OnVu™

OnVu™ (www.onvu.com) is one of the leading TTIs for a number of reasons. It can be activated easily at the point of application by exposure to UV light from a special activator supplied by the manufacturer. A UV blocking filter should then be permanently applied over the label to prevent subsequent UV exposure from causing reversion to the starting colour. The sensitivity of the indicator colour change can be controlled by the length of the UV activation exposure period. It utilises a photochromic ink that loses its blue colour in a time-temperature dependent spontaneous solid state crystal phase reaction following activation by UV light. This means that it can be printed directly onto packaging if required, in a very cost-effective manner. The fading of the active colour zone can be assessed visually, relative to a stable reference surrounding colour (www.onvu.com/content/en/retailer#2).

A strong correlation was reported between the shelf life of fresh domestic marinated salmon trout and the end-point of an OnVu™ indicator activated by exposure to UV light for 2 seconds (Smolander, 2009). A similar validation study was carried out using OnVu™ to predict the quality of seabream fillets (Taoukis et al., 2008). Both studies were part of the EU FRESHLABEL project aiming to develop tailor-made TTIs for specific fish and meat products.

A variant of OnVu™ in the advanced stages of development (CoolVu Active Barcode) initially obscures part of a two-stage barcode. If the product has been subjected to temperature abuse, the pigment coating the barcode becomes transparent so that the previously obscured portion of the barcode can then be read (http://www.freshpoint-tti.com/product/CoolVu-Active-Barcode.apx).

TT Sensor™

TT Sensor™ (Avery Dennison Corp, USA) is another TTI that can be distributed and stored in an inactive form, which is an important feature. It is supplied as two separate components; an indicator label and a transparent activator overlay. These are brought together automatically by a special dual-spindle applicator to activate the sensor at the time of application to the package to be monitored (Pacquit et al., 2008). Unlike some TTIs, TT Sensor™ labels do not need to be refrigerated prior to application, and offer a six-month shelf life. Once the two layers are brought together, diffusion of an acidic species from one polymer layer to the other results in a progressive, irreversible colour change in a pH indicator from fluorescent yellow to bright pink at a rate dependent on time and temperature (Anon., 2008, 2009). Different sensitivities are available to suit particular applications, and TT Sensor™ tags can be decorated, printed and cut to suit specific design requirements. They are intended for application to the outside of product packaging, not to the product itself. The Avery Dennison online listing of important things to know about the sensor (Anon., 2008) emphasises that ‘TT Sensor™ monitors temperature exposure over time – not product quality’.

Limitations of TTIs

Whereas TTIs are generally used to reveal exposure to undesirably high temperatures, most fruit and vegetables may also be damaged by storage below their safe minimum temperatures. Detection of such treatment is not a strength of most TTIs. Soon-to-be-released OnVu Ice™ is intended to demonstrate whether cold-sensitive produce has been exposed to sub-zero temperatures. However, it is difficult to use TTIs to demonstrate that tropical fruits like mango and banana have been subjected to sub-optimal temperatures in the 1–10 °C range. This task of unequivocal monitoring would be better performed by a downloadable temperature logger, such as an iButton™, or a disposable temperature sensor connected to an RFID that can permit automatic, regular downloading of temperature data to the internet. Systems like X-Sense™ (http://www.coldchainquality.com/how-xsense-works) and X_Tract™ (Anon, 2004; ****Anon, 2004, 2008) permit remote, real-time monitoring of RFIDs connected to disposable temperature sensors that could be distributed amongst packages at various representative points in a truck or container in transit. Such active RFID-linked temperature sensors cost about $US1 each, 10–50 times more than typical TTI labels, but they offer a more complete, unambiguous temperature record, revealing the time spent above or below the optimum transit or storage temperature.

Struggle for survival

FQIs offer a more direct indication of food freshness and safety than do correlative indicators such as TTIs and leak detectors. However, despite this apparent advantage, FQIs have generally been far less successful than TTIs in gaining a market foothold. Dermot Diamond, a prominent research leader with many years of involvement in the field, recently observed that the food quality indicator arena ‘is littered with abandoned ideas and technologies; probably because food related research is not as heavily funded as other areas, and research can be subject to stop-start-stop patterns that disrupt bringing technologies to the market; also the margins are very tight in the food industry, and it is often the case that a great idea cannot be properly developed at the right price’.

For most products that conform to the adage ‘freshest is best’, a colour change in a freshness sensor is generally bad news from the perspective of a wholesaler or retailer who at that point owns the product and wishes to sell it for the best price. Retailers are therefore typically unwilling to absorb the additional cost that such a sensor adds to a product line. Instead, they simply pass it on to consumers, who also often baulk at paying the extra price premium necessary to make a sensor profitable. Added to these problems are the unnecessary wastage or discounting that result from false positives, and, more seriously, possible liability claims that might result from false negatives if customers believe they can depend on the sensor as a guarantee of safety.

In view of the above maze of hurdles and obstacles, it is hardly surprising that very few freshness sensors for use on perishable foods have survived for more than a year or two as commercial products. Of the seven that were listed as commercially available in a review of the topic two years ago (Smolander, 2008), only a couple are still represented by up-to-date websites, and even they do not respond to invited inquiries via ‘contact us’, suggesting that they are no longer actively seeking sales.

One means of survival might be to target niche markets of socio-economic groups or supermarket chains that place particular emphasis and value on food freshness and safety. For example, It’s Fresh! Inc. has taken the novel approach of selling its Food Freshness Indicator mounted on a card, or pre-attached inside empty storage containers (http://www.foodfreshnesstechnology.com/researchanddevelopment.html). These are intended to be purchased by consumers for detecting signs of microbial spoilage in stored meats and meat-containing products after they have purchased them.

Fruit ripeness sensors open up fresh opportunities

Fruit marketing opportunities offer a ray of hope in an otherwise bleak outlook for freshness sensor developers. Many fruits are picked in a mature but unripe condition. Their eating qualities improve and ideally become optimal while on display, or shortly after a consumer has purchased them. Fruit ripeness sensors are therefore, in principle at least, viewed positively by all concerned; wholesalers, retailers and consumers alike. Occasional failure to give perfectly accurate information about the state of ripeness is unlikely to lead to a costly liability lawsuit. Anything that generally distinguishes ripe from unripe fruit is better than nothing for fruits that exhibit little or no significant changes in skin colour as they ripen.

Since fruit normally needs to be packed and distributed at a pre-optimal level of ripeness to permit damage-free handling and shipping, a ripeness indicator should ideally undergo a progressive change, making it useful at its different stages to inventory managers, produce managers and finally the consumer. Alternatively, the differing requirements of inventory managers and consumers might be better served by two distinct indicator systems. A sensor intended for inventory management use might preferably be machine readable (e.g. by optical or wireless scanning) and not meaningful to consumers, whereas sensors for consumer use ideally should be colorimetric and self-evident.

Other targets and challenges for fruit quality sensors

Although they have proven popular with consumers and produce managers alike, ripeness sensors and their associated packaging remain a challenge to produce at a cost that will permit widespread use on commodity fruits. Sensors for use in inventory management may prove easier to produce cost-effectively since a single sensor may then be used to monitor whole cartons, rather than small packs typically containing just 4–8 fruit. It should be possible to detect rots and other abnormal fruit based on their volatile signatures (Li et al., 2009, 2010; Moalemiyan et al., 2007). This type of behind-the-scenes application also opens up the attractive possibility of employing machine-readable transduction interfaces for monitoring the fruit quality without requiring a vivid associated colour change.

Wireless sensors employing printable nanoparticle inks

Printable metallic nanoparticle inks have opened up new, cheaper ways to produce printed circuits on packaging. Once printed, sintering is necessary to partially melt and fuse adjacent nanoparticles together to form a conductive structure. A new method of electrical sintering of printed conductive silver nanoparticles to produce printed conductor structures such as antennas, circuits and sensors on substrates offers a number of significant advantages over thermal oven sintering (Allen et al., 2008). This approach has been used to produce a prototype wirelessly readable spoilage sensor for poultry that is based around the reaction of silver with hydrogen sulphide (Alastalo, 2008). The resulting impedance change in the printed pattern of the sintered silver nanoparticles comprising the sensor within the package was wirelessly read through the package utilising capacitive coupling between sensor and an external reading device (Alastalo, 2008). Unlike most other RFID-linked sensors, which require a battery to power the sensor and the active RFID that transmits the sensor data, the above type of printed sensor is passive, meaning that the only powered component in the system is the reader.

Inkjet printed single wall carbon nanotube patterns (‘electronic coding’) have also been found suitable for wireless capacitive read-out applications (Helisto, 2008). A single printed electronic code is estimated to cost less than US 0.1 cents and can be invisible to the naked eye. Initially commercialised as an anti-counterfeiting measure (http://www.nicanti.com/technology.html), such electronic codes appear to have the potential to form the basis of a very cheap, wireless sensor system. One or more of the several different conductive inks used in producing a unique code might be made inherently sensitive to the target analyte in a way that affects its conductivity and thus the complex impedance of the code pattern when interrogated by a wireless reader. Alternatively, the analyte-responsive component might comprise an overlay or bridge between elements of the coding pattern sufficient to alter the complex impedance of the code.

Wireless sensors based on standard passive RFID tags

Whereas the above wireless sensor systems require a component of the printed conductive pattern to be altered by interaction with the target analyte, a team at GE Global Research has shown that ubiquitous passive 13.56 MHz RFID tags can be used for diverse sensing applications with little or no modifications (Potyrailo and Morris, 2007; Potyrailo et al., 2009a, 2010). With the aid of a special reader, the complex impedance of the RFID resonant antenna is measured and correlated to physical, chemical or biological properties of interest (Fig. 10.3). By measuring several parameters from the resonant antenna and applying multivariate data analysis, it is possible to perform multianalyte sensing and rejection of environmental influences using a single sensor (Fig. 10.4).

Generally, a sensing film is deposited onto the resonant antenna of the RFID tag to alter its impedance response in a manner that varies with the concentration of the analyte(s) of interest. Various overlay film compositions can be employed to give the best response and selectivity for the particular target analyte(s). Sensitivity varies between analytes. Alcohols can be detected at low ppm concentrations, while the lower limit for ammonia so far achieved was calculated to be 20 ppb (Potyrailo et al., 2009b). The sensitivity is also dependent on the power of the RF emission from the reader and the distance between the reader and the RFID sensor (Potyrailo et al., 2009b). These distances can be up to several centimetres, so a sensor RFID monitoring the headspace inside a sealed cardboard or plastic container should be readable from outside without the need for line of sight contact.

Additional bonuses of using industry standard passive RFID tags in sensors are (i) they are comparatively cheap (currently around US 5 cents each); and (ii) the memory chip can be used to store a product’s pedigree, along with sensor calibration and end-user data.

It is not always necessary to overlay a sensor film on the antenna in order to achieve sensor functionality from a passive RFID tag. Milk can be monitored non-invasively simply by attaching a standard disposable RFID tag to the outside of the milk carton (Fig. 10.5). Changes in the solution dielectric constant as a function of age of the milk can be monitored by its effects, through the wall of the container, on the electromagnetic field penetration depth out of the plane of the sensor (Potyrailo et al., 2009a). These initial studies need to be followed up by similar comparisons of milk with known differences in microbial contamination to determine whether the sensor is providing a measure of the extent of microbial growth or simply responding to biochemical changes associated with ageing of milk. It would also be interesting to discover whether this non-invasive system is more broadly applicable to the monitoring of other beverages.

DNA nanobarcodes

A group at Cornell devised a system for specific pathogen detection utilising short Y-shaped strands of DNA linked together into a tree-like structure with many branching ends, to which antibodies or molecular probes and fluorescent dye molecules can be attached (Li et al., 2005). Since both dye type and dye number can be precisely controlled, multicolour fluorescence intensity-encoded ‘nanobarcodes’ could be fabricated, distinguished by the ratios of the different types of dye. Using just three fluorescent dyes that can be distinguished optically, up to 1,000 different coded structures can be created and linked to specific antibodies or molecular probes (Steele, 2005). Consequently, the resultant DNA nanobarcodes not only have coding capacity, but also contain molecular recognition elements that can be used for molecular detection.

Several methods of detection and evaluation of specific binding were tested, including fluorescence microscopy, dot blotting and flow cytometry. However, simple visual observation could distinguish four different pathogens based on colours of dot blots after each pathogen was applied to a separate spot on the membrane (Li et al., 2005). Although originally intended for rapid medical diagnostics, this technology certainly has the potential for wider applications (Lee et al., 2010), and it may be feasible to simplify and adapt it for use in quality and safety indicators on packaging.

Quantum dots in fluorescence resonance energy transfer systems

Quantum dots (QDs) are luminescent semiconductor nanocrystals in the size range 2–6 nm with unique electro-optical properties that confer on them better brightness and photostability than conventional fluorescent dyes and make them better suited for multicolour applications (Algar et al., 2010). They have shown great potential in the development of novel fluorescence resonance energy transfer (FRET) assays, for, among other things, chemical sensing (Snee et al., 2006). The narrow, size-tuneable emission spectrum of QDs enables them to be FRET donors that can be easily customised to match the acceptor absorption features of various environmentally sensitive dyes that can be conjugated to or otherwise co-immobilised with QDs (Algar et al., 2010). QDs entrapped in sol–gel matrices have been shown to have potential to constitute a sensitive new type of pH sensing label with a number of advantages over the use of conventional pH indicator dyes alone, including high stability, luminescence and robustness (Coto-Garcia et al., 2009). These authors concluded that ‘our experiments indicate that the co-immobilization of an acceptor dye with the QDs in the sol–gel matrix could provide a novel, simple, general and advantageous approach for the future development of a wide variety of FRET-based sol–gel sensors’. It remains to be seen whether sensors containing QDs will become practical and cost-effective in package label applications. If low concentrations of QDs are used to reduce cost, their photoluminescence may fluctuate with time, a troublesome phenomenon known as blinking (de Dios and Díaz-García, 2010). Also, the toxicity of cadmium selenide, commonly found in QDs, may preclude its use in food-contact situations.

QDs as an optical transduction system for detection of analyte binding on molecularly imprinted polymers (MIPs) offer a number of advantages over organic fluorescent dyes that have been used widely in this role (Lin et al., 2004). Binding of the analyte (or template) to the corresponding imprint in the MIP can result in quenching of photoluminescence, presumably due to FRET between QDs and adjacent template molecules. In spite of the potential functionality, there are few published studies on MIP materials in conjugation with QD as probes (de Dios and Díaz-García, 2010). Perhaps the most relevant of these to food labelling has been the recent demonstration of a MIP attached to photoluminescent gold-silver nano-clusters being used for Bacillus cereus spore detection (Gultekin et al., 2010). Binding of dipicolinic acid, a characteristic component of B. cereus spores used to imprint the MIP, caused significant decreases in fluorescence intensity because it induced photoluminescence emission from the adjacent Au–Ag nanoclusters in the crosslinked nanoshell polymer matrix. B. cereus can cause food poisoning, so novel sensor labels for food safety monitoring could conceivably exploit this approach.

Polydiacetylene

Polydiacetylene (PDA) is in the recently resurrected category, having been highlighted again recently as a polymer with unique optical properties that seem to make it ideal for use in sensors (Pires et al., 2010). Its potential for use in sensors has long been known. Diacetylenic lipid monomers are readily polymerised in monolayers by UV irradiation to form a conjugated PDA backbone that changes colour from deep blue to red in response to various stimuli (e.g. increasing temperature or changes in pH) because of conformational changes in the conjugated backbone (Tieke et al., 1977). More remarkably, Charych et al. (1993) demonstrated the potential of PDA to undergo a similar colour transition from blue to red solely due to recep-tor–ligand interactions when the PDA layer was functionalised with a receptor ligand for the influenza virus haemagglutinin.

The basis of the above sensor principle was a bilayer assembly containing both the receptor-binding ligand and the capability (inherent in the adjacent PDA layer) to signal the specific binding event through a colour transition readily visible to the naked eye. Since other ligands can be incorporated into the film, ‘affinitychromism’ was claimed to offer the possibility of a general method for the direct detection of receptor-ligand interactions (Charych et al., 1993). That same team at the Ernest Orlando Lawrence Berkeley National Laboratory made further headlines by demonstrating a new sensor that offered the first instant test for a toxic strain of E. coli (Kahn, 1996). The method provided an instant analysis, and proponents predicted it could be formulated into a very inexpensive sensor that could be placed on plastic, paper or glass. They suggested that it could, for example, be part of a bottle cap or container lid. A colour change from blue to red would indicate contamination. They went on to demonstrate that polydiacetylenic lipids incorporating gangliosides that specifically bound cholera toxin or botulinum neurotoxin would change colour from blue to red when exposed to the toxins (Charych et al., 1996). They concluded that ‘poly-diacetylenic lipid membranes offer a general “litmus test” for molecular recognition at the surface of a membrane. A concentration of 20 ppm of protein could be detected using polymerized thin films. The speed, sensitivity and simplicity of the design offer a new and general approach towards the direct colorimetric detection of a variety of different molecules.’

Several patents were submitted and the technology was expected to be licensed to private industry. However, recent enquiries revealed that this much publicised, innovative technology was in fact never commercialised and its inventors have long since left the laboratory. One is left to speculate as to the possible reasons for yet another apparently great idea failing to become a commercial product.

Despite this initial commercialisation failure, scientific interest in the potential of PDA for use in novel detection systems has continued. An Israeli group more recently demonstrated that by combining a PDA film with various phospholipid bilayers, it is possible to obtain colorimetric detection of bacteria without the need to include specific recognition elements (Scindia et al., 2007). This utilises the general affinity of bacteria and bacterially secreted amphiphilic compounds for thin films of lipids. Their binding can induce chromatic blue-red/fluorescence transformations of interspersed PDA domains that serve as built-in optical reporters. They found that the degrees of bacterially induced colour transformations depended both on the bacterial strains examined and the lipid compositions of the films. This permitted bacterial fingerprinting through pattern recognition obtained by recording the chromatic transformations in an array of lipid/PDA films with different lipid components. These authors envisaged utilisation of their colorimetric bacterial film sensor in high-throughput medical screening, but it would seem possible to use this concept in a package label to detect bacteria.

Poly(thiophene)

Poly(thiophene) as a single cross-reactive conjugated polymer has been shown to generate a multidimensional response capable of identifying and differentiating between 22 structurally similar amines with 97% accuracy in a competitive aqueous environment, as well as detecting these compounds in real life assays of fish samples (Maynor et al., 2007). The poly(thiophene) polymer was found to exhibit various analyte-induced aggregation patterns with different amines, each with its unique optical signature (the shape of its absorption spectrum). Entire absorption spectra were used in linear discrimination analysis (LDA), permitting a single polymer to distinguish between 22 different amines, each applied separately. It seems unlikely that analysis of a single spectrum could identify all the component biogenic amines if applied together as a mixture. It was shown to be possible to use a ratiometric approach to determine the amount of amine present in fish samples spiked with known amounts of histamine. If this approach were to be used to produce a freshness sensor label for fish packaging, it is likely that a scanner or hand-held spectrometer with capabilities to assess label absorption across the entire visible spectrum would be required.

Dye-labelled single-stranded DNA

Single-stranded (ss) DNA oligomers (20–24 bases) stained with a fluorescent dye and dried onto a polyethylene substrate have been shown recently to change in fluorescence intensity upon exposure to various organic volatiles in a manner that reflected the precise sequence of bases (White et al., 2008). Individual DNA oligomer-dye probes typically do not offer a specific response to a particular volatile compound. Rather, they each exhibit a different characteristic pattern of both positive and negative changes in fluorescence in response to a range of different volatiles. An array comprised of microdots of 29 ss-DNA sequences, each covalently attached to a fluorescent dye, produced a complex pattern of responses that amounted to a multivariate fingerprint of each volatile (White et al., 2008).

Because of the intrinsically combinatorial nature of DNA, an oligomer 21 bases long can yield 4²¹ different sequences, each with the potential to be unique in some respects of its multivariable interactions with a range of organic volatiles. Since DNA oligomers can be readily engineered, amplified and replicated precisely, this discovery appears to open the door to a wide spectrum of new sensors for detecting volatile compounds in many applications. Oligomers can be selected that offer the most sensitive responses to particular target compounds and then replicated precisely. For example, one oligomer proved remarkably sensitive to 2,4-dinitrotoluene, permitting detection down to 6 ppb for this vapour characteristic of TNT-containing landmines (White et al., 2008). Generally, an array of fluorescent DNA oligomers will be needed to give non-ambiguous information about a product’s vapour profile. Tethering of the DNA oligomers to gold nano-particles may offer advantages, including improved sensitivity and selectivity, based on experience with other sensors employing linear DNA (Lee et al., 2010).

Practical applications of the selective odorant interactions of single-stranded DNA oligomers are likely to initially be in electronic odorant sensors, probably in combination with single-walled semiconducting carbon nanotubes (e.g. Johnson et al., 2010). For sensor label applications, it may be possible to print microarrays of different fluorescent DNA oligomers onto packaging to sensitively and specifically detect and quantify particular target analytes. Responses of dye-stained DNA oligomers reported so far have all been transient, and completely and rapidly reversible, so quantification would provide a current concentration rather than cumulative information about a particular volatile. The oligomer arrays could conceivably be cheaply produced and applied to packaging, but the reader would probably need to be a precise fluorometer, possibly hand-held. This could initially restrict this technology within food and beverage package labelling to applications in inventory management, rather than on the consumer interface. However, exciting advances are being made in adapting cell phones so that their cameras can be used as microscopes and colorimeters (Korhonen, 2009; Makinen, 2006). Such devices may one day allow consumers to interpret instantly a microscopic array of fluorescent DNA dots that might constitute a sensor label on packaging.