Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
IEEE Trans Nanotechnol. Author manuscript; available in PMC 2017 April 28.
Published in final edited form as:
PMCID: PMC5409134

Monitoring the Environmental Impact of TiO2 Nanoparticles Using a Plant-Based Sensor Network

Scott C. Lenaghan, Yuanyuan Li, Hao Zhang, Student Member, IEEE, Jason N. Burris, C. Neal Stewart, Jr., Lynne E. Parker, Fellow, IEEE, and Mingjun Zhang, Senior Member, IEEE


The increased manufacturing of nanoparticles for use in cosmetics, foods, and clothing necessitates the need for an effective system to monitor and evaluate the potential environmental impact of these nanoparticles. The goal of this research was to develop a plant-based sensor network for characterizing, monitoring, and understanding the environmental impact of TiO2 nanoparticles. The network consisted of potted Arabidopsis thaliana with a surrounding water supply, which was monitored by cameras attached to a laptop computer running a machine learning algorithm. Using the proposed plant sensor network, we were able to examine the toxicity of TiO2 nanoparticles in two systems: algae and terrestrial plants. Increased terrestrial plant growth was observed upon introduction of the nanoparticles, whereas algal growth decreased significantly. The proposed system can be further automated for high-throughput screening of nanoparticle toxicity in the environment at multiple trophic levels. The proposed plant-based sensor network could be used for more accurate characterization of the environmental impact of nanomaterials.

Index Terms: Biosystems, environmental monitoring, nanobioscience, nanobiotechnology

I. Introduction

CONCERNS about the impact of nanoparticles to environmental health and safety are rapidly increasing commensurate to their use in cosmetics, food, and paint industries [1]–[7]. Despite their widespread use, it is still unclear how these nanoparticles might impact the environment. Much research has focused on the effects of nanoparticles in mammalian systems, with few efforts examining the environmental impact in either the aquatic environment or on terrestrial plant species [8]–[11]. Specifically, most research has focused on the in vitro toxicity of metal nanoparticles to the lung, liver, brain, and skin [12]–[19]. Only recently have researchers begun to examine organic nanoparticles as an alternative to their metal-based counterparts [20]. Through the course of these studies, like metal nanoparticles, the toxicity of the organic nanoparticles has been found to depend on their structures and bioavailability [21]–[26]. While these studies are crucial for determining the direct impact on human and mammalian health, it does not address the issue of environmental toxicity, or develop a method for determining the level of nanoparticle exposure in the environment.

There is an increasing need to develop a high-throughput, reliable system for monitoring the environmental impact of nanoparticles at various levels. However, significant difficulties exist in the development of a procedure to determine if an environment has been exposed to potentially hazardous nanoparticles. Often the use of large and expensive equipment, such as aerosal-mass spectrometry systems, is necessary to detect nanoparticle exposure [27]. Other options include delayed detection through inductively coupled plasma atomic emission spectroscopy, size-exclusion chromatography, microfiltration, field-flow fractionation, and capillary electrophoresis [28]. These detection systems have several disadvantages that make them impractical to use for broad landscape-level environmental monitoring: 1) detection cannot be conducted at the potential source of exposure; 2) the use of these techniques is relatively costly and time-consuming; 3) detection requires an experienced operator to conduct the testing; and finally, 4) even if the systems detect nanoparticles in the environment, they are not capable of determining if the nanoparticles have a negative effect on the ecosystem. In an effort to develop a more effective sensor for environmental monitoring of the effects of nanoparticles, we propose a sensor network-based approach in this paper. The purpose of the sensor network is to broadly interrogate biomes for potential hazards.

Sensor networks have gained broad attention in recent years with their ability to measure and gather information from the environment, and, based on local decision-making processes, transmit the data to a user. With the advent of the internet and smart phones, the applicability of these systems has rapidly been embedded into our daily lives. These systems may acquire data continuously and transfer high-fidelity data across a network. As a result, the sensor nodes should strive to process the raw sensor signal locally and perform local decision making to determine the most “interesting” signals/events, such as detecting anomalous events. Local processing and decision making avoids wasting resources on “uninteresting” data, such as sending normal raw sensor readings to a human operator for interpretation, which could be regarded as noise. Distributed decision-making algorithms for anomaly detection in natural environment greatly simplify the jobs of human operators. In anomaly detection applications, a wireless sensor node in the network can monitor its local region, and communicate through a wireless channel with other nodes to collaboratively produce a high-level representation of the environment. Using such a network, a large area can be monitored at a relatively low cost. The challenge for implementing the sensor network lies in the fundamental understanding of the sensors, modeling of the sensor data, and adherence to appropriate engineering principles. Data modeling in such distributed sensor networks is critically important for determining the “normal” operation mode of the system. This will allow the system to perform autonomous anomaly detection to recognize when unexpected sensor signals are detected. The advantage of this approach is that the designer does not have to characterize the anomalous signatures in advance. Instead, the system of sensor nodes can learn this characterization autonomously, for application to new domains. For environmental monitoring of the effects of nanoparticles, we have chosen to build a sensor network using biological system sensors that can be easily embedded in the environment.

Instead of deploying costly hardware, biological sensors such as plants and algae are capable of detecting chemical and physical toxins in the environment at relatively low levels, extending the “vision” of the sensor network, and integrating signals [29]. Plant-based biosensors (“phytosensors”) have previously been used to detect environmental exposure from a variety of contaminants, including radiation, heavy metals, and chemical toxins [30], [31]. Many of these systems use microarray studies to determine genes that are up regulated by a specific toxin. Once identifying the genes, a fluorescence protein (FP) reporter gene can be coupled to the gene or promoter of interest and the expression of the gene can then be monitored externally. This kind of transgenic biosensor requires a significant amount of time and resources to develop, and is generally highly specific to a certain toxin. This approach has led to the development of very specific biosensors, but not one that can determine exposure from a wide range of toxins—a generalist sensor that acts as a “check engine light.” To circumvent this problem, we have developed a phytosensor that is capable of detecting the environmental impact of nanomaterials on plants using a variety of cues to reflect plant health. At the early stages, this system can be used to identify plants that are sensitive to the nanoparticles before costly FP-based transgenic systems are developed. These generalist phytosensors have the advantage of being noninvasive, directly demonstrate positive and negative responses from the biosensors, and are cost effective. A network of biosensors has great potential for applications, such as environmental monitoring [32], water quality monitoring [33], and health monitoring [34]. Since plants have not previously been used as sensors to monitor the environmental impact of nanoparticles, and differences may occur between different plant species, online learning algorithms are desirable to develop and employ for the system.

For the initial testing, Arabidopsis thaliana has been used as the biosensor. This species has been chosen for the following reasons: 1) previous studies have demonstrated that it is susceptible to a variety of metal nanoparticles, including ZnO, Al2O3, SiO2, Fe3O4, and SWCNT’s (see Table I); 2) it is fast growing and compact; 3) it is the most widely adopted model plant for research; and 4) it has clear morphological changes after toxic exposure that can be easily analyzed through image processing. In general, toxicity in plants can be detected by monitoring changes in the shape, texture, color, and orientation of the phytosensors. These factors have previously been proven effective at determining stress or toxicity induced by pathogens, toxins, and radiation [35], [36].

Known Nanomaterial Toxicity in Arabidopsis

In this study, individual sensor nodes were established using A. thaliana as the phytosensor. An array of sensor nodes was then combined to create a phytosensor network. A learning algorithm was used to train the system to identify changes in the environment. Finally, the system was used to test if toxic effects could be observed from the administration of TiO2 nanoparticles. The results from this study demonstrate the feasibility of using a phytosensor network to monitor the environmental impact of nanoparticles at multiple trophic levels. The design of this system is readily expandable to a large number of sensor nodes capable of monitoring a large geographic area. By using machine learning techniques, the system is also highly adaptable and provides a robust way to detect abnormal changes in a defined set of plants.

II. Materials and Methods

A. Phytosensor Setup

Arabidopsis thaliana ecotype “Columbia” plants were grown from seeds. Plants were chosen at four days of growth, and potted individually in 6 cm2 pots. The plants were then moved to an environmental growth chamber set at a constant temperature of 25.2 °C and humidity level of 80%. The environmental growth chamber used a 16–8 day–night cycle to simulate realistic lighting conditions in temperate summer. Three plants were placed in a Tupperware container and compromised a single sensor (experimental unit). The use of a separate Tupperware container for each sensor allowed the introduction of the control (tap water) and test solutions (TiO2 nanoparticles) without the risk of cross contaminating the sensors.

B. Establishment of the Sensor Nodes

A complete sensor node consisted of phytosensor units, as defined earlier, an RGB camera, and a computer to analyze the acquired images (see Fig. 1). Logitech QuickCam Pro4000 cameras were used to monitor the biosensors and capture images for subsequent analysis. This camera is an off-the-shelf commercial product that provides photo quality even at low light levels. Additionally, since it is inexpensive, a large number of cameras can be easily deployed into the environment. The cameras were triggered, by the attached laptop computer, every half hour for one month to constantly determine the health of the plant (sensor). In the preliminary setup, the cameras were connected to the laptop through a USB connection; however, wireless data transfer could be used instead of USB transmission in future studies. A crucial step in establishing this type of sensor node is translating the image data into a measurement of plant health.

Fig. 1
Experimental platform. A biosensor is composed of three separately potted Arabidopsis plants exposed to the same environment containing either tap water, or tap water mixed with 1000 mg/ml TiO2 nanoparticles. Images of all four biosensors are collected ...

Typically, the key metrics used to determine the health of the plant include: shape and orientation of the leaves (an indicator of wilting), spectral characteristics of the leaf (color and temperature, indicators of photosynthesis), and the texture of the leaf (an indicator of the water content). In this study, we have chosen to focus only on the spectral characteristics associated with leaf color as a measure of the plant health. Previous studies have demonstrated the potential for machine vision to be used in the detection of plant toxicity due to plant pathogens, fungi, etc., [39].

The machine learning algorithm used in this study focused on comparison of the current captured image with that of the image captured at the previous time point. If the difference between the two pictures exceeded a threshold, then an abnormal change reading was detected. In addition, the picture frames were analyzed over time, to detect sequential anomalies associated with normal plant behavior associated with growth. To assess the color changes of the leaf, we utilized a histogram thresholding technique. First, the RGB image was converted to a grayscale intensity image. The conversion eliminates the hue and saturation information while retaining the luminance. The algorithm converts RGB values to grayscale values by forming a weighted sum of the R, G, and B components. The weights for R, G, and B components were 0.2989, 0.5870 and 0.1140, respectively. After converting the images, they were segmented based on the distribution of “green” pixels. Finally, the segmented images were used to train the fuzzy adaptive resonance theory (ART) neural network.

C. Training of the Sensor Nodes

Fig. 2 shows the proposed learning and sequential detection procedures on a single sensor node. The first step in training is to determine the baseline values for the phytosensors by growing them in the optimal conditions in the environmental growth chamber. This allows the sensor nodes to develop a classifier for normal and expected signals. In addition to the classifier, a time analysis model was built to represent dynamic changes over the course of time. First, a likelihood-ratio sequential detection was used to detect anomalies. Next, the system used a classifier to categorize the stress cues into different environmental states and perform sensor fusion, using the fuzzy ART neural network [40]. The fuzzy ART system is an unsupervised artificial neural network that can perform dimensionality reduction and pattern classification simultaneously. There is no offline training phase required by the fuzzy ART neural network. The learned class prototypes can be updated at each sampling point. Therefore, it is adaptive to changes in both the environment and the phytosensors themselves. Moreover, the algorithm is simple enough to run in real time, while maintaining an acceptable performance level. Fuzzy ART was first implemented in the wireless sensor network [41], and has since been updated to improve anomaly detection performance by incorporating temporal and spatial information [40], [42], [43]. Being able to model and detect time-related changes allows periodic detection of the leaf color change, indicating exposure to the nanoparticles. In order to accomplish the detection of abnormal leaf color, we propose a three-step detection scheme. First, the semantics were extracted out of the temporal sequence using a symbol compression technique called Lempel–Ziv–Welch [44]. Next, a variable memory length Markov model (VMM), i.e., a probabilistic suffix tree (PST) [45], was used to model the compressed temporal sequence. A variable memory Markov model preserves the minimal subsequences (of variable lengths) that are necessary for precise modeling of the given statistical source, resulting in a more flexible and efficient sequence representation than a traditional Markov model [46]. Finally, a universal background model (UBM) [47] likelihood-ratio detector was used to detect anomalies in the time patterns.

Fig. 2
Machine learning framework used for the biosensors. The top portion of the figure diagrams the learning procedure, whereas the bottom portion illustrates the detection procedure.

Phytosensors may have long periods of inactivity, i.e., nondetection, which would be computationally costly to model using the Markov-based time model alone. To avoid this expense, a symbol compression technique was used to map these temporal events/sequences into semantic symbols, prior to analysis with the VMM model. The advantages of the current model, over continuous-measuring models are: 1) digital image data are often discrete; 2) environments that are modeled with discrete states have clear physical interpretations and are therefore natural and easy for humans to interpret (e.g., toxic or nontoxic); 3) data compression techniques, which the system uses to reduce the size of the event sequence, typically require discrete state representations. To detect time-related changes, ideally, the system would build a normal life span of healthy plants and an abnormal lifespan after the plants have been exposed to the nanoparticles. In reality, it is difficult to model all different types of abnormal events, since all possible abnormal situations are unknown and cannot be modeled in advance. To account for this, the model builds a normal temporal model and a universal temporal model (all training sequences that may have mixture of normal and abnormal sequences). The UBM detector then compares the current phytosensor observation sequences against the normal PST model and matches the sequence against the universal PST model. If the likelihood ratio is below a threshold, then an abnormal state is detected. Note that the system can incrementally build the model during the online detection stage to make the model more adaptive to the current environment.

D. Nanoparticle Testing

For nanoparticle testing, spherical rutile TiO2 nanoparticles (50 nm in diameter, Stock#5485MR) were purchased from Nanostructured and Amorphous Materials, Inc. (Houston, TX). The morphology and size distribution of the TiO2 nanoparticles was conducted by the manufacturer using transmission electron microscopy and laser scattering, and provided on the certificate of analysis. Solutions of nanoparticles for toxicity testing were formed by the addition of 1 g of nanoparticles to 1 l of tap water. Tap water alone was used as a control for normal growth of the plants. The nanoparticle solution was added to the plant by pouring into the Tupperware container. As indicated earlier, each phytosensor had a separate container for the solution to prevent cross contamination. The nanoparticle and control solutions were added to the plant every two days to ensure that plants remained hydrated. This procedure was maintained throughout the course of the study.

E. Analysis of Algae Growth

After completing the initial analysis of the images, it was observed that in the control samples, there was an abundance of algae that was not present in the nanoparticle samples. Since the goal of this study was to analyze the environmental impact of the TiO2 nanoparticles, it was important to analyze this effect. To analyze the algae growth, first the raw RGB images were segmented to separate the area containing the plants from the water only images. This segmentation left a 19 × 83 pixel area containing only the image of the water supply where the algae would grow, and were read into the MATLAB code. Next, the true color RGB image was converted to a grayscale intensity image by eliminating the hue and saturation information while retaining the luminance. Although the size of the image, 19 × 83 pixels was small, there were a significant number of images 891, for analysis. Due to the low resolution of the images, it was necessary to increase the size of the images by using a magnification factor of 5, increasing the image size to 95 × 415 pixels. This enlargement allowed smoothing of the image, which would have been impossible in the original image. After smoothing, the image was cropped with a four-element position vector [20 65 45 320], which removed the shadow effect from the corner of the water tray, and prevented any bleed over from the growing leaves in the original image. This edge cropping was necessary to reduce false positives associated with the dynamic shadows present on the edges of the container. The next step consisted of enhancing the grayscale image by transforming the values in the intensity image so that the histogram of the output image matched the specified histogram. In this study, the intensity values were originally spread over the entire range from [0, 255]. After enhancing the contrast, a threshold was set to detect the emergence of the algae; in this particular case, a value of 7 was used. All pixels with a value less than 7 were considered to be algae, and appeared as white in the resulting images. In the final step, the image stream was saved as a video, and the amount of algae over time was determined based on the number of pixels detected at each time point.

III. Results

A. Effect of TiO2 Nanoparticles on Arabidopsis Health

Experiments were conducted over 28 days, yielding 891 total frames for analysis. Using the detection system, it was possible to detect changes in the control biosensors, Biosensors 1 and 2, at day 15. In these samples, the leaves began to change from green to yellow, indicating decreasing plant health. On day 16, greater than 20% of the leaves had turned yellow, with increasing amounts of yellow observed during the continuation of the study. As indicated in Fig. 3, the change in green level in Biosensors 1 and 2 could be clearly observed. For the fuzzy ART detection, the plants were classified into two categories, either increased or decreased health. The model was tuned so that if >20% of the leaves turned yellow, the plant was deemed to be unhealthy, indicating a lack of growth. Based on the fuzzy ART detection, at 500 min Biosensors 1 and 2 changed their status from healthy to unhealthy (see Fig. 4). They remained in the unhealthy state for the remainder of the study. In contrast, Biosensors 3 and 4, the TiO2 nanoparticle treated samples, remained in the healthy state for the duration of the study. It should be that the increase in growth of Biosensors 3 and 4 did not affect their state classification. The increased growth of the TiO2 nanoparticle treated samples was an unexpected result, but demonstrates the strength of the sensor-network approach. The vigilance value for Biosensors 1, 3, and 4 was 0.6, while the vigilance value for Biosensor 2 was 0.77. The vigilance value is a measure of the sensitivity of the range of 0–1. 0 means that the network is not sensitive to any change, while 1 indicates that a new class would be created for each pixel change. Biosensor 2 had a higher vigilance factor because the starting plant was smaller, providing fewer green pixels. Since Biosensors 1, 3, and 4 used similar size plants, it was possible to analyze them with the same vigilance value. To ensure that the trends observed were reproducible, the study was replicated two more times with similar results.

Fig. 3
Images captured during the course of the study. At Day 16, the percentage of yellow leaves was greater than 20% of the green pixels in Biosensors 1 and 2. Biosensors 2 and 4 maintained their green color throughout the course of the study.
Fig. 4
Fuzzy ART neural network detection results. Class 1 indicates all plants were green. Class 2 indicates more than 20% of the leaves changed into yellow. Sampling rate is 30 min per frame. The algorithm detected changes in Biosensors 1 and 2 around day ...

B. Effect of TiO2 Nanoparticles on Algae Growth

In addition to analysis of the terrestrial plant health, the biosensor system also allowed for visualization of algae proliferation in the solution used to maintain the hydration of the phytosensors. Analysis of the image segments that monitored the water level in the plant containers showed that there was an inverse relationship between the growth of algae in the control samples and the decreased growth of the control plants (see Fig. 5). Due to the addition of the TiO2 nanoparticles into the experimental biosensors, there was a significant decrease in the amount of algae, indicating a toxic effect from the TiO2 nanoparticles. Considering that the algae will compete with the terrestrial plants for the availability of nutrients, the emergence of algae in the water supply is expected to decrease the growth of the slower growing terrestrial plants. This is a common phenomenon observed with eutrophication induced from environmental contaminants, where the rapid growth of algae will outcompete slower growing plants. While the exact mechanism used by the TiO2 to reduce algae growth in the biosensors cannot be identified from this study, previous studies have demonstrated that nanoparticles are toxic to many species of green algae [48]–[50]. Clearly, the reduction in the amount of algae between the controls and experimental biosensors illustrates a potential toxic response from the introduction of the nanoparticles into the system. This environmental impact might lead to changes at other trophic levels leading to a greater environmental impact over a long time course.

Fig. 5
Analysis of algae growth over time in both the control and nanoparticle treated samples. Note that at the completion of the study, there is significantly more algae present in the control samples. In addition, the samples begin to diverge around 12 days, ...

IV. Discussion

In this study, we have demonstrated the feasibility of using phytosensors to detect the environmental impact of nanomaterials, specifically, TiO2 nanoparticles, in a system similar to that found in the real world. In the proof-of-concept design, the biosensor node was designed as a two component system that has a proven capability to detect the impact on both a terrestrial plant and algae. In this study, the TiO2 nanoparticles, at a concentration of 1000 mg/ml, demonstrated an increase in the green content of these samples, potentially indicating an increase in the health of Arabidopsis thaliana when compared to untreated controls. At the same time, the reverse trend was observed for the growth of green algae in the biosensor system, with the control samples having pronounced algal growth when compared to the TiO2 treated biosensors. Several mechanisms exist in the literature to explain the observed effects from the introduction of the TiO2 nanoparticles. Previously, TiO2 nanoparticles have been shown to induce the isotropic growth of root cells through disregulation of microtubule assembly in Arabidopsis [51]. The increased speed of Arabidopsis growth after introduction of the TiO2 nanoparticles may be the result of this rapid root growth, or, potentially, the response to the decreased algae content. As observed in eutrophic systems, the growth of algae can lead to a reduction in available nutrients for other slower growing plant species. It is possible that the decreased amount of algae in the experimental samples lead to more nutrient availability, and increased growth. In either instance, the goal of this study was to develop a robust environmental monitoring system that could detect any environmental impact from the introduction of the nanomaterials. In this way, the biosensor system demonstrated its strength, whereas a more specific detection system may have missed the observation of the impact.

Another key aspect of the system designed in this study is the expandability, and the ease of use of the system. The developed biosensor detection system can autonomously detect anomalies adaptively using sensor data that are collected by distributed biosensors. In the proof-of-concept design of the biosensors, the biosensor node is a two component system that has proven capable at detecting the impact on a terrestrial plant and algae. In future studies, the goal will be to expand the system to include invertebrate grazers, Daphnia, and vertebrates (fish), Danio rerio, both of which are key species used to determine aquatic toxicity [50], [52]–[54]. This will expand the platform for analysis of three trophic levels, and will also help to gauge if there is a keystone species, to which a nanoparticle effect will have a more dramatic effect. Typically, such complicated ecosystem analysis is difficult, and often requires a significant amount of time from an observer, and generates qualitative data. Because the system developed in this study uses cameras to observe changes over time, and machine learning techniques are used to automatically making sequential detection decisions, there is a decreased investment of human monitoring of the system. By monitoring the system through a sequence of image data, it is also possible to obtain quantitative data from the system, which would otherwise not be possible. As indicated earlier, it is also possible to introduce any nanomaterial into the system to determine if a potential environmental impact may result from this nanomaterial. Considering the cost effectiveness of the developed system, it would also be possible to screen a wide range of nanomaterials, and nanomaterial concentrations, in a high-throughput manner without significant monetary or labor costs. Considering the speed at which new nanomaterials are currently being created and used, a robust system for determining the environmental impact of nanomaterials is crucial.

Perhaps the most significant application for the phytosensor system developed in this study is the potential large-scale monitoring of wide geographic areas to determine the health of the system. With the advent of satellite imaging, it would be possible to survey a large area and use machine learning and image processing techniques to determine the health status of an entire crop. Using such an approach, it would be possible to determine if there is a change in crop health in near real-time and also to determine the area of the crop most affected. In the case of a toxic event, using the concept developed in this study, it would be possible to identify the toxic effect in a time-dependent manner, which would indicate where the toxin was introduced into the environment. Determination of the source of introduction of a toxin into the environment is very difficult to ascertain when on the ground, and delays in determining the source can lead to significant environmental damage, increased costs, and further spreading. Large-scale environmental monitoring, such as described earlier, would be extremely difficult without an adaptable automated system, and represents the true strength of a system such as that designed in this study.

V. Conclusion

Using machine learning, phytosensors, and remote monitoring techniques, we established a proof-of-concept system for monitoring the environmental impact of nanoparticles on both a terrestrial plant and algae. We observed increased leaf growth in Arabidopsis thaliana with the introduction of 1000 mg/ml TiO2 nanoparticles, when compared to control plants. Corresponding with the occurrence of this growth trend, there was a toxic response to green algae causing a significant reduction in biomass in the TiO2 treated samples. By using the machine learning framework, it was possible to detect this environmental impact using an automated system, without a substantial investment of time or monitoring from a human operator. The system demonstrates a cost-effective approach for the monitoring of the environmental impact of nanomaterials using sensor network and presents a scalable alternative for large-scale and wide-area monitoring. This system could be used for future studies to identify the risk associated with the large-scale use of nanomaterials.


This work was supported in part by the UTK-ORNL Science Alliance under Award 3318039, the US Army Research Office under Grant W911NF-10-1-0114, and the National Science Foundation under Grant CBET: 0965877 and Grant CMMI: 1029953. The review of this paper was arranged by Associate Editor L. Dong.


An external file that holds a picture, illustration, etc.
Object name is nihms819641b1.gif

Scott C. Lenaghan received the B.S. degrees in marine science and biology from the University of Miami, FL, USA, in 2000 and the Ph.D. degree in biological sciences from Auburn University, Alabama, USA, in 2008.

In 2009 he joined the Department of Mechanical, Aerospace, and Biomedical Engineering at the University of Tennessee, Knoxville, USA, as a Postdoctoral Research Associate in the Nano Bio-systems and Bio-mimetics Lab. In 2011, he joined the faculty at the University of Tennessee as a Research Assistant Professor, within the same lab, where he currently works. His research interests include nanorobotics, bio-inspired material development, advanced diagnostics, cellular biology, and biomedical imaging. He has authored more than 30 journal publications.

Dr. Lenaghan was awarded a distinguished NSF STEM Fellowship in 1998 for excellence in teaching.

An external file that holds a picture, illustration, etc.
Object name is nihms819641b2.gif

YuanYuan Li received the B.S. degree in computer science from the Minnesota State University, Mankato, USA, and the M.S. and Ph.D. degrees in computer science from the University of Tennessee, Knoxville (UTK), USA, in 2010.

She was a Research Assistant in the distributed intelligent laboratory from 2003 to 2010 and she was a postdoctoral researcher in the mechanical, aerospace, and biomedical engineering department at the UTK from 2010 to 2011. Since 2011, she joined biostatistics branch at the National Institute of Health. Her research interests are in machine learning, biostatistics, bioinformatics, and artificial intelligence. She has been a reviewer for the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), the IEEE International Conference on Robotics and Automation (ICRA), and the Journal of Robotics and Autonomous Systems (JRAS) and Plos One.

An external file that holds a picture, illustration, etc.
Object name is nihms819641b3.gif

Hao Zhang (S’09) received the B.S. degree in electrical engineering from the University of Science and Technology of China, Hefei, China, in 2006 and the M.S. degree in electrical engineering from Chinese Academy of Sciences, China in 2009. He is currently a Ph.D. student in computer science at the University of Tennessee, Knoxville, USA.

He interned at Oak Ridge National Laboratory in summer 2011. His research interests are in computer vision, machine learning, and human perception.

An external file that holds a picture, illustration, etc.
Object name is nihms819641b4.gif

Jason Burris received the B.S. degree from North Carolina State University, USA in horticultural science in 2005 and the M.S. degree in plant sciences in 2010. He is currently working toward the Ph.D. degree in plant sciences with a research focus on identification and molecular characterization of English ivy nanoparticles.

An external file that holds a picture, illustration, etc.
Object name is nihms819641b5.gif

Neal Stewart received the B.S. degrees from North Carolina State University, in 1984, and after a short stint of public school teaching received the M.A. degree in education from Appalachian State University, NC, USA, in 1988, and the M.S. and Ph.D. degrees from Virginia Tech., Blacksburg, USA, in 1990 and 1993, respectively.

He is a Professor of plant sciences and holds the Ivan Racheff Chair of Excellence in plant molecular genetics, and he is the Director of the Tennessee Plant Research Center. He has postdoctoral experience at the University of Georgia in Wayne Parrotts laboratory. He joined the faculty of UNC-Greensboro and was an Assistant Professor, and then Associate Professor from 1995–2002 in biology, after which he assumed the Racheff Chair and at the appointment of professor, and then with tenure in 2006. His Labs research spans plant biotechnology, genomics, and ecology. He has been performing agricultural biotechnology and biotechnology risk assessment research since 1994. His research has been supported by various granting agencies including the DOE, USDA, NSF, EPA, NASA, and various US military agencies. He has participated in biotechnology ventures in the private sector during much of his science career. He teaches graduate level courses in plant genomics and research ethics and an undergraduate course in plant biotechnology and genetics. He has given scientific and lay-presentations around the US and in 16 countries. He has authored or co-authored over 200 publications, including four books. He has editorial duties for 7 journals.

An external file that holds a picture, illustration, etc.
Object name is nihms819641b6.gif

Lynne E. Parker (S’92–M’95–SM’05–F’10) received the B.S. degree from Tennessee Technological University, Knoxville, TN, USA, in 1983, the M.S. degree from the University of Tennessee, in 1988, and the Ph.D. degree in computer science with a minor in brain and cognitive science from the Massachusetts Institute of Technology (MIT), USA, in 1994.

She is a Professor and an Associate Head with the Department of Electrical Engineering and Computer Science, University of Tennessee, where she serves as the Director of the Distributed Intelligence Laboratory. She also holds an appointment as an Adjunct Distinguished Research and Development Staff Member at Oak Ridge National Laboratory, Oak Ridge, TN, where she was a Full-Time Researcher for several years. Her research interests include distributed robotics, heterogeneous teams, humanrobot interaction, sensor networks, and machine learning.

Dr. Parker is a Former Editor and Associate Editor of IEEE TRANSACTIONS ON ROBOTICS.

An external file that holds a picture, illustration, etc.
Object name is nihms819641b7.gif

Mingjun Zhang (S’98–M’01–SM’06) received the M.S. degrees in bioengineering and electrical engineering from Stanford University, Stanford, CA, USA, the D.Sc. degree from Washington University, St. Louis, USA, and the Ph.D. degree from Zhejiang University, P.R. China.

After working 7 years in industry, he started his academic career as an Associate Professor at the University of Tennessee, Knoxville, TN, USA, in 2008. He and colleagues made several original contributions to ivy nanoparticles, fungus-based nanoparticles, tea nanoparticles, and sundew nano-scaffolds for applications in drug delivery, tissue engineering and cancer therapy. They also discovered the unique energy-efficient actuation and propulsion mechanisms of giardia and whirligig beetle. He is interested in building biologically engineered nanoparticle-based systems with energy-efficient propulsion mechanisms inspired by microorganisms. His research benefits from the groups unique expertise on integration of multiple nano-instrumentation platforms, and systems theory. His research has been supported by ARO, ONR, NSF, National Academies’ Keck Future Initiative, ORNL, and industries. His research have been published in PNAS, Nano Letters, Advanced Functional Materials, PLoS Computational Biology, and drawn international media attention, including Science, Nature, AAAS Science Update, BBC news, Science Daily, Science News, and many others.

Dr. Mingjun was awarded the Young Investigator Program Award by the Office of Naval Research, and the Early Career Award by the IEEE Robotics and Automation Society.


Color versions of one or more of the figures in this paper are available online at

Contributor Information

Scott C. Lenaghan, Department of Mechanical, Aerospace, and Biomedical Engineering, The University of Tennessee, Knoxville, Knoxville, TN 37996 USA.

Yuanyuan Li, Department of Electrical Engineering and Computer Science, The University of Tennessee, Knoxville, Knoxville, TN 37996 USA.

Hao Zhang, Department of Electrical Engineering and Computer Science, The University of Tennessee, Knoxville, Knoxville, TN 37996 USA.

Jason N. Burris, Department of Plant Sciences, The University of Tennessee, Knoxville, Knoxville, TN 37996 USA.

C. Neal Stewart, Jr., Department of Plant Sciences, The University of Tennessee, Knoxville, Knoxville, TN 37996 USA.

Lynne E. Parker, Department of Electrical Engineering and Computer Science, The University of Tennessee, Knoxville, Knoxville, TN 37996 USA.

Mingjun Zhang, Department of Mechanical, Aerospace, and Biomedical Engineering, The University of Tennessee, Knoxville, Knoxville, TN 37996 USA.


1. Berube D. Rhetorical gamesmanship in the nano debates over sunscreens and nanoparticles. J Nanoparticle Res. 2008;10:23–37.
2. Contado C, Pagnoni A. TiO2 in commercial sunscreen lotion: Flow field-flow fractionation and ICP-AES together for size analysis. Anal Chem. 2008;80:7594–7608. [PubMed]
3. Faunce T, Murray K, Nasu H, Bowman D. Sunscreen safety: The precautionary principle, the Australian therapeutic goods administration and nanoparticles in sunscreens. NanoEthicsl. 2008;2:231–240.
4. Schmid K, Riediker M. Use of nanoparticles in swiss industry: A targeted survey. Environ Sci Technol. 2008;42:2253–2260. [PubMed]
5. Sozer N, Kokini JL. Nanotechnology and its applications in the food sector. Trends Biotechnol. 2009;27:82–89. [PubMed]
6. Tiede K, Boxall A, Tear S, Lewis J, David H, Hassellov M. Detection and characterization of engineered nanoparticles in food and the environment. Food Additives Contaminantsl. 2008;25:795–821. [PubMed]
7. Palanivel V, Zhu D, van Ooij WJ. Nanoparticle-filled silane films as chromate replacements for aluminum alloys. Prog Organic Coatingsl. 2003;47:384–392.
8. Corredor E, Testillano P, Coronado M-J, Gonzalez-Melendi P, Fernandez-Pacheco R, Marquina C, Ibarra MR, de la Fuente J, Rubiales D, Perez-de-Luque A, Risueno M-C. Nanoparticle penetration and transport in living pumpkin plants: In situ subcellular identification. BMC Plant Biol. 2009;9:45–56. [PMC free article] [PubMed]
9. Lee W-M, An Y-J, Yoon H, Kweon H-S. Toxicity and bioavailability of copper nanoparticles to the terrestrial plants mung bean (Phaseolus radiatus) and wheat (Triticum aestivum): Plant agar test for water-insoluble nanoparticles. Environ Toxicol Chem. 2008;27:1915–1921. [PubMed]
10. Lin D, Xing B. Phytotoxicity of nanoparticles: Inhibition of seed germination and root growth. Environ Pollutionl. 2007;150:243–250. [PubMed]
11. Lovern SB, Strickler JR, Klaper R. Behavioral and physiological changes in daphnia magna when exposed to nanoparticle suspensions (Titanium Dioxide, Nano-C60, and C60HxC70 Hx) Environ Sci Technol. 2007;41:4465–4470. [PMC free article] [PubMed]
12. Auffan M, Rose J, Wiesner MR, Bottero J-Y. Chemical stability of metallic nanoparticles: A parameter controlling their potential cellular toxicity in vitro. Environ Pollutionl. 2009;157:1127–1133. [PubMed]
13. Cho W-S, Cho M, Kim SR, Choi M, Lee JY, Han BS, Park SN, Yu MK, Jon S, Jeong J. Pulmonary toxicity and kinetic study of Cy5.5-conjugated superparamagnetic iron oxide nanoparticles by optical imaging. Toxicol Appl Pharmacol. 2009;239:106–115. [PubMed]
14. Hussain SM, Hess KL, Gearhart JM, Geiss KT, Schlager JJ. In vitro toxicity of nanoparticles in BRL 3 A rat liver cells. Toxicol Vitrol. 2005;19:975–983. [PubMed]
15. Kim S, Choi JE, Choi J, Chung K-H, Park K, Yi J, Ryu D-Y. Oxidative stress-dependent toxicity of silver nanoparticles in human hepatoma cells. Toxicol Vitrol. 2009;23:1076–1084. [PubMed]
16. Lin W, Huang Y-w, Zhou X-D, Ma Y. In vitro toxicity of silica nanoparticles in human lung cancer cells. Toxicol Appl Pharmacol. 2006;217:252–259. [PubMed]
17. Migdal C, Rahal R, Rubod A, Callejon S, Colomb E, Atrux-Tallau N, Haftek M, Vincent C, Serres M, Daniele S. Internalisation of hybrid titanium dioxide/para-amino benzoic acid nanoparticles in human dendritic cells did not induce toxicity and changes in their functions. Toxicol Lett. 2010;199:34–42. [PubMed]
18. Park MVDZ, Annema W, Salvati A, Lesniak A, Elsaesser A, Barnes C, McKerr G, Howard CV, Lynch I, Dawson KA, Piersma AH, de Jong WH. In vitro developmental toxicity test detects inhibition of stem cell differentiation by silica nanoparticles. Toxicol Appl Pharmacol. 2009;240:108–116. [PubMed]
19. Wu J, Liu W, Xue C, Zhou S, Lan F, Bi L, Xu H, Yang X, Zeng F-D. Toxicity and penetration of TiO2 nanoparticles in hairless mice and porcine skin after subchronic dermal exposure. Toxicol Lett. 2009;191:1–8. [PubMed]
20. Xia L, Lenaghan S, Zhang M, Zhang Z, Li Q. Naturally occurring nanoparticles from english ivy: An alternative to metal-based nanoparticles for UV protection. J Nanobiotechnol. 2010;8:1–9. [PMC free article] [PubMed]
21. Dhana lekshmi UM, Poovi G, Kishore N, Reddy PN. In vitro characterization and in vivo toxicity study of repaglinide loaded poly (methyl methacrylate) nanoparticles. Int J Pharmaceuticsl. 2010;396:194–203. [PubMed]
22. Huang G, Zhang N, Bi X, Dou M. Solid lipid nanoparticles of temozolomide: Potential reduction of cardial and nephric toxicity. Int J Pharmaceuticsl. 2008;355:314–320. [PubMed]
23. Kedmi R, Ben-Arie N, Peer D. The systemic toxicity of positively charged lipid nanoparticles and the role of Toll-like receptor 4 in immune activation. Biomaterialsl. 2010;31:6867–6875. [PubMed]
24. Plard J-P, Bazile D. Comparison of the safety profiles of PLA50 and Me. PEG-PLA50 nanoparticles after single dose intravenous administration to rat. Colloids Surfaces B: Biointerfacesl. 1999;16:173–183.
25. Semete B, Booysen L, Lemmer Y, Kalombo L, Katata L, Verschoor J, Swai HS. In vivo evaluation of the biodistribution and safety of PLGA nanoparticles as drug delivery systems. Nanomedicine: Nanotechnol Biol Med. 2010;6:662–671. [PubMed]
26. Singh P, Prasuhn D, Yeh RM, Destito G, Rae CS, Osborn K, Finn MG, Manchester M. Bio-distribution, toxicity and pathology of cowpea mosaic virus nanoparticles in vivo. J Controlled Releasel. 2007;120:41–50. [PMC free article] [PubMed]
27. Kane DB, Oktem B, Johnston MV. Nanoparticle detection by aerosol mass spectrometry. Aerosol Sci Technol. 2001;34:520–527.
28. Simonet B, Valcárcel M. Monitoring nanoparticles in the environment. Anal Bioanal Chem. 2009;393:17–21. [PubMed]
29. Hollis RP, Killham K, Glover LA. Design and application of a biosensor for monitoring toxicity of compounds to eukaryotes. Appl Environ Microbiol. 2000 Apr 1;66:1676–1679. [PMC free article] [PubMed]
30. Bontidean I, Mortari A, Leth S, Brown NL, Karlson U, Larsen MM, Vangronsveld J, Corbisier P, Csöregi E. Biosensors for detection of mercury in contaminated soils. Environ Pollutionl. 2004;131:255–262. [PubMed]
31. Giardi MT, Koblízek M, Masojídek J. Photosystem II-based biosensors for the detection of pollutants. Biosens Bioelectron. 2001;16:1027–1033. [PubMed]
32. Bilitewski U, Turner APF, editors. Biosensors for Environmental Monitoring. New York: Taylor & Francis; 2000.
33. Sanchez B, Yale G, Chatni R, Ochoa-Acuña H, Porterfield D, Mclamore E, Sepúlveda M. Oxygen flux as an indicator of physiological stress in aquatic organisms: A real-time biomonitoring system of water quality. Proc SPIEl. 2009;7304:730405-1–730405-8.
34. Milenkovi A, Otto C, Jovanov E. Wireless sensor networks for personal health monitoring: Issues and an implementation. Comput Commun. 2006;29:2521–2533.
35. Perfus-Barbeoch L, Leonhardt N, Vavasseur A, Forestier C. Heavy metal toxicity: Cadmium permeates through calcium channels and disturbs the plant water status. Plant J. 2002;32:539–548. [PubMed]
36. Terry N. Physiology of trace element toxicity and its relation to iron stress. J Plant Nutritionl. 1981;3:561–578.
37. Lee C, Mahendra S, Zodrow K, Li D, Tsai Y, Braam J, Alvarez P. Developmental phytotoxicity of metal oxide nanoparticles to arabidopsis thaliana. Environ Toxicol Chem. 2010;29:669–675. [PubMed]
38. Shen C-X, Zhang Q-F, Li J, Bi F-C, Yao N. Induction of programmed cell death in arabidopsis and rice by single-wall carbon nanotubes. Amer J Botanyl. 2010 Oct 1;97:1602–1609. [PubMed]
39. Chaerle L, Leinonen I, Jones HG, Van Der Straeten D. Monitoring and screening plant populations with combined thermal and chlorophyll fluorescence imaging. J Exp Botanyl. 2007 Mar 1;58:773–784. [PubMed]
40. Li Y, Thomason M, Parker LE. Detecting time-related changes in wireless sensor networks using symbol compression and probabilistic suffix trees. Proc IEEE/RSJ Int Conf Int Robots Syst. 2010:2946–2951.
41. Kulakov A, Davcev D. Tracking of unusual events in wireless sensor networks based on artificial neural-networks algorithms. Proc Int Conf Inf Technol: Coding Comput. 2005;02:534–539.
42. Li Y, Parker LE. Detecting and monitoring time-related abnormal events using a wireless sensor network and mobile robot. Proc IEEE/RSJ Int Conf Int Robots Syst. 2008:3292–3298.
43. Li Y, Parker LE. A spatial-temporal imputation technique for classification with missing data in a wireless sensor network. Proc IEEE/RSJ Int Conf Int Robots Syst. 2008:3272–3279.
44. Sayood K. Introduction to Data Compression. San Francisco, CA, USA: Morgan Kaufmann; 2000.
45. Mazeroff G, Gregor J, Thomason M, Ford R. Probabilistic suffix models for API sequence analysis of windows XP applications. Pattern Recog. 2008;41:90–101.
46. Ron D, Singer Y, Tishby N. The power of amnesia: Learning probabilistic automata with variable memory length. Mach Learn. 1996;25:117–149.
47. Reynolds DA, Quatieri TF, Dunn RB. Speaker verification using adapted Gaussian mixture models. Digital Signal Process. 2000;10:19–41.
48. Aruoja V, Dubourguier H-C, Kasemets K, Kahru A. Toxicity of nanoparticles of CuO, ZnO and TiO2 to microalgae pseudokirchneriella subcapitata. Sci Total Environ. 2009;407:1461–1468. [PubMed]
49. Hund-Rinke K, Simon M. Ecotoxic effect of photocatalytic active nanoparticles (TiO2) on algae and daphnids (8 pp) Environ Sci Pollution Res. 2006;13:225–232. [PubMed]
50. Hall S, Bradley T, Moore JT, Kuykindall T, Minella L. Acute and chronic toxicity of nano-scale TiO2 particles to freshwater fish, cladocerans, and green algae, and effects of organic and inorganic substrate on TiO2 toxicity. Nanotoxicologyl. 2009;3:91–97.
51. Wang S, Kurepa J, Smalle JA. Ultra-small TiO2 nanoparticles disrupt microtubular networks in arabidopsis thaliana. Plant, Cell Environ. 2011;34:811–820. [PubMed]
52. Zhu X, Chang Y, Chen Y. Toxicity and bioaccumulation of TiO2 nanoparticle aggregates in daphnia magna. Chemospherel. 2010;78:209–215. [PubMed]
53. Zhu X, Zhu L, Duan Z, Qi R, Li Y, Lang Y. Comparative toxicity of several metal oxide nanoparticle aqueous suspensions to zebrafish (Danio rerio) early developmental stage. J Environ Sci Health, Part Al. 2008;43:278–284. [PubMed]
54. Griffitt RJ, Luo J, Gao J, Bonzongo J-C, Barber DS. Effects of particle composition and species on toxicity of metallic nanomaterials in aquatic organisms. Environ Toxicol Chem. 2008;27:1972–1978. [PubMed]