Glycosylated gold nanoparticles in point of care diagnostics: from aggregation to lateral flow

Current point-of-care lateral flow immunoassays, such as the home pregnancy test, rely on proteins as detection units ( e.g


Introduction -point of care
The diagnosis of infectious diseases should be achieved by; determining infection sites, considering patient needs and gaining a microbiological diagnosis. In many cases, however, diagnosis is guided by symptomatic presentation rather than therapeutic approaches as many of these methods can take hours or days to produce a result. Observational approaches alone can in some cases reduce patient outcomes and the injudicious prescribing of broad-spectrum drugs which can contribute to antimicrobial resistance (AMR). 1 Alexander N. Baker Dr Alexander N. Baker is a research fellow in the Department of Chemistry at the University of Warwick. He completed his MChem (2007), and PhD (2022) in collaboration with Iceni Glycoscience at the University of Warwick. Alex's work, in the Gibson group, is focussed on organic and polymer synthesis to design medical diagnostics for low resource settings. His most recent work has developed diagnostic tools for the rapid detection of COVID-19 in collaboration with NHS partners and Iceni Glycoscience. Beyond his research, Alex is a founding fellow of the Warwick Institute of Engagement (WIE).

George W. Hawker-Bond
Dr George W. Hawker-Bond is an Academic Foundation (AFP) doctor at Oxford University Hospitals NHS Foundation Trust and the Oxford University Clinical Academic Graduate School. He obtained his MChem (2017) degree from the University of Warwick, before qualifying in Medicine (MBChB) from Warwick Medical School in 2021. George's main research interests include the interplay between Chemistry and Medicine -with a particular focus in neuroscience and psychiatry -and medical education. He has also published on the involvement of under-represented groups in clinical trials.
The first point-of-care (POC) diagnostic was developed in 1962 to measure blood glucose levels. 2 This was followed by the rapid pregnancy test circa 1969. 3,4 Both of these diagnostics revolutionised care: allowing for testing and diagnosis by or with the patient present. In 2011, point-of-care tests (POCT), or ''bedside testing'' devices were worth $15 billion USD with a projected increase of 4% per annum. 5 While there is no accepted definition for point-of-care testing, it can be considered to be rapid testing at, or near to, the point of need that is used to make a medical decision. [6][7][8][9] Current POCT devices include the aforementioned glucose biosensor, dipsticks and lateral flow devices (LFDs); 10 such as Malaria Rapid Diagnostic Tests (MRDTs) 11 and pregnancy tests, that separate samples through a solid phase. 5 Surveys of medical professionals, in relation to POCT, have indicated that sensitivity (4 90%) is considered the most important attribute, followed by a price of less than $20 USD and a short detection time. 12 This is summarised by the World Health Organisation (WHO) guidelines for the design of POCT in the ASSURED acronym (Fig. 1). 13 The cost-effectiveness of POCT, compared to traditional laboratory techniques, has been demonstrated in the UK National Health Service (NHS) Health Checks, an early detection tool for cardiovascular disease; 14 a disease which costs the NHS approximately d7 billion GBP per year. 15 The total expected costs (TEC) per completed Health Check was markedly reduced for POCT compared to laboratory techniques -d18 GBP per completed Health Check compared to d25 GBP, respectively, and attendance rates were higher for POCT. 14

Simone Dedola
Dr Simone Dedola is General R&D Manager at Iceni Glycoscience where he leads the chemistry and the assay development team. His specific expertise lies in the chemical synthesis of carbohydrates, total synthesis of glycoproteins, and the chemoenzymatic synthesis of glycoconjugates for vaccine and diagnostic programmes, assay development, in vitro diagnostics. Dr Dedola has experienced the journey from academia to SME, from academic high science to pragmatic SME R&D, he has been with the company from day one (4 6 years) and has a full appreciation of the company's activities, from concept and research through development to product prototyping.

Robert A. Field
It is equally important to consider the perspectives of patients and health care professionals (HCPs) to POCT. During an increase of Dengue fever cases in Singapore, POCT were found to be more convenient, and marginally better than non-POCT in improving the speed and accuracy of diagnoses. 16 The alternative to POC testing is (highly accurate) laboratory-based testing, which require trained staff, embedded health infrastructures and specialised facilities. This makes lab-based methods, considered standard in wealthier nations, cost-prohibitive in low-and middle-income countries. This is evidenced by the COVID-19 (caused by SARS-COV-2) pandemic. PCR-based approaches have been the primary tool for diagnosis but is not ideal for mass testing due to price, turnaround time and the need for centralised infrastructure. 17,18 It is estimated that laboratory-independent POCT for four common infections (syphilis, malaria, tuberculosis and bacterial pneumonia) could prevent more than 1.2 million deaths per annum in low-and middle-income countries. 4 Chromatographic lateral flow paperbased immunoassays, as highlighted by MRDTs and pregnancy tests, are of particular interest as POCT because of their low cost, speed of use, laboratory-independence and wide range of devices and analytes currently in use. 19 This review article introduces the potential of glycans (sugars/carbohydrates) to be deployed in rapid, lateral flow assays, compared to the currently used antibody-based LFDs. We first summarise the chemical basis of POC lateral flow immunoassay (LFIA) diagnostics that are based upon antibodies. We then provide a discussion on how a lateral flow assay has been deployed in the real world. Finally, we discuss glycosylated materials, in particular glycosylated gold nanoparticles. We focus on gold particles as these are the basis of commercial tests, however other material cores are highlighted. We cover the rapidly emerging literature showing the development of paper-based lateral flow glyco-assays, LFGA.

Lateral flow -an overview
Lateral flow devices (LFDs) typically employ antibodies as sensing components; these devices are often termed lateral flow (immuno) assays (LFIA), or ''sol particle immunoassays'' as originally termed by Leuvering et al. 20 LFIAs have been developed for a variety of targets, including; complement; 21 anabolic steroids; 22 prostate-specific antigen; 23 leprosy 24 and meningitis 25 in serum; evidence of kidney injury in urine; 26 V. cholerae in faeces; 27 and other marker proteins 28,29 or pathogens 30 in whole blood samples. 31 This ability to target a broad range of analytes has led to the use of lateral flow devices in policing illicit drug use, for example. 32,33 LFDs typically contain two components; a mobile phase and a stationary phase. When a sample is applied to the strip's sample pad, it is drawn by capillary forces caused by the wick, through the device ( Fig. 2A). Then the sample passes through the conjugate pad. If the sample contains the target analyte ( Fig. 2Aa), it will bind antibody-coated gold nanoparticles (or other coloured nanomaterials) within the conjugate pad ( Fig. 2A). The sample and nanoparticles (both bound and unbound) pass into the nitrocellulose membrane, where they move through two lines of the deposited material. One of these lines is a control line that containing antibodies specific to the primary antibody in the conjugate pad (Fig. 2Ad). Binding of the nanoparticles only to the control line indicates a successful test, but a negative sample (Fig. 2B). The test line contains antibodies specific to the analyte. A line will form if the analyte is ''sandwiched'' between antibodies immobilised in the test line and antibodies immobilised on the nanoparticles (Fig. 2Ac).
There are examples of multiplexed lateral flow tests. 34 This multiplexed approach allows for greater efficiency and greater diagnostic precision and may reduce overall cost when tests are combined into one device. An earlier form of the lateral flow test, called a flow-through assay, utilised an immobilised sample as a test line -these devices are often less sensitive than lateral flow devices. However, they are employed when

The components of lateral flow
The stationary phase The stationary (immobile) phase comprises a receptor bound to the surface, usually an antibody. Larger antibodies do not require an immobilisation agent, with immunoglobulin G having an absorption of 4 100 mg cm À2 on nitrocellulose; however, protein absorption can decrease with decreasing molecular weight. 36 Bovine serum albumin (BSA), or less commonly ovalbumin, are used as carriers/immobilisation agents for small molecules such as nucleotides and small proteins. 19,37 Less common immobilisation strategies include covalent epoxide-thiol and using fusion-proteins. 38 Strategies to immobilise polynucleotides have also been reported. 39,40 The low absorption of unconjugated antibodies was noted in polymeric systems by Aoyama et al. 41 To overcome this they produced high surface area microcone architectures on polycarbonate sheets. The increased surface area enabled high antibody immobilisation levels. This step was inspired by the use of plastic and paper microfluidic systems, 42 a rapidly growing technique that has been used to sense Ebola virus RNA, 43 Salmonella typhimurium 44 and glucose. 45 Studies with cellulose paper have also exploited biotin and the ability to covalently modify the cellulose surface for easy test line immobilisation. 46,47 Tanaka et al. and Nagatani et al. have utilised the localised surface plasmon resonance (LSPR) phenomenon of gold nanoparticles in the stationary phase to enhance the signal-to-noise ratio when sensing for human chorionic gonadotrophin (HCG) and prostate-specific antigen respectively. 23,48 This approach allowed for detection limits of 1 pg ml À1 of HCG, 48 the hormone detected in the standard pregnancy test.
Reverse immunoassays that utilise ''proteinticles'' (proteincoated nanoparticles) in the stationary phase have also been developed. 49 In these systems, antibodies against HIV and hepatitis A and B were tested in the mobile phase. However, the patient's immune system must mount an adaptive immune response for the test to be successful. Other tests, such as the Epstein-Barr Virus Antibody Test, also sense for an immune response to the pathogen. Although not POC, it does provide precedence for reverse immunoassay approaches. 50 The mobile phase Nanoparticles for biosensing have received significant attention due to their unique optical properties, making them ideal for sensing activities, 51 and their tunability in terms of morphology and composition. 52 Plasmonic, especially gold, particles are the most common signal generating units. Plasmonic nanoparticles are intensely coloured because their free electron density can couple with electromagnetic photons with a wavelength greater than the size of the nanoparticle. [53][54][55] The colour of a nanoparticle is not only dependent on its size but also its shape and the dielectric constant of the metal and medium it is in, as shown by the Fröhlich condition. [56][57][58] Only light resonant to electron oscillation can excite the localised surface plasmon and be absorbed by it; this absorption leads to localised surface plasmon resonance (LSPR), producing the distinctive colour of the nanoparticle system, which in the case of gold is from red to blue. Plasmonic nanoparticles have high molar attenuation (extinction) coefficients (e), hence their use in LFDs. For example, 35 nm gold nanoparticles (AuNPs) and silver nanoparticles (AgNPs) have attenuation coefficients far higher than common organic dyes such as Sudan III, methylene blue and crystal violet. The best porphyrin dyes, such as 5,10,15,20-tetra-21H-23H-porphine zinc, also have lower attenuation coefficients than AuNPs (Table 1). High attenuation coefficients lead to bright, vivid colours even at low concentrations in nanoparticles systems, making them ideal for sensing applications.
Gold nanoparticles (AuNPs, 15-800 nm) are the most widely used plasmonic particles 31 due to their biocompatibility, simple synthesis, low cost (at the concentration used) and straightforward functionalisation with organic molecules, particularly thiols. The most common method for AuNP synthesis uses HAuCl 4 , and sodium citrate as both a reducing and capping agent. [62][63][64] By using seeded growth, anisotropic nanocrystals, such as rods, wires and triangles can be obtained and used in lateral flow systems. 65 AuNPs can be surface-modified with the analyte targeting components by surface passivation using sulphur-linked (for example, thiols and thioctic acids) 66 or electrostatically charged antibodies or DNA strands for example. 39 Capping ligands include; zwitterions, polymers, peptides, proteins, sugars and nucleic acids (Fig. 3). 51 The vast majority of lateral flow systems utilise antibody capping groups and have been extensively reviewed. 19,31,37 In addition to gold, but outside of the scope of this review, other nanoparticles have also been tested in lateral flow systems, including; core-shell gold nanoparticles, 67 quantum dots, 68 graphene oxide 69 and carbon nanotubes. 70 Furthermore, Dou et al. have produced nanomaterial-free antibody systems 71 using crystal violet-stained antibodies, enabling a test to run in 14 minutes.
While many lateral flow devices are designed for qualitative analysis, often by eye, it is possible to quantify analyte concentrations. Urusov et al. describe three methods for quantitative detection; optical detection using a camera, scanner, or smartphone etc and image analysis software, or magnetic labels or electrically conductive labels. 72 In these methods the intensity of the test line is used as an indication of the analyte concentration, versus a calibration curve. The benefits of quantification, as described by Urusov et al. include the removal of human subjectivity, a lower limit of detection and an understanding of the target's concentration. However, the use of an additional device for analysis does decrease the applicability of some setups in low-resource scenarios. The simplicity, diversity and broad applicability to many targets have made lateral flow systems an attractive diagnostic. Many other POCT devices and lab-based diagnostics have been designed for use in more economically developed countries (MEDCs) but are not compatible in less economically developed countries (LEDCs). 73,74

MRDTs in the field
Malaria Rapid Diagnostic Tests (MRDTs) have been deployed in low-and middle-income countries as a successful example of the benefits and drawbacks of lateral flow compared to other POC devices. The World Health Organisation (WHO) estimated 438,000 deaths from malaria in 2014, with 90% of these deaths occurring in sub-Saharan Africa (Fig. 4). 30,75 In 2001, artemisinin-based combination therapies (ACTs) were recommended by WHO as first-line treatments. 76 Within the decade (2009), resistance to first-line ACT treatments was observed in the Greater Mekong Sub-region. [77][78][79] Resistance to ACTs has not been observed in Africa. However, there is concern that resistance will be observed. 80,81 During the period from 2008-2015, 35% of suspected malaria cases in Africa were not confirmed with diagnostic tests, which led to antimalarial overuse, increasing the risk of ACT resistance. 30,81,82 In the same period, resistance was further compounded by broader economic issues. Most malaria-affected countries use treatments costing less than $0.5 USD per capita, while artemisinin-based combinatorial therapies cost B$5 USD per capita. [83][84][85] A shift to a diagnostics-led approach could alleviate this problem, with research in Nigerian community pharmacies demonstrating the cost-effectiveness of MRDT-led treatment. 86 The United Nations Children's Fund (UNICEF) increased its procurement of Malaria Rapid Diagnostic Tests (MRDTs) from 3.8 million tests to 15 million. 87 MRDTs are POC immunochromatographic or photometric lateral flow devices utilising peripheral blood samples and selenium dyes or AuNPs conjugated to monoclonal antibodies as sensing units, and an example test is shown in Fig. 5.
They are handheld POCT for malaria detection and can differentiate four out of five malaria species (Plasmodium vivax, P. ovale, P. malariae and P. falciparum). The devices are,  however, insensitive to P. knowlesi: a simian parasite known to infect humans. MRDTs target a variety of antigen analytes such as histidine-rich protein 2 (HRP-2 is synthesised by all P. falciparum parasites as an abundant water-soluble protein, unlike HRP-1 and HRP-3), 92 Plasmodium lactate dehydrogenase (pLDH) and Plasmodium aldolase. 93,94 Alternatives to MRDTs and other lateral flow systems MRDTs are less sensitive than alternative non-POCT techniques, such as microscopy of thick blood films and PCR techniques which can detect low parasite concentrations of 50 parasites per mL and 5 parasites per mL, respectively. While WHO does not test MRDTs at concentrations below 200 parasites per mL. 30,93 MRDTs are also less capable of determining parasite species beyond P. falciparum and P. vivax, unlike Giesma-stained blood microscopy, 95 the most common diagnostic technique, that can have a sensitivity of 95% and a specificity of 98%. 30,93,96 Furthermore, this high sensitivity and specificity can be achieved at $0.2 USD per sample at a rate of one sample every 20 minutes. 96 However, microscopy is a labbased technique that requires an expert to carry out the analysis, plus the sensitivity can drop to 10% in the field. 97 Due to the COVID-19 pandemic, diagnostics have been rapidly rolled out globally, with rRT-PCR (real-time reverse transcription-polymerase chain reaction) being the gold standard, but turn-around time can be a limiting factor. In July 2020, in the United States, the average wait time for a COVID-19 RT-PCR test was 4 days, and only 37% of people received results within 2 days. 98,99 A comparison of rRT-PCR to LFDs showed that lateral flow device sensitivity was 4 85% and its specificity 4 95%, with a run time of just 15 minutes. 100 Several other techniques have emerged to address the time and cost issues of (the very accurate) rRT-PCRs. Loop-mediated isothermal amplification 101 (LAMP) can be run on a benchtop with various outputs generating the signal, 101,102 in between 20-90 minutes. 101 A LAMP device for malaria diagnosis was field-tested in India and Thailand and showed similar sensitivities and specificities to rRT-PCR. LAMP devices have also been developed for SARS-COV-2 and compared well to RT-PCR 103 and for malaria diagnosis too. 104 Aerts et al., 105 when discussing the cost-effectiveness of LAMP, microscopy and lateral flow tests in diagnosing cutaneous leishmaniasis in Afghanistan, suggested that all these diagnostic platforms have a role to play in different healthcare settings. They found that lateral flow systems reduce overall cost when used outside the clinic. Whereas microscopy is favourable in the clinic except in periods of high demand where LAMP and lateral flow devices are more cost-effective.

MRDT summary
In summary, MRDT lateral flow systems are cost-effective, fast to run and do not require a specialist to perform nor a dedicated laboratory. MRDTs provide an interesting case study in applying POC lateral flow as a low-cost, fast diagnostic versus lab-based and more expensive POC systems in LEDCs. Moreover, studies carried out in MEDCs (the United States of America) have found that MRDTs can outperform blood smear microscopy, 106 showing the potential of lateral flow systems in more economically developed countries where patients and health care professionals value short turnaround times. 5,12 Therefore, while more expensive reusable POCTs and labbased diagnostic systems have the potential to provide higher specificities, they do not have the same ease of use, minimal power needs, token maintenance requirements, and low costs of LFDs. It is essential to highlight, though, that some reports of MRDTs suggest they are still costly and challenging to use. 107 Crucially, antibody-based MRDTs can degrade when not transported and stored at low temperatures (''cold-chained''), becoming less-sensitive in conditions common in the field 108highlighting a potential weakness of protein-based LFDs and hence there is interest in developing antibody-free lateral flow devices.

Beyond antibodies: glycans as recognition units
Current POC diagnostics, particularly lateral flow systems (above), rely on antibodies as recognition units, owing to their high specificity and established methods for their production and nanoparticle conjugation. However, they are not the only biological recognition motifs; nucleic acids 109 and lectins have also been incorporated but are still far less commonly used than antibodies. 110 Glycans (carbohydrates) offer an alternative to these recognition units, and their study might bring new opportunities.
Glycans are ubiquitous in biological systems; 111 with over half of all human proteins glycosylated. 112 Glycosylation covalently adds glycans to proteins, lipids or other molecules (including RNA 113 ) to form glycoconjugates in a dynamic process, regulated by glycosyl transferases and glycosidases. 114 As it is not template driven, glycosylation cannot be (readily) predicted from genomic information alone. Glycoconjugates perform a vast range of roles, including; cell signalling, 115 hormonal actions, 116 cancer progression, 117 correct protein folding/structure 118,119 and mediating immune responses. 120 Many of these processes occur at the glycocalyx (Fig. 6), found on the surface of eukaryotic and bacterial cells, providing the primary interface for cells with their external environment. 121 The nature of glycan presentation on mammalian cells is a crucial marker of a disease state, 122,123 influenza species specificity 124 and influenza A viruses' zoonotic potential. 125 Antigens from P. falciparum have been shown to bind to sialyllactoses and other derivatives found on human erythrocytic glycoprotein A. 126 Consequently, there is a significant opportunity to target the glycans, or the ''glycan-readers'' termed lectins, in diagnostics. Lectins are a broad family defined as glycan-binding proteins which are neither enzymes, transporters or antibodies and often have highly conserved carbohydrate-binding domains (CBD). 128,129 Examples of lectins (or lectin-like proteins) relevant to healthcare include the Shiga toxins, 130 cholera toxin 131 and ricin, 132 while also being found in snake venoms 133 and biocides in marine algae. 134 Notably, the use of lectins for staining histology samples has been established for decades. 128 The use of lectins in biosensor design has been extensively reviewed. 135,136 For example, Damborský et al. have reported a lectin-based LFD for prostate-specific antigen (PSA). 110 There is extensive literature on the use of glycans in various applications, which is beyond the scope of this focussed review, but includes; anti-adhesion therapies, 137,138 glyconanotechnology, 139,140 antimicrobial applications 141 and, influenza 142 and human immunodeficiency virus (HIV) 143 vaccine development. A challenge (and opportunity) lies in the synthesis of oligosaccharides, which inhabit a vast chemical space beyond that of other biological macromolecules of nucleic acid and protein origin. 144 However, access to glycans for recognition studies is rapidly expanding, 145 thanks to the emergence of automated glycan synthesis [146][147][148] and the use of chemo-enzymatic strategies for greater glycan diversification. [149][150][151] Multivalency and glyco-materials Glycan-lectin interactions are typically weak with K d (dissociation constants) ranging from mM to mM, 139,152 which is significantly weaker than antibody-ligand interactions (in the nM to pM range). This is overcome in Nature by the presentation of multiple copies of glycans leading to statistical rebinding and chelation, resulting in significant enhancements to binding affinities -known as the ''cluster glycoside effect''. 153,154 The observed affinity enhancement (per glycan) as valency increases is non-linear and is dependent on the architecture of the glycan, its linker and its accessibility to the target. Due to this, there has been significant interest in using materials chemistry (polymers, particles and surfaces) to present multiple copies of glycans. In the context of diagnostic devices, the cluster glycoside effect is crucial to obtaining the necessary affinity and selectivity, which is not possible using individual glycans. To show the power of the cluster glycoside effect, Fig. 7 shows a summary of 37 glycoconjugates from a previous review. 152 This simple representation highlights enhancement effects, ranging from 2 to 410 000-fold increases associated with multivalency. However, it also shows that the simplified assumption that ever larger, higher valency systems lead to increased affinity is not valid: the effects are more subtle. For example, one of the most potent inhibitors of the cholera toxin has just five precisely placed glycans. 155 This is crucial to understand how glycosylated gold (or other) nanoparticles may apply in lateral flow diagnostics and how they can be designed more effectively.   Table 1'' from Lundquist et al. 152 highlighting the maximum enhancement in binding (corrected for ligand valency) as valency changes.

Structural and synthetic strategies for glycan presentation
In order to integrate glycans into nanoparticles for LFDs, it is crucial to consider the synthesis (and glycan-binding potential) of multivalent glycosylated materials, which is essential to generate the needed affinity/avidity. Whilst current LFDs use gold cores, other nanomaterials can be employed, and it is crucial to also consider these routes to multivalency, including polymeric methods. It is important to note that beyond multivalency/presentation on polymeric scaffolds, glycan arrays have allowed the probing of multi-antennary/branched glycans. 156 The benefits of this approach are well illustrated by Wang et al., who synthesised a series of asymmetrically substituted multiantennary glycans and found differential binding between the structures despite presenting similar terminal sugars. 157

Glycoclusters and dendrimers
Glycoclusters provide well defined and controlled glycan presenting architectures. André et al. designed a series of glycoclusters to increase affinity relative to the free sugar and enhance selectivity towards galectins, in some cases more than halving the IC 50 in a tetravalent arrangement. 158 More complex glycocluster systems have allowed linker length and valency tuning to target virulence factors such as LecA of Pseudomonas aeruginosa 159 and the fimbriae-mediated adhesion of E. coli, 160 respectively. Other groups have harnessed non-synthetic cores and directed evolution to design glycoclusters. 161 Dendrimers are oligomeric or polymeric structures produced iteratively, forming tree-like, highly branched regular structures and have been used as platforms for glycodendrimers. 162,163 A study by Woller et al. measured the binding enhancement realised as dendrimer generation increased in mannose-functionalised dendrimers and saw a 660-fold increase in activity per sugar, versus free sugar, when targeting Concanavalin A (ConA). 164 However, glycodendrimers often present a compromise -higher valences are accessible at the cost of decreased structural control and more challenging syntheses, unlike in glycoclusters, where structural control is often favoured over high valences. The high tunability of dendrimers also makes them an attractive approach for sugar presentation, with void spaces in larger dendrimers providing room for molecular cargo, potentially for clinical applications. 165,166 Polymeric approaches Glycopolymers present glycans either on their side chains or end-groups. Due to the relative ease of polymer synthesis, glycopolymers present a convenient route to engineering multivalency. 167,168 Controlled radical polymerisation techniques have been crucial to glycopolymer development enabling the synthesis of macromolecules with predictable chain lengths, compositions, and architectures -essential to dissect the impact of multivalency. These techniques include atom transfer radical polymerisation (ATRP), 169 reversible additionfragmentation chain-transfer (RAFT) 170 polymerisation, ringopening metathesis polymerisation (ROMP) 171 and nitroxidemediated polymerisation (NMP). 172 Glycopolymers have been reported (non-exhaustively) to bind with enhanced affinity to lectins and as inhibitors of influenza, 173,174 to sequester (and inhibit quorum sensing of) bacteria, 175 and to prevent bacterial adhesion. 176 It is emerging how the polymer backbone, which presents the glycans, can itself play a key role in fine-tuning affinity and selectivity. For example, Kiessling showed that cis-backbones (from ROMP) in mucin mimetics were more potent inhibitors than trans equivalents (Fig. 8A). 177 While amide-linked mannose side-chains were found to bind ConA less than ester-linked side-chains, this difference was attributed to subtle mobility differences. 178 Macromolecular engineering approaches have also been deployed to engineer specificity. For example, side-chain length (i.e. linker length between backbone and glycan) was varied to enable control over the access of galactose into the relatively deep GM1 binding pocket of the cholera toxin. 181,182 While adding branching units on the glycan-polymer linker, the allosteric sialic acid binding site could be targeted to further increase selectivity, Fig. 8B. 179,180,183 This approach also utilised multiple different glycans within a single polymer chain and was shown to increase affinity through a range of mechanisms 184 beyond just targeting a second binding sitepotentially steric shielding too. 185 An alternative approach to increase affinity with heterogeneous glycopolymers was introduced by Mahon et al. Polymer-Scaffolded Dynamic Combinatorial Libraries were used to dynamically select for high-affinity pendant groups (Fig. 9). 186 Glycopolymers have been deployed, or show potential for, therapeutic roles and biomedical applications, including enhanced immune responses in vaccine candidates 187 MRI contrast agents, 188,189 drug delivery, 190,191 anti-cancer agents 192 and to re-program the glycocalyx. 193 Chikae et al. 194 demonstrated that glycopolymer-coated nanoparticles immobilised on carbon electrodes, can detect amyloid-b, a peptide produced in Alzheimer's disease. Others have shown the potential of glycopolymer-coated AuNPs as anti-cancer 195 and transfection agents. 196 Each of the above examples shows how the underpinning high-affinity binding from glycopolymers to targets has the potential to be translated into innovative biosensors and diagnostics. The structural versatility of multivalent sugar presenting systems can be used to display a glycan and precisely adjust the nature of the presentation in 3D space 197 (as seen for native glycans) providing opportunities to tune the interaction.

Glycosylated gold nanoparticles
As discussed in the previous sections, antibody-functionalised gold nanoparticles are the most common core of lateral flow devices. There is limited literature on glyco-gold nanoparticles for LFD (discussed in the section below), but there is a broad literature on the synthesis and use of glycosylated gold nanoparticles, particularly for lectin binding, which is summarised here. Notably, gold nanoparticles are not the only nanoparticles systems used to detect lectins; examples include quantum dots and magnetic nanoparticles. 198 Whilst many glycosylated nanoparticle syntheses report direct immobilisation, or short linkers to attach the glycan, this can lead to colloidal stability challenges and irreversible aggregation when in biologically relevant media, such as saline buffers or blood plasma. Hence, the use of polymeric tethers that provide steric stabilisation and act as non-fouling 199 interfaces has been explored.
Polymer chains can be added to the surface of AuNPs by three conceptual methods; ''grafting to'' (a polymer is added to the surface of pre-prepared AuNPs), in situ (a polymer is added to the surface of growing AuNPs) and ''grafting from'' (polymerisation occurs on CTAs or initiators bound to the preprepared AuNP surface), Fig. 10. 200 The advantage of ''grafting to'' is that the precursor polymer can be synthesised and characterised, and installation of the glycan (or other functionality) can be confirmed, which is challenging when added directly to a nanoparticle surface. It is, however, important to note that the diffusion of a macromolecule onto the surface and steric constraints limit the grafting density. This is overcome with ''grafting from'', which gives high densities, but it must be isolated from the surface to characterise the polymer. Finally in situ methods, to produce particles in the presence of a polymeric capping agent, can be used which is a balance between the two other approaches.
Poly(ethylene glycol) (PEG) is widely used as a nanoparticle coating due to its low-biofouling, low cytotoxicity and commercial availability. Russel and co-workers used a PEG linker to immobilise a tri-functional sialoside to gold nanoparticles, enabling tuning of avian versus human influenza binding (dictated by the 2,3 versus 2,6 linkages). 201 Penadés and coworkers have used PEGylated gold nanoparticles 202 to investigate glycosphingolipid mediated carbohydrate-carbohydrate interactions when coupled with an SPR sensor. 203 Whilst PEG is widely used, there are thousands of other potential monomeric building blocks for the polymer tether. RAFT polymerisation has attracted interest in this respect, as it installs an a-terminal (latent) thiol on every chain, which is suitable for immobilisation onto gold surfaces. 204,205 Cameron and co-workers used poly(galactosides) derived by RAFT for the in situ formation of small (o20 nm) glycosylated AuNPs, capable of lectin recognition and to generate glyco-conjugate cancer vaccines. 206 RAFT polymerisation has also been used to make pH responsive 207 and thermoresponsive 208 glycopolymers on AuNPs. Other (non-RAFT) approaches using disulphide, double-headed ATRP initiators 209 or ''grafting from'' methods have also been used. 210,211 The versatility of these approaches enables a mix and match approach whereby polymer ligands can be applied, and the coated gold nanoparticles are easily isolated by centrifugation/washing cycles (Fig. 11).

Glycosylated AuNPs in sensors and diagnostics -from aggregation to lateral flow
Considering the large body of research on the synthesis of glycosylated materials, and integration into (gold) nanomaterials there has been obvious interest in their deployment for diagnostics and biosensing applications, such as colorimetric bioassays. 218 When nanoparticles become localised, the surface plasmon absorption energy maximum of a plasmonic nanoparticle is altered by collective conduction-band electron oscillation i.e. when nanoparticles aggregate there is a strong colour change. 51  showing how specific DNA sequences can be identified by a simple colour change in solution. 219 Due to the fact that lectins often have multiple binding sites (i.e. Concanavalin A (ConA) has 4, 220 Soybean agglutinin (SBA) has 4), 221 multivalent glycosylated gold nanoparticles have been explored as colourimetric sensors for lectins based on red-blue colour shifts. Many studies have focused on model (plant) lectins, such as G-rutin functionalised AuNPs (via a catechol) or mannosylated polymers to bind ConA, 222,223 galactose-functionalised glycopolymers to bind Ricinus communis agglutinin I (RCAI or RCA 120 ), 223 galactose-functionalised glycopolymers made by copper-free click immobilised on core-shell AuNPs to bind peanut agglutinin (PNA) 224,225 and lactose-functionalised glycopolymers to bind cholera toxin. 226 By using polymeric linkers, nanoparticle (steric) stability in complex media is introduced, avoiding false positives and unwanted aggregation which can occur with direct-surface conjugation. 227 The use of polymer tethers also enables additional functionality and properties, such as thermal control of glycan presentation. 228 Toyoshima et al. 229 synthesised variable density mannoseand N-acetylglucosamine-functionalised glycopolymers immobilised on 15, 40 and 100 nm AuNPs. Aggregation experiments were carried out with target and off-target lectins. The mannose functionalised AuNPs were assayed against ORN178 (an E. coli species specific for a-mannose) and aggregated. 229 While Richards et al. showed how fimbrae-differing phenotypes (FimH + and FimH À ) of E.coli can be discriminated by glycosylated AuNPs. 230 Aggregation is a crucial consideration of lateral flow (below) but the use of gold nanoparticles for SERS (Surface enhanced Raman scattering) for other device formats is also possible but is beyond the scope of this review. SERS has been demonstrated, 231 for example, for both ConA 232 (to pM levels) and human galectins. 233

Lateral flow
Despite the above, showing the significant body of literature on how multivalent presentation of glycans leads to large binding responses with glycosylated nanoparticles, the incorporation of these into lateral flow glyco-assays (LFGAs) has not been widely explored yet. As glycans can be chemically synthesised, there is no need to raise antibodies against emergent pathogens and the sheer range of tools to alter their presentation make them appealing targets. However   217 Toyoshima et al. 234 utilised a-galactoseand a-mannosepacrylamidophenyl pyranosides against lectins in aggregation assays and surface plasmon resonance (SPR) Fig. 12(A). In this work, however, they also showed a flow-through assay, where Shiga toxin-1 was deposited as the test line, facilitating some binding against the galacto-AuNPs Fig. 12(B). This was taken further by Mirua and co-workers using manno-AuNPs to detect ConA. In this case, the glyconanoparticles were used in the mobile phase and a rabbit anti-ConA antibody as the stationary phase test line Fig. 12(C). The dipsticks gave a visible 'red line' response to the naked eye. Whilst not a full glyco lateral flow, this study clearly demonstrated that glycoLFDs are possible. 235 Further studies varied the mannose density providing optimisation of the output signal. 236 A key learning point from this work, is the challenge of maintaining a good signal-to-noise ratio in the final device, with smearing of the nanoparticles as they run up the nitrocellulose due to non-specific interactions. This shows how the tuning of the polymer coating is essential alongside the use of conventional blocking agents such as BSA, to reduce non-specific interactions. The importance of the polymer coating has previously been observed in aggregation assays where a compromise must be reached between stability in aqueous media and rapid optical readout, usually tuned by adjusting polymer length. 227,230 Similarly, alterations to the polymer itself i.e., to a more sterically bulky polymer can be used to stabilise AuNPs 237 and favour binding but avoid aggregation. 238 Baker et al., reported that SARS-COV-2 (the causative agent of COVID-19) spike protein has affinity towards certain sialic acids, 239 (confirmed with microarrays 240 and STD NMR 241 ) discovered using glycosylated nanoparticles against spike protein immobilised on a biolayer interferometry sensor. N-Acetyl neuraminic acid functionalised poly(hydroxyethyl acrylamide) were immobilised onto AuNPs and were used to detect a SARS-COV-2 spike protein bearing pseudovirus in what was the first report of a complete lateral flow glyco-assay device (using a BSA-glycoconjugate test-line). 239 This was developed further into a flow-through device, whereby viral samples from patient swabs were dried as the test line, so no glycan test line was needed, allowing correct identification of positive/negative patients (Fig. 13A). 242 Controlled radical polymerisation (RAFT) was employed to undertake a detailed structure-function study exploring how the polymeric glyco tether impacted the performance of a lectin-detecting lateral flow system. 243 Using galactose as the glyco-ligand, the delicate balance between polymer chain  length to give specificity, but avoid non-specific binding was revealed, demonstrated by differential binding profiles to SBA and RCA120 using the same glycan on different polymers (Fig. 13A).
Further work has also been carried out by Baker et al. to remove proteins from all sensing components of an LFD i.e., no protein in the test line or on the particle surface. They did this by using poly(vinyl pyrrolidone) anchors decorated with glycans. 244 This was further developed by Kim et al. who used glycosaminoglycans from the glycocalyx, to make a protein-free test line for SARS-COV-2. 245 This approach of using nonprotein-based polymers, both synthetic and biological, to present sugars marks a step change in LFD development and could lead to improved workflows.

Conclusions and perspectives
The purpose of this review was to introduce lateral flow diagnostic technology, including its chemical basis, and the real-world application of these devices as low-cost diagnostic tools. We then set out to show that rather than just considering these as immuno-assays, with antibodies as the detection systems, that glyco-assays are also possible by exploiting the diverse range of biochemical interactions driven by glycans. Whilst this is a very new field, with less than 10 publications, this makes it the ideal time to introduce the topic, and in particular highlight the key considerations of multivalent presentation and the importance of how glycans are tethered to the nanoparticle (e.g. gold) core. The potential for glycopolymerfunctionalised AuNPs (or other cores) integrated into lateral flow systems, for use as POCT devices is vast, underpinned by the fast-growing field of glycomics. These low-cost, yet potentially robust glycopolymer-based lateral flow diagnostic devices could be ideal for low-and middle-income, and more economically developed countries alike.
As this is a new and exciting interdisciplinary area, spanning materials chemistry, analytical science, biochemistry, and biomedicine there exist several ''challenge areas'', or areas where continued scientific exploration is needed to drive the field forward.
The integration of new core materials, not just those based on gold or plasmonic nanoparticles.
A greater understanding of how the linker interface impacts the performance of devices and in particular to reduce non-specific binding.
The development of chemical tools to integrate glycans into particles more efficiently and at scale, including increasing access to rarer (often low quantity) glycans, alongside the development of the corresponding analytical tools.
Greater integration of multiple recognition modalities and further development of multiplexed tests.
The integration of glycan-based sensing into the various 'flavours' of lateral flow, or flow-through devices. (Fig. 14) both as all glyco systems, or in combination (''hybrid'') with antibodies, nanobodies, nucleic acids or other small molecule binders.
The demonstration of function in clinically relevant situations and settings, including in complex media.
Ensuring the devices are sufficiently sensitive and specific to be deployed for real world applications.
In summary, we have introduced the concept of lateral flow glyco-assays and their chemical basis, whilst laying out potential areas of application and the role of chemistry in developing these. It is the hope of the authors that lateral flow glyco-assays will become useful in the coming years, especially in resource-limited settings.

Literature review methods
A review of the literature was carried out to find papers containing AuNP aggregation assays and glycopolymer use in lateral flow assays.
A review of the literature was conducted through the following databases: Web of Science (all years), Embase  and Scopus (all years). Search date November 2021.

Conflicts of interest
RF is a shareholder and director of Iceni Glycoscience limited. ANB and MIG are named inventors on a patent relating to underpinning technology reported here.