Apart from the PUFs, the sampler is equipped with a final back up filter (which is usually Teflon) for collecting the particles ≤0.1μm diameter. The HCCI is most suitable to collect particles in all microenvironments due to its high flow rate (30 LPM)), easy installation, more compact design and comparatively low noise levels due to the dry vacuum pump. In addition, impactors such as Micro-Orifice Uniform Deposit Impactors (MOUDI; MSP corporation), are used extensively to collect PM mass for toxicological studies of UFPs (Table   3). These impactors generate a sampling flow rate of 10 to 30 LPM (Wang et al., 2013) and they use both Teflon and Quartz filters according to the focus of their studies on PM (either physicochemical characterisation or toxicological studies) (Shirmohammadi et al., 2016).

Physio-chemical characterisation of fine and ultrafine particulate matter
Physicochemical characterisation of PM is essential since the elemental composition, ionic composition and organic matter content will play a vital role in influencing the toxicological responses. In addition, the PM composition and hence bio reactivity will vary according to the type of the microenvironment they are collected from.
Although initial measurements of weight, number, concentration and size distribution of particles are carried out during sample collection (Sections 4.1 and 4.2), the measurements are often insufficient to characterise the full complexity of PM that is a mixture of various organic and inorganic compounds, which contribute different ways to the overall toxicity (Park et al., 2018).
The constituents exhibit a wide range of sizes, morphologies and chemistries (bulk and surface), which are averaged with bulk methods. J o u r n a l P r e -p r o o f

Destructive versus non-destructive
Several techniques used to characterise the physical and chemical characteristics of PM require an amount of material for testing that can be difficult to collect practically. The minimal amount of material required and the information generated by a given method is often set by the physical principles of the instrumentation (e.g. the diffraction limit) or engineering constraints such as the detector sensitivity. Some methods are intrinsically destructive, e.g. ICP-MS and TGA and other theoretically non-destructive methods may cause some mass loss due to the sample preparation required or limit further applications, e.g. samples for TEM must be mounted on grids. Therefore, the order in which the PM is characterised must start with the nondestructive methods first before proceeding with destructive techniques such as ICP-MS for bulk analysis of trace metal content. In Table 4, a subset of commonly used characterisation techniques and required PM amounts are listed.

Selection of characterisation techniques
The techniques selected for analysis depend on the mass of PM collected and the composition and mass of each component of PM within the mixture. For example, the low detection limit of techniques such as ICP-MS, TOR and TD-GC/MS analysis makes it feasible to characterise very small amounts of trace metals or gases in PM. It is, however, difficult to estimate the total mass needed for each technique, as PM contains a mixture of elements. In practice, 1 cm 2 punch-outs of a filter could be used for these analyses. However, these methods are destructive and provide information averaged from the whole sample. On the other hand, pyrolysis properties of PM and their gaseous decomposition products, e.g. to give information about the % of organic and metals in a sample, are both destructive and require high amounts of PM (Lapuerta et al., 2007). In contrast to the methods described above, electron microscopy needs very small mass, has high spatial and energy resolutions for chemical analysis and mapping and provides information at the length scale of the individual particles within the mixture (Liati et al., 2016). However, the sampling volumes are very small, so these methods need to be correlated with bulk techniques to ensure that the collected data are representative of the whole sample collected.

Analysis of PM on, and extracted from, the filter to more accurately understand the cellular response
Some techniques such as SEM, Raman spectroscopy and XRF (Satsangi and Yadav 2014;Hamdan et al. 2018) can characterise PM directly on the collection filters, so the PM is in its near-native state. Other on-filter methods include SEM-EDX, XRF, micro-XRD, which are used to image the size distribution, chemistry and phase of the particles. However, in practice, the filter will also influence the PM, principally by altering its aggregation state, so that it is no longer truly representative of ambient PM. The use of impactors for PM collection could also change the morphology of the PM during the high-speed collision processes. Many techniques, e.g. ICP-MS requires that the PM is removed from the collection filter, often using sonication, and is suspended in a liquid such as methanol mounted on a substrate such as a TEM grid.
Extraction by sonication into organic solvents will change the properties of the PM, i.e. the original size distribution of the sampled PM and its chemistry. Extraction into aqueous solvents could also dissolve some of the inorganic fraction due to ion redistribution. Other issues with particle extraction include contamination from the filters, i.e. generation of broken PTFE fibres during sonication, preferential extraction of certain components of the PM (depending on the extraction medium is used), and contamination from glassware used for the extractions which J o u r n a l P r e -p r o o f will also alter the outcomes of cell viability assays. When the particles are transferred to cell culture media, they also need to be freeze-dried by lyophilization, and transferred into cell culture media which can also change the chemistry and aggregation state of the particles, so the PM must also be characterised in this media (Theodorou et al., 2013).
Depending on the separation method, filter type used and the properties of collected material, recovered amounts of PM can vary significantly. Some constituents might be lost or retained within the filter during separation. Gravimetric analysis of total PM 2.5 mass has shown that extractions are around 80% efficient. However, specific PM constituents like metals and organics have much lower extraction efficiencies (on average, 47±22% and 25±14%, respectively) (Roper et al. 2015). There is no perfect method for PM extraction, and available techniques need to be adjusted for the optimal extraction of constituents of interest.
These artefacts generated during the extraction of PM from filters make it challenging to relate the toxicological outcomes of cell culture experiments to the health effects of exposure to aerosolised PM direct from the air. Another issue is that collection/extraction-induced aggregation of the PM will change sedimentation and diffusion of the PM in the cell culture media as they settle on the cells and also cellular uptake which will alter the dose of PM delivered extracellularly and to the inside of the cells affecting their bioreactivity (Hinderliter et al., 2010). Electrostatic filters have also been used to collect PM, but these also induce aggregation, and there is evidence that these filters produce ozone as a by-product due to corona discharge (Rim et al., 2013), which is damaging to cellular health (Wiegman et al., 2014). In practice, it is not possible to fully overcome these issues to yield PM which are fully representative of ambient PM, but by fully characterising the PM at each stage of collection/extraction into the in vitro system, more accurate relationships can be drawn between PM properties and the cellular response. Thus, toxicological studies can require substantial amounts of mass (up to 10 mg) to assay the effects of the PM on cellular health, including assays of oxidative stress, inflammation, cell death and mitochondrial activity. Another issue, discussed in the previous section (section 5), is that the PM has to be extracted off the filters and dispersed into solvents such as methanol, which could alter their physico-chemistry. Although alternative solvent-free methods, such as electrostatic collectors can be used to overcome this issue (Ning et al. 2008;Sillanpää et al. 2008). The identification of cellular modifications caused by PM is essential to understand its harmful effects and related respiratory and cardiovascular diseases. In this section, we describe and list the assays and techniques used to evaluate the toxicity of PM 2.5 and PM 0.1 to illustrate how the selection of toxicological endpoints being assayed determine the mass of PM needed.

Optimisation in particle collection
The knowledge of the average concentrations of the particles in the different microenvironments (Section 4) can enable the researchers to plan the sampling duration to collect the required amount of particle mass (Section 5) and their subsequent toxicological assessments (Section 6). These studies involving particle mass collection for toxicity and physico-chemical characterisation assessments demonstrate that the particles collection rate highly depends on the instrument's flow rate, site morphologies and the prevailing atmospheric conditions. The following section discusses, firstly, the necessary ways to collect the particles effectively. Secondly, it provides specific recommendations to increase the likelihood of obtaining the required quantities of particles.
J o u r n a l P r e -p r o o f

Fine particles
The toxicological studies of PM 2.5 discussed in the previous sections reveal that these studies require particles in mg quantities. In general, the PM 2.5 particles were collected mainly using instruments with a medium volume flow rate of 16.7 LPM (Table 2). However, some studies use instruments with a high volume flow rate ranging from 100 to 400 LPM ( Table 2).
The average concentrations extracted from the previous literature can be combined with the instrument flow rate to plan the field campaign.
Furthermore, the maximum sample mass which can be collected on a single filter paper (be it PTFE or glass fibre or quartz or polycarbonate) should also be calculated; this data can work as a guide in changing the filter paper effectively, especially for continuous outdoor measurements.
However, preliminary monitoring on the particle mass concentration has to be carried out (using optical instruments) as this helps in planning the duration to change the collecting medium (filter paper) which is highly dependent on the concentration in respective microenvironments.
The toxicological assessments are carried out in different doses such as 5, 10, 20, 50, 100 and 200 μgmL -1 (Islam et al., 2019b)) for estimating the exposure to different compounds and also for multiple experiments. These dosage proportions may vary according to the interest of the researcher and their objectives. However, commonly the dosage levels start from 1 to 5 μgmL -1 and extend up to 200 to 300 μgmL -1 . Hence, as noted above, there is a need to obtain particle mass in mg quantities for these assessments. Therefore, by considering the dosage levels mentioned in a wide variety of literature, it is evident that an average of 1000 μg (or 1 mg) of particle mass is at least required for conducting a toxicological study. The time required for collecting the mentioned 1000 μg for each microenvironment has been calculated using the typical concentrations obtained ( Figure 2) from various sources and have been tabulated in Table   4.

J o u r n a l P r e -p r o o f
The schematic representation of the time taken by various samplers (Figure 3) demonstrates that the samplers with a high flow rate ranging from 400 to 1050 LPM require the least time to collect the targeted PM 2.5 particle mass. In contrast, the fine particulate sampler (16.6 LPM) and Harvard impactors (30 LPM) require longer to collect the targeted mass. This is because the samplers with higher flow rate possess a larger pump capacity resulting in reduced sampling duration. However, these samplers with larger pump capacity are not suited for all types of microenvironments. For instance, samplers with larger pumps are usually not suitable for an indoor environment due to factors such as increased noise levels (noise pollution) creating an unpleasant ambience, vibrations on the floor, potential deterioration of air quality (if oil pumps are used), usage of space, and difficulty in transportation, etc. However, these factors need not be considered when high volume samplers are used in a roadside or traffic intersections microenvironment (generally outdoors). The fine particulate samplers and Harvard impactors possessing a lower flow rate compared with high volume samplers are much better suited for indoor conditions, especially the Harvard impactors since these impactors were mounted on a tripod and possess a higher pump capacity with less noise compared to fine particulate samplers.
Also, the Harvard impactors collect a significant amount of particle mass when used outdoors (traffic intersections or roadside). However, the high-volume samplers will collect more mass, compared to these samplers.

Ultrafine particles
The mass concentrations of UFPs are significantly low when compared with PM 2.5 .
However, the mass of UFPs is collected using methodologies similar to PM 2.5 mass collection.
The UFPs are generally collected using impaction sampling techniques coupled with high volume samplers ranging from 400 to 1100 LPM. However, samplers with flow rates ranging from 9 to 70 LPM are used, but mostly in indoor environments. The sampling time will generally be higher for UFPs due to their lower concentration, and high-volume samplers are usually J o u r n a l P r e -p r o o f preferred to reduce the sampling duration while collecting the required amount of particle mass.
In general, the toxicological assessments require samples of large quantities (in mg), for multiple experimental investigations to study their exposure efficiently. Similarly, the physico-chemical characterisation of UFPs requires an average of at least 100 μg of particle mass for estimating elements, ions and carbon content present. Therefore, it can be said that an average of at least 1000 μg and 100 μg of particle mass will be required to conduct toxicological and physicochemical assessments, respectively. This particle mass quantity might increase or decrease according to the research requirements and study objectives.
The particle mass collection can be optimised by first calculating the sampling time required to collect the required particle mass. This can be calculated by conducting preliminary measurements using handheld optical instruments showing the mass concentrations of the UFPs.
These mass concentrations can be used in calculating the sampling time for each of the commonly used samplers (Table 4). The obtained sampling time (Table 6)  Other samplers (Pan et al., 2019;Verreault et al., 2008).
Among these sampling techniques, impactors, filters, impingers and cyclone samplers, aid in collecting airborne viruses effectively (Haig et al., 2016;Reponen et al., 2011). The impactor sampling technique uses Andersen samplers, which works on the principle of inertial impaction and traps particles of specific cut off diameter from 0.65 to 7 µm (Tseng and Li 2006), with air passing through the six stages of impactor at a flow rate of 28.3 LPM on to a petri dish consisting of liquid agar medium (White et al., 2020). The slit samplers draw particles through narrow holes or slits onto a petri dish which holds a culture medium (Booth et al., 2005). In the case of sampling by liquid impactors, all-glass impinger (AGI) samplers are used, where the sampler draws the air stream inside the liquid collection medium through diffusion (Hermann et al., 2006).
The filter sampling mechanisms are used widely when the particle sizes of airborne viruses are diameter. The virus collection efficiency is found to be higher for RNA viruses (Zhao et al., 2014). However, these samplers use gelatin filters which, as described earlier, are highly RH dependent; when exposed to higher HR it dissolves the sample but dessicates the sample when exposed to lower RH (Verreault et al., 2008). Thus, at higher RH, the gelatin dissolves the J o u r n a l P r e -p r o o f sample and at lower RH it desiccates the sample. Further information on a wide array of biosamplers for the sampling and detection of viruses, especially coronaviruses, can be found elsewhere (Rahmani et al., 2020, Pan et al., 2019Verreault et al., 2008  Airborne pathogen (especially viruses) collection are found to be efficient when a liquid medium is used with a high flow rate of air (ranging from 100 to 300 LPM). A collection medium using filters is unfeasible because of filter desiccation (in case of PTFE, polycarbonate and glass fibre filters) whereby the microorganisms get dehydrated during this process and cause a greater loss in recovering the viruses. Gelatin filters are, however, efficient in collecting these viruses provided the measurements are taken in an RH controlled (which is mostly indoors) environment. The impaction and impinger sampling technique is found to be less efficient compared to the cyclone sampling technique (using centrifugal forces), where the airborne particles are collected on the sides of the centrifugal tube with the collection medium inside.
These types of samplers operate at a high flow rate and the collection medium used in most of these samplers can be readily used for further gene analysis processes such as PCR for virus detection. Hence, the samplers with cyclone sampling techniques (Coriolis μ air sampler and NIOSH BC 251 samplers) using centrifugal forces are effective in collecting airborne viruses in any kind of environment (both indoors and outdoors).

Optimisation of particle mass sampling via modelling
The strategy of particle mass sampling can be optimised by predicting the air pollution concentrations by predictive models, which can enable to identify the pollution hotspots (places with increased pollution levels). These hotspots can be used as suitable sampling locations since there is a higher probability to obtain large amounts of particles in a significant time frame.
These predictive models In terms of accuracy computed as a mean squared error with respect to a control variable, this approach shows a reduction of the error which goes from 8.90✕10 -1 for locations randomly chosen to 3.35✕10 -02 for the optimal locations. A similar approach has been implemented in Tajnafoi et al. (2020) where an ML model is trained using data from a CFD simulation of air pollution in a room within the Clarence Centre building located in Elephant and Castle in London, UK. The simulated training dataset consists of 10,000 grid points, which represent potential instrument locations distributed in a 3D space. The number of selected locations, in this case, is 7. In terms of accuracy, this approach showed a reduction of the error, which goes from 1.70✕10 -1 for the seven random locations to 5.00✕10 -4 for the optimal locations. The accuracy of a modelling technique is vital since it provides an estimation of the misfit between the measures collected from the selected location and the estimates provided by a forecasting model.
Reductions of the errors are mandatory for effective prediction.

J o u r n a l P r e -p r o o f
In the last decade, these methods have been accepted as powerful tools in forecasting the outcomes. However, these DA methods quantify certain uncertainties, and these can be customdeveloped to minimize the discrepancy between numerical results and observations, where forecast and observations are assumed as the sources of information having errors that will be Hence, the DA models used for the sensor placement can be adapted to estimate the hotspots for particle mass collection by customising the data assimilations to reduce the errors in estimating those hotspots. In the case of particle mass collection, the number of placement could be reduced to a few locations. These technologies provide a list of the coordinates of the best locations in decreasing order with respect to the impact they have in terms of mutual information. In case the possible location is only one, the best location to consider is the first in the list provided by the technologies. At the moment, these approaches have not been used for selecting optimal sites for mass collections. However, these technologies are general and can be used if a predictive model can be implemented to simulate the particle mass dispersion.
The modelling techniques and overall approach described above have not been used for the purpose of an effective mass collection yet. This is primarily due to the lack of access to model results that require specific expertise, resources and knowledge base. However, it can be a novel approach for selecting effective locations to collect particle mass on filters when relevant information is available.

Summary, Conclusions and Future Work
This article focused on the methodologies involved in PM 2.5 and UFP mass collection to study their physicochemical characterisation and toxicological effects. We provided a critical The key conclusions drawn from this study are as follows: • Over the last two decades, it has become increasingly necessary to perform toxicological assessments of PM 2.5 and UFP particles. Airborne PM exposure to these two size fractions has been demonstrated to significantly correlate with the occurrence of adverse cellular effects, cardiovascular and chronic respiratory diseases. The toxicity of PM is generally higher in small size fractions (such as PM 2.5 and UFPs), and a strong association exists between the physico-chemical characterisation and their toxicity effects. These studies require a relatively large amount of particle mass to analyse these biological mechanisms, and it is often difficult to obtain the targeted particle mass in a reasonable period. Limited studies have discussed the particle mass (which is an important aspect) collection effectively by selecting an appropriate type of instrument according to the characteristics of each microenvironment in urban areas; this review provided viable suggestions to use the appropriate type of instrument for the respective microenvironments.

J o u r n a l P r e -p r o o f
• We reviewed the typical concentrations in five different urban microenvironments (roadside; traffic intersections; indoors; parks; roadside behind GIs). As expected, the mass concentrations near to traffic-related activities such as roadsides and traffic intersections were found to be higher compared than those indoors where elevated mass concentrations were observed at study sites with poor ventilation and cooking activities.
The park and roadside sites behind GI showed a lower mass concentration, due to the presence of urban vegetation and roadside GI. These typical concentrations served as a viable input in calculating the sampling time required by each instrument for obtaining the targeted particle mass.
• The sampling instrument and its volumetric flow rate play an essential part in the sampling duration for collecting the targeted mass of particles for the analysis. The volumetric flow rate on these samplers can range from 16 to 1100 LPM for both PM 2.5 and UFP, and these samplers largely use either impactor or cyclone sampling techniques.
The particles collected in indoor environments generally use impaction sampling techniques (with flow rates 16 to 30 LPM) in a medium volumetric flow rate for collecting particles effectively without causing any increased noise levels or unpleasant conditions during sampling (in indoor environments). The impaction sampling technique, coupled with high volume suction pumps ranging from 400 to 1100 LPM, can be deployed in most of the outdoor microenvironments. Certain samplers, such as Harvard impactors, can be used for both indoor and outdoor environments since they have a significant flow rate and can collect particles in both fine and ultrafine size range. These types of samplers are as efficient as high-volume samplers and easy to operate and handle compared to high volume samplers.
• Though there are many methods available for physicochemical examination of PM, their use can be restricted by numerous factors. In particular, the amount of PM collected may J o u r n a l P r e -p r o o f limit viable options. When only limited amounts of PM are available, the experimental workflows to characterise the PM needs to be carefully planned. The PM samples need to be cycled through non-destructive experiments before attempting destructive approaches.
On-filter experiments might be favoured over experiments requiring PM extraction. In general, if more mass of PM can be collected, more comprehensive and accurate analysis can be performed.
• Different mechanisms such as inflammatory processes, oxidative stress and genotoxicity are involved in PM cytotoxicity. Fine and UFPs exhibit significant toxicological effects due to their particle diameter, which aids their entry into the airways and bloodstream and causing cellular effects. The cellular toxicity of PMs and their subsequent damaging effects in different organs is attributed to the type of cell line, culture conditions, and the particle size, composition and concentration. In short, higher the PM concentration and exposure time lead to highest toxicity, and it strictly depends on the absorbed toxic pollutants. The oxidative stress seems to be generated mostly by the metal fraction while the genotoxicity by the organic fraction. Higher levels of PM and different cell lines are used in cell toxicity studies compared to a real-life situation. Methodological differences hamper direct comparison of results between different studies. The PM 0.1 toxicity grade is correlated not only with the relative mass, as for bigger PM, but also with particle number and surface area. Due to the higher capacity to absorb organic pollutants, PM 0.1 result in more toxic effects compared to PM 2.5 , due to their small size and ability to produce damaging effects in the lung and the brain. More studies about its neurotoxicity and correlation with neurodegenerative diseases are needed. The identification of cellular modification caused by PM is essential to understand its harmful effects and correlated diseases. Bioassay and imaging techniques are essential tools employed to understand the cytotoxicity mechanism of PM.
J o u r n a l P r e -p r o o f • Toxicological assessments are carried out in different doses. Hence, they require an average of 1000 µg of particle mass in both the size fractions. Similarly, it is possible to say that the physico-chemical characterisation needs an average of 100 µg of particle mass (for elements, ions and carbon analysis). PM 2.5 mass was largely collected by six different types of samplers and the UFP mass by seven different samplers. The samplers with higher volumetric flow rates (1100 LPM) require very short duration compared to samplers with low volumetric flow rates (9 LPM). However, the high volume samplers cannot be used in all the urban microenvironments, due to difficulties in handling and operation. The Harvard impactors are easy to handle, equipped with a quieter pump and have a higher collection efficiency, demonstrating that it is a viable option to collect PM particles effectively.
• With respect to airborne pathogens, in particular viruses, collection conducted using samplers with cyclone (centrifugal forces) sampling techniques, has been proven to be effective. The samplers with cyclone sampling techniques use centrifugal forces, which cause a cyclone, by collecting the particles on the surface of the collection vessel in a medium such as DMEM. Thus, the cyclone sampling technique (Coriolis µ air samplers and NIOSH BC 251 samplers) with a suitable medium is viable for virus collection.
• Predictive modelling approaches using AI can be used in identifying locations for effective measurements of the particle mass. In particular, data assimilation and machine learning techniques can help to identify the potential hotspots for effective particle mass collection. The effectiveness of these technologies has been mathematically proven and tested for some test cases (e.g. measurements of CO 2 ). These technologies have not been used for effective mass collection yet. However, there is potential for these technologies to aid particle mass collection.

J o u r n a l P r e -p r o o f
This review identified the different PM collection methodologies and recommended appropriate sampling strategies for various microenvironments for effective particle collection. These microenvironments will not possess similar characteristics since some will have higher concentration due to proximity to emission sources as opposed to others with very low concentrations. The morphology of certain environments facilitates the use of samplers with high flow rate and certain indoor microenvironments do not favour using these types of instruments; hence a low flow rate instrument is used, and the sampling duration extends significantly, not to mention the particle mass collection will also be affected. Also, monitoring stations, sporadic unattended measurements are not feasible in all the microenvironments due to vandalism or any other local difficulties. Hence, there is a need for a sampler with high flow rate (as high volume samplers) as well as portable (as low volume samplers) with easy handling and good collection efficiency, which will facilitate the user to utilize in all kinds of microenvironments. Usage of these kinds of appropriate samplers supports in conducting a more comprehensive and accurate analysis of physico-chemical characterisation and toxicological effects of PM. Prior to conducting any sampling activity, the morphology of sampling environments and their prevailing average ambient concentration can be studied by pilot investigations using handheld monitors.
The duration for collecting the targeted particle mass can be calculated using these pilot investigations, which in turn will help in finding the appropriate sampler to complete the sampling in a significant time frame. Future investigations should focus on using predictive modelling techniques such as data assimilation and machine learning, whose effectiveness has been proved mathematically and can be used for identifying the potential hotspots for effective particle mass collection.

9.
This work has been supported by the EPSRC funded project, Health assessment across biological length scales for personal pollution exposure and its mitigation (INHALE; Grant No. EP/T003189/1).  Table S1 for (a) PM 2.5 , showing a higher deviation in indoor (n = 9) concentrations, followed by roadside (n = 11), traffic intersections (n = 10), road side with GIs (n = 4) and parks (n = 5). (b) UFPs from Table S2, showing a higher deviation in traffic intersections (n = 7), followed by roadside (n = 8) indoor (n = 9), parks (n = 9) and roadside behind GIs (n = 5).  Tables   Table 1. Summary of existing reviews on the physicochemical and toxicological studies on fine and ultrafine particles.

Study focus and topic areas covered Author (year)
• Toxicological effects of PM 2.5 to understand the molecular pathways generating oxidative stress in particles and their transmission to the target organs. Leni et al. (2020) • Summarised the association of PM with pediatric asthma and reviewed their fundamental molecular mechanisms.  • Toxicological studies of PM obtained from whole diesel exhaust and particle filtered exhaust involving animal and human exposure studies.
• The physico-chemical composition, exposure (in vivo) and toxicological (in vitro) studies on PM from underground railway emissions of PM and its effects on human health.

Loxham and
Nieuwenhuijsen (2019) • Summarized the time series and cohort study on PM 2.5 on their effects of cardiovascular and chronic respiratory effects and other health endpoints.
Yang et al. (2019) • Summarised the in vivo/vitro studies on the immunotoxic effects of PM 2.5 exposure to pulmonary cellular effectors.
Wei and Tang (2018) • Inflammatory health effects from various types of PM sources and different environments and their pathogenesis of diseases. Wu et al. (2018) • Collected data for studies on the divergent responses to PM exposure and the effects on different types of cell deaths and their assessment in PM-exposed models (in vivo). Peixoto et al. (2017) • Toxicological studies of PM 2.5 (both in vitro and in vivo) and its biological response in humans.
Mohammed et al.
• Toxicological effects of PM comparing reactive oxygen species (ROS) production, inflammation and oxidatively damaged DNA in different experimental systems. Møller et al. (2014) • Methods to evaluate the various toxicological effects of PM on the respiratory, cardiovascular, and nervous systems using in vivo and in vitro experimental models. Nemmar et al. (2013) • Toxicological effects of PM from various sources, especially their components and bioreactivity of PM from traffic emissions.

Kelly and Fussell
• Studies relating to different PM sources and their effects on human health. Stanek et al. (2011) • Studies on ambient PM and the effects of PM on oxidative stress, activation of antioxidant defences, inflammation and toxicity. Li et al. (2008) • Studies linking the in vitro toxicological effects and mechanisms of PM from traffic emissions, focusing on PM size fractions < PM 10 .
de Kok et al. (2006) J o u r n a l P r e -p r o o f J o u r n a l P r e -p r o o f J o u r n a l P r e -p r o o f Note: ICP-MS (Inductively coupled plasma mass spectrometry); IC (Ion chromatography); TOR (Thermal/optical reflectance); TGA (Thermogravimetric analyzer); TD-GC/MS (Thermal desorber combined with gas chromatography coupled to mass spectrometry); XRF (X-ray fluorescence); XRD (X-ray powder diffraction); SEM (Scanning electron microscope); TEM (Transmission electron microscope).
J o u r n a l P r e -p r o o f J o u r n a l P r e -p r o o f