The experiments conducted for this project employ a novel microarray-based technology known as surface plasmon enhanced fluorescence (SPEF). This technology can be used to simultaneously assess the levels of multiple distinct biomarkers that are characteristic of specific disease states in a given cell population. In previous studies, we have shown that metallothionein (MT; a critical stress response protein) levels will alter the progression of both beneficial and deleterious immune activities. We hypothesize that the gene dose of metallothionein will significantly influence the transcription factor (TF) biomarker signatures (biosignatures) that are a component of the activities. Protocols for SPEF TF measurement were initially developed using the human Jurkat T lymphocytes as a prototype cell line. Subsequent experiments then employed primary murine splenocyte lysates from strains of mice with distinct MT gene doses (wild type, MT-knockout, and a transgenic strain over-expressing MT). These lysates were directly fluorescently labeled using the Alexa Fluor 647-NHS dye and were passed over SPEF arrays that have been prepared with antibodies against various TFs in order to characterize TF biosignatures. In some instances, cell lysates from the three mouse strains were captured on these arrays after in vitro activation using T-cell mitogens in order to further interrogate the functional capacities of the cells. Our results suggest that this technology has significant promise as a means of specific biosignature detection for the purposes of both diagnosing disease states and exploring molecular interactions.

The need for efficient, sensitive, and reliable protein-based microarrays to diagnose disease states early in their progression has become increasingly apparent over the last decade. While DNA-based microarrays are relatively ubiquitous, RNA may at times be a poor predictor of specific protein activity, indicating the need for direct arrays which can simultaneously survey levels of expression of multiple proteins in a given sample (1,2). Currently extant proteomic microarrays have been successfully optimized to detect antigen-specific IgE levels with similar sensitivity to that of commercially available ELISA assays (2). These microarrays have great potential predictive clinical value, as shown by Quintana et al, who used an antigen-based microarray to establish patterns of IgG expression in nonobese diabetic mice which were strongly predictive of each mouse’s susceptibility to cyclophosphamide-accelerated diabetes (3). One method of disease detection relies upon biomarkers: cellular molecules which have levels of expression that are characteristic of a current or future disease state or disease susceptibility. Biomarkers are often used in drug discovery and have obvious value in a clinical setting, allowing for advance treatment of a probable disease state before its more insidious symptoms can manifest. For example, increased levels of IL-18 as detected by ELISA in urine following cardiopulmonary bypass were recently found to be predictive of acute kidney injury, yielding the potential for physicians and patients to take this susceptibility into account when assessing the risks of such surgical procedures (4). Biomarkers are assessed on the basis of both their specificity (the number of correctly established true negative results) and their sensitivity (number of correctly established true positive results) (5). Oftentimes, the specificity or sensitivity of a given biomarker may not be high; for example, a urine IL-18 level >50pg/mL 12 hours after cardiopulmonary bypass had a specificity of 94% but a sensitivity of only 50% as a predictor of acute kidney injury (4).This suggests that only 50% of cases of acute kidney injury would be correctly diagnosed (a high rate of false negatives), while only 6% of negative cases would be falsely diagnosed as positive. Thus, while urine IL-18 is a good predictor of acute kidney injury, it alone is insufficient to diagnose such a condition.

A solution to single biomarker sensitivity/specificity problems is the study of patterns of multiple biomarker activities in disease states. These biomarker signatures (biosignatures) of disease serve to increase both specificity and sensitivity of disease detection by simultaneously measuring levels of multiple biomarkers to provide a more precise indication of physiological states. A recent review on tuberculosis (TB) diagnosis and drug development stresses that the toxic effects of the disease lack a single reliable biomarker, and as a result early diagnosis can be a challenging process. The authors instead propose the use of a panel of well-defined biomarkers of TB in order to ameliorate rapid diagnoses and drug discovery (6). Other biosignature discovery projects currently underway include establishing biosignatures of the prodromal stage of Alzheimer’s disease in order to allow for advance treatment of the disease (which currently cannot be diagnosed with certainty until autopsy), and detecting biosignatures of renal transplant tolerance in order to custom tailor treatment methods to individual transplant recipients (7,8). The ability to predict disease onset from specific protein biosignatures has myriad applications in the current healthcare system. At birth, infants in all US states have blood drops collected that are screened for specific metabolic disorders, and are then stored for a defined period of time. As initial testing generally leaves the majority of these dried blood spot samples (DBSS) unused, numerous multiplexed protein- and DNA-based disease screens could be performed without requiring any additional sample collection from neonates (9). Janzi et al were able to use microarrays to detect C3 deficiency in Swedish DBSS which had been stored for up to 20 years, with a strong correlation between microarray and ELISA data, demonstrating the diagnostic value of these blood spots were they to be screened upon collection (10). While complement deficiency is a lack of a single defined protein, strategies are currently being developed to screen for biosignatures of much more complex diseases such as post-traumatic stress disorder; the ability to identify this and other debilitating conditions early in their course (or even before disease onset) would elucidate many potential early intervention treatments (11).

Many other clinical applications of protein microarrays based on DBSS analysis can easily be imagined. DBSS-derived assays have already been developed which can test for the presence of specific toxins on blood collection cards, as demonstrated by Bottein et al. who were able to detect ciguatoxin in stored blood collection cards collected up to 12 hours post-toxicant exposure (12). Occupations inevitably risk exposing employees to toxic compounds, producing myriad negative health outcomes. At times, compounds may not initially be known to possess carcinogenic properties, such as naphthalene, which was only classified as “potential human carcinogen” in 2000 by the US EPA despite having been a ubiquitous environmental and occupational compound for decades, especially in certain industrial settings (13). If all employees of a particular company were to be screened for levels of a toxicant such as naphthalene after being exposed for a period of time, the relative values would be uninformative without a baseline reading from the beginning of each employee’s tenure, as preexisting exposure would vary on an individual basis. Storing a liquid serum or blood sample for each employee would be cost- and space-prohibitive; however storing DBSS samples would be affordable and would require very limited storage space. Stored DBSS would allow for baseline screens of compounds which are not known to be toxic at the onset of employment, providing a clear advantage over performing only specific screens after a hazard has been identified. A similar screening technique can be imagined for soldiers entering the armed forces, with DBSS being stored before each soldier goes to war in order to serve as a baseline to test toxicant exposure during wartime operations.

There are a variety of proteomic microarray platforms currently under development, including those based on the phenomenon of grating coupled surface plasmon resonance (GCSPR). A surface plasmon is an evanescent electromagnetic wave which may exist at a metal-dielectric interface as the result of charge-density oscillations as a result of excitation, and which can be best detected at a specific angle known as the SPR angle where reflectivity of a region is at a minimum (14,15). The SPR angle will change as the result of any event which alters the index of refraction at the metal-dielectric interface, and this shift can be calculated, allowing for the detection of individual molecular interactions such as antibody-antigen binding on the surface of a metal-dielectric interface.

Traditional SPR is limited in its capacity as a multiplexed biosensor system, allowing for the simultaneous measuring of very few ligand-specific interactions, whereas the more recent GCSPR platform is able to simultaneously assess hundred of such interactions. GCSPR utilizes gold sensor chips with a diffraction grating at the surface, diffracting incident light into higher orders and allowing for the simultaneous generation of surface plasmons for multiple dissimilar ligand-specific interactions (14). GCSPR microarray chips are prepared by depositing multiple specific antibodies onto a chip surface in nanoliter volumes at specific regions of interest (ROIs). Incident light is then shined onto a chip from a variety of angles, with the SPR angle being the angle with the lowest intensity of reflected light, and the shift in SPR angle being directly correlated with the mass of antigen bound to each individual ROI (16). The technology has been demonstrated to be capable of detecting mouse IgG concentrations with slightly lesser sensitivity than standard ELISA assessment, and has been applied to capture cells based on cell surface antigen expression as an alternative to standard flow cytometry-based methods of cell sorting (14,16).


Although GCSPR is a powerful label-free technology, it has limitations owing to its sensitivity, which is below that of a standard ELISA and which is especially low for assays that have yet to be optimized, inevitably limiting sensitivity in a multiplexed format (14). A novel alternative approach is the use of surface plasmon enhanced fluorescence (SPEF), a technique which combines the versatility and sensitivity of standard fluorescence microarrays with the signal specificity and multiplexed capabilities of GCSPR. In standard fluorescence immunoassays, light of an excitation wavelength corresponding to a specific fluorophore will produce diffuse fluorescence of the resultant emission wavelength. When a surface plasmon is generated by light of the proper excitation wavelength at a metal-dielectric interface, the plasmon will additionally excite a fluorophore in a directional manner, enhancing signal without the use of standard fluorescence amplification schemes that often result in low net sensitivity gain (17). When used with a standard fluorescence microscope, a slide with a metal grating on the surface resulted in a 24-fold enhancement in fluorescence intensity when compared with an uncoated slide (18).

For the SPEF assays currently under development in our laboratory, sensor chips are blocked and ligands of interest are captured on these chips as in standard GCSPR assays. Samples are either directly fluorescently labeled with Alexa Fluor 647 dye or a labeled secondary antibody dye is flowed over the chip in a sandwich immunoassay format. The fluor is then excited using a 650nm excitation laser and the resultant fluorescence is captured by with a camera and analyzed. Early results using this experimental design have produced a 50-fold increase in fluorescence detection over standard emission conditions, with sensitivities of detection that have reached 10-15g/mL (17). This places SPEF assays on-par with standard ELISA techniques, and with continued optimization the technology has the potential to achieve a 100 to 1000-fold increase in fluorescence detection, with the ultimate goal of achieving zeptomolar levels of sensitivity.

As previously mentioned, there are two standard approaches to fluorescent immunoassays; the direct and indirect (sandwich) approaches. Each experimental methodology has unique advantages and disadvantages that make them ideal in certain circumstances. In the context of SPEF, sandwich immunoassays consist of a specific capture antibody spotted on the chip to which unlabeled sample binds. Unlabeled primary antibodies are then flowed over the chip, attaching to the sample on the sample-capture antibody complex. Lastly, fluorescently labeled anti-species secondary antibodies are flowed over the chip, binding the appropriate primary antibodies and allowing for fluorescent imaging of the sample. The direct labeling approach consists of labeling a sample (such as a whole cell lysate) directly with a fluorescent dye, and then flowing it over capture antibodies which then bind labeled proteins of interest, allowing for fluorescent imaging of the sample (21).

The primary advantages of the indirect immunoassay technique are its increased sensitivity and amenability to amplifications schemes (19,20). These assays are however limited by the increased time required for the multiple incubation and wash steps, the unavailability of matched pair capture and primary antibodies, and the potential for cross-reactivity between primary, secondary, and capture antibodies (21). This potential for cross-reactivity results in the need for a large number of controls which increases exponentially proportionally to the number of capture antibodies, constraining this technique in the context of a multiplexed biosensor format (20,21). Additionally, although amplification schemes do serve as a powerful tool, they often amplify background noise along with true signal, resulting in a limited net sensitivity gains that makes this advantage less salient (17).

The direct labeling approach has the obvious advantages of significantly reduced time per assay and, most notably, the lack of need to control for nonspecific secondary antibody binding, These advantages vastly reduce the number of controls required for each experiment, thereby reducing the cost-per-assay and making this technique easily accessible as a proteomic biosignature detection format (19,20). The primary disadvantage of this technique is the potential a reduction in net sensitivity. Li et al. compared relative background-subtracted fluorescence produced by multiple cytokine immunoassays in both direct and indirect formats and discovered an inconsistent signal between cytokines in the direct but not indirect labeling experiments, despite presenting equal concentrations of all 5 cytokines in all assays. They posit that this inconsistency is due to differential labeling of cytokines by the fluorescent dye, a result which they confirm with standard SPR techniques (19). Thus, while direct labeling makes samples readily usable in the SPEF format, individual ROI fluorescence may be a poor indicator of relative protein concentration as different proteins may bind different numbers of fluorescent molecules depending on the number of available amine groups (when using an amine-reactive dye chemistry). There is no clearly superior assay technique between direct and indirect immunolabeling, and both are currently being explored in connection with SPEF assays in the laboratory.

While any physiologically regulated molecules have the potential to serve as biomarkers, different classes of molecules present unique advantages and limitations in this regard, making the composition of biosignature candidate assays a critical consideration. Transcription factors (TFs) represent a broad class of over 2000 DNA binding proteins in humans which serve to regulate gene expression by binding specific transcription control regions of DNA. TFs assemble into enhancer complexes in vivo to create unique combinations of TFs, with the potential for each gene to be regulated by its own unique enhancer complex (22). Each TF thereby regulates numerous genes, and thus individual TFs would serve as weak biomarkers since their individual levels could be indicative of a plethora of intracellular states. Although they would serve as poor individual biomarkers, TFs are ideal biosignature candidates, since the ability to simultaneously assay the levels of numerous TFs would theoretically allow for the determination of which TFs regulate specific genes in a reproducible manner. Ideally a biosignature assay would be developed which could simultaneously assay the levels of all known TFs, providing the potential to generate comprehensive biosignatures of gene regulation in individual disease states for human cells regardless of the genes in question. For obvious cost and materials-associated reasons, it is not feasible to create such an assay at this point in time, and thus the TFs to be surveyed in a given biosignature discovery assay must be carefully selected based on knowledge of the experimental system. The primary experimental systems in use in the laboratory are mice (strains C57BL/6J, C57BL/6-Mt1tm1Bri Mt2tm1Bri, and C57BL/6-Tg(Mt1)174B ri) which have received different gene doses of metallothionein (MT).

MT is a low molecular weight cysteine rich metalloregulatory protein found ubiquitously in mammalian cells, with pleiotropic intracellular and extracellular roles in metabolic and protective processes (23). Extracellularly, MT has been implicated as serving as an antioxidant and a mediator of heavy metal toxicity, two roles with which it has also been commonly associated intracellularly (24). It has also been shown to suppress T-cell activation thereby moderating humoral immune responses, and it has been found to serve as a potential leukocyte chemoattractant, acting as a “danger signal” which promotes leukocyte movement to sites of inflammation (24,25). On an intracellular level, MT is primarily found in association with zinc molecules, but it is able to bind and subsequently sequester toxic heavy metals such as cadmium, mercury, arsenic, lead, and others (26). By sequestering these toxicants, MT serves a protective role against acute heavy metal toxicity, reducing susceptibility of mice to cadmium-induced liver injury (27).

One of the most studied intracellular functions of MT is its role in the maintenance of zinc homeostasis. Each molecule of MT can complex with seven zinc ions which bind to the sulfur-containing cysteine molecules of MT, forming a coordinated 3-dimensional structure; up to 20% of intracellular zinc may be bound to MT (28, 29). Excess free zinc levels cause protein misfolding and general cytotoxicity, whereas low levels will impair the functioning of the up to 10% of proteins that bind or interact with zinc ions, demonstrating the need for careful control of zinc availability (29). As zinc concentrations increase, MT production is induced by the zinc-dependent transcription factor MTF-1 (metal-responsive transcription factor 1), which binds to the metal responsive elements (MREs) of the MT promoter region and enhances transcription (30). MTF-1 contains six zinc finger domains and is constitutively produced from a TATA-less housekeeping promoter, indicating that the gene itself is not upregulated by excess zinc. Rather, it is proposed to activate MT transcription under such conditions as a result of zinc binding the zinc finger domains, leading to conformational changes which make the nuclear localization signal of MTF-1 available, resulting in nuclear translocation of the protein and successive transcriptional activation (30,31). This nuclear localization can also occur under other stress conditions and alone is insufficient to induce MT promoter binding, an interaction which is additionally suggested to rely on an acidic activation domain of MTF-1 (32).

Once induced, MT effectively serves as a buffer system which both maintains intracellular free zinc within physiologically critical ranges and protects cells from the detrimental effects of oxidative stress. Due to the mutual affinity of sulfur and transition metals, zinc binds to the cysteine residues of MT in a very thermodynamically stable manner which predicts that MT would be an ideal zinc reservoir but a poor buffer for free zinc concentrations as it would not readily release zinc (33). The zinc buffering capacity is thought to result from the oxidation of the cysteine residues of MT by free radicals or electrophiles, leading to the release of zinc ions from the protein (26). An MT-redox cycle has been proposed wherein three forms of MT exist: a form of MT which has bound zinc molecules, a fully oxidized form of apo-MT (thionin) which cannot accept zinc and forms intramolecular disulfide bridges, and a fully reduced form of apo-MT (thionein) which can readily accept zinc ions (34). It is unlikely that the latter two of these forms exist in significant quantities in vivo, but rather that intermediate redox forms of MT are in constant flux. Newly induced apo-MT is in the reduced thionein state, and will readily accept zinc ions unless oxidizing conditions are predominant within the cell (35). The proposed mechanism which links MT, zinc buffering, and oxidative stress is as follows: as oxidative stress levels increase within the cell, zinc will be released but will unable to re-bind to the oxidized MT. As a result, this increased free zinc concentration will induce MTF-1 binding to MREs as previously described, thereby activating the transcription and resultant translation of apo-MT/thionein which can either accept free zinc or be oxidized to reduce the oxidative stress levels within the cell (36).

As a critical regulator of zinc within cells, MT is a prime candidate for interaction with TFs which contain zinc finger domains to which the binding of zinc ions is critical; up to half of eukaryotic TFs are predicted to bind zinc ions, primarily doing so to maintain their structure (37). One crucially important zinc-binding TF is p53, a pleiotropic tumor suppressor protein which also serves antioxidant functions and may have a role in the regulation of intracellular glucose metabolism (38). p53 is composed of seven distinct domains, including two largely disordered N-terminal transactivation domains and a DNA-binding domain (DBD) (39). This DBD binds a single Zn2+ ion, stabilizing protein structure in the proper conformation for minor groove interaction. While insufficient intracellular zinc concentrations result in weak nonspecific DNA binding, excess free zinc results in equally deleterious misfolding and protein aggregation, attesting to the need for finely tuned control of zinc levels by MT and related metallochaperones (37). Recently, p53 has been found to interact with MT in vivo in breast cancer epithelial cells, where MT-p53 complexes were observed. These complexes specifically consisted of apo-MT and p53, suggesting that the sulfhydryl groups of apo-MT may bind the zinc ion in the DBD of p53, thereby inhibiting its ability to bind DNA (40). Conversely, functional p53 has been shown to be required for metal induced MTF-1-mediated activation of MT transcription, suggesting a potential negative feedback system in MT regulation and further demonstrating the complexity of MT-p53 interactions (41).

Specificity protein 1 (Sp1) is a ubiquitously expressed TF which contains three C2H2 zinc finger domains that function in nuclear import and as a DBD, making Sp1 another prime candidate TF for MT-related regulation (42). Originally identified as binding GC boxes of housekeeping genes and other genes with TATA-less promoters, Sp1 has more recently been shown to be a copper sensitive TF with broad roles in gene regulation and chromatin remodeling, blocking the spread of heterochromatin and interacting directly with certain histone deacetylases (43,44). As with p53, free zinc-bound Sp1 has been shown to bind apo-MT and prevent specific DNA binding, whereas Sp1 which has already bound DNA is protected from this interaction (45). Additionally, Sp1 and MTF-1 were found to form a co-activator complex which bind MREs and thereby regulate MT gene expression (46). This suggests that MT and Sp1 regulate one another in vivo, and while the zinc buffering properties of MT are likely an aspect of this interaction, the overall process is ultimately far more intricate, making Sp1 an ideal TF biomarker of MT activity.

An exhaustive list of zinc-binding TFs likely to interact with MT in unfeasible as it would contain hundreds of compounds; as such the following are a selection of notable zinc-sensitive TFs which have the potential to be regulated in part by MT. Sp3 is functionally and structurally similar to Sp1 and has been shown to interact with Sp1 to co-activate certain promoters while competing with Sp1 to repress others, making the Sp1:Sp3 ratio critical and necessitating careful regulation of these proteins (44). Peroxisome proliferator-activated receptor (PPAR) is a nuclear hormone receptor containing two zinc finger motifs which together form a DBD. PPARγ activation decreases the expression of members of the MT protein family, whereas zinc deficiency results in PPARγ inactivation, suggesting the mutual regulation of MT and PPAR as a function of zinc concentration (47,48). Transcription factor IIIA (TFIIIA) is a common TF that contains two C2H2 zinc finger domains; Huang et al. have found that apo-MT can bind and extract zinc from both free and DNA-bound Zn-TFIIA at physiologically relevant concentrations (49).

As MT also functions as an antioxidant, it is likely that redox-sensitive TF expression will additionally be affected by MT gene dose. AP-1 (Activator Protein 1) is a TF which was initially identified as a regulator of MT2A gene expression. It has since been characterized as a redox sensitive TF which forms homo- and hetero-dimers of basic leucine zipper proteins of the Jun, Fos, and ATF families (50,51). AP-1 functions in the regulation of numerous genes, including many involved in cellular proliferation and extracellular matrix formation (52). Under H2O2-induced oxidative stress, c-Jun but not c-Fos transcripts were found to be upregulated, demonstrating a selective induction of AP-1 subunits under redox strain (53). A more recent study suggests a mechanism whereby c-Jun is induced by heme oxygenase 1(HO-1); a protein which is rapidly induced in response to oxidative stress, heavy metals, and other stressors. Induced HO-1 is cleaved and translocates to the nucleus, resulting in phosphorylation of c-Jun and AP-1 nuclear translocation, leading to subsequent AP-1-mediated gene regulation as a function of oxidative stress (54).

NF-κB is a rapidly inducible TF and which consists of dimers of proteins of the Rel and NF-κB familes (55). Inactive NF-κB is sequestered in the cytoplasm where it is bound to proteins of the IκB family which block both the nuclear localization signal and the DNA binding cleft of the p65 (RelA) subunit (56). IκB has been found to be phosphorylated in response to oxidative stress induced by H2O2, leading to IκB dissociation and subsequent NF-κB nuclear translocation and DNA binding, resulting in the regulation of a broad spectrum of genes, including many related to immune and inflammatory responses (57,58). IκB can be phosphorylated in response to myriad sources of oxidative stress; for example, TNF-α (Tumor Necrosis Factor-α) has been demonstrated to induce NF-κB activation in a redox dependent manner (58). Additionally, Kim et al. found that increased free zinc is inhibitory of IκB dissociation and that this response can be modulated by altering levels of MT expression (59). This, coupled with the ubiquitous nature of this TF and its broad redox-responsiveness, suggest a multifaceted interplay between MT, redox strain, zinc levels, and NF-κB activation in vivo, making this and ideal TF of interest in the context of MT dose-related TF biosignature discovery.

As with zinc-sensitive TFs, it is not feasible to review every TF which is sensitive to oxidative stress, as a large number have already been identified, and far more are undoubtedly sensitive to redox stress (60). Nrf2 and Stat3 are noteworthy redox-sensitive TFs which are likely to be expressed in T-cells and are consequently likely to be directly relevant for biosignature discovery as part of this study. Nrf2 (nuclear factor E2-related factor 2) is an important regulatory and detoxification-related enzyme that largely mediates the activity of the antioxidant response element (ARE), which may be induced by a broad range of oxidative stressors. In non-stressed conditions, Nrf2 is continuously produced but is rapidly degraded due largely to Keap1, an actin-associated protein which promotes Nrf2 ubiquitinylation (61). Under oxidative stress conditions, specific cysteine molecules on Keap1 undergo alkylation and form disulfide bridges, significantly reducing Nrf2 ubiquitinylation and thus stabilizing its activity (62). Stat3 is a TF involved in the Jak-Stat signaling pathway, serving a protective role against oxidative stress (63). H2O2-induced oxidative stress has been found to lead to Stat3 nuclear translocation and subsequent gene regulation in human T lymphocytes (64).

While these aforementioned transcription factors represent a broad sample of TFs expected to have their regulation affected by MT gene dose, it is clear that many other TFs are similarly altered by changes in intracellular redox states and zinc equilibrium. As this project seeks to confirm the potential of TF biosignatures of MT gene dose, it will be important for future studies to expand the SPEF microarrays performed herein in order to establish a clearer picture of the role of MT in transcriptional regulation as a whole.

In this project, TF biosignatures were generated using directly Alexa Fluor-647-labeled splenocyte lysates from mice with different gene-doses of MT in tandem with the SPEF microarray technology. The goal of this study was three-fold: to assess the efficacy of SPEF microarrays as means of generating reproducible biosignatures, to establish the reliability of TF biosignatures of MT gene dose, and to determine the feasibility of using a directly labeled fluorescent immunoassay approach for such novel applications.

If successful, this project will offer insight into the role of MT in transcriptional regulation, with the potential for novel drug applications and therapeutic approaches. Additionally, success will verify the capability of SPEF technology to generate reproducible biosignatures in a relatively low-cost manner with minimal sample volume requirements. The clinical applications of this technology are obvious, with the immediate possibility of advance disease screening and diagnosis, and the long term potential of implementing dried blood spot samples to assess changes in toxicant exposure within individuals over extended periods of time.


Antibodies and Reagents

Mouse MOPC-141 IgG2b, teleostean fish gelatin, and ovalbumin were obtained from Sigma. Rabbit anti-ovalbumin was purchased from Chemicon International. Monoclonal mouse antibodies against SP-1, ssDNA, and dsDNA were purchased from Millipore. Mouse anti-c-Jun, rabbit anti NF-kB p65, goat anti NF-kB p50, and rabbit anti NF-kB cRel were purchased from Santa Cruz Biotechnology. Monoclonal mouse antibodies against human CD4 and human CD3e were purchased from R+D Systems. Mouse anti-NF-kB p65 was purchased from Abcam. Mouse anti-human CD28 was purchased from BD Pharminigen. DMSO was purchased from Fisher Scientific. Mammalian Protein Extraction Reagent (M-PER) and dithiobis (succinimidyl) propionate (DSP) were purchased from Thermo Scientific. Biogel P10 size exclusion resin was purchased from Biorad. Complete protease inhibitor cocktail tablets were purchased from Roche Diagnostic.


The mouse strain C57BL/6J was received from Jackson Laboratory (Bar Harbor, ME). Mice were maintained in sterile laminar flow cages on 12 hour day/night cycles with free access to food and water.

Cell Cuture

Human Jurkat T Lymphocytes were cultured in RPMI 1640 containing 10%FBS, 12 mM HEPES, 4mM L-glutamine, 100μg/mL penicillin, and 100μg/mL streptomycin. Cells were grown in at 37oC in 5.0% CO2 conditions in an incubator (Fisher Scientific) and were passed regularly.

T-Cell Stimulation

Jurkat T lymphocytes or mouse splenocytes were brought to a concentration of 106 cells/mL in complete RPMI media, and 1mL of cells were added to wells of a 24-well plate. To each well, Concanavalin A was added at concentrations of 10μg/mL, 5μg/mL, or 1μg/mL and cells were incubated for 24 hours. Stimulation was confirmed by collecting supernatant from these wells after 24 hours, removing extant cells by centrifugation at 1000rpm for 10 minutes, and testing this supernatant for elevated levels of IL-2 and IFN-γ using an ELISA.

ELISA Confirmation of T-Cell Stimulation

96-well plates were incubated overnight at 4oC with antibodies specific to human IL-2 or IFN-γ at concentrations recommended by the manufacturer (COMPANY). Plates were then washed 3 times with PBST and blocked for 2 hours at 37oC with 2% bovine serum albumin. Plates were again washed 3 times with PBST and protein standards or supernatant from stimulated cells, unstimulated cells, or media control were incubated on the plate for one hour at 37oC. Plates were then washed 3 times with PBST and a biotinylated secondary antibody specific to IL-2 or IFN-γ was incubated for one hour at 37oC at concentrations recommended by the manufacturer (COMPANY). Plates were washed 3 times with PBST and incubated with horseradish peroxidase-streptavidin for 20 minutes at room temperature. Plates were washed a final 3 times and were then incubated with color developing reagents (COMPANY) for 20 minutes. The reaction was then terminated used 1N H2SO4 and the plate was read at 450nm.

Cell Lysis

Aliquots of Jurkat lymphocytes or murine splenocytes were assessed for viability by mixing 20μL of cells with 20μL of trypan blue dye (COMPANY), which is readily taken up by dead but not live cells. Cells were then counted using a hemocytometer in order to determine viability and overall cell titer. Murine splenocytes were additionally counted using a (Z2 – COMPANY) in order to confirm cell count. Samples of 10 million cells were then pelleted at 1000rpm for 10 minutes in a (COMPANY) centrifuge, and resuspended in 500μL mammalian protein extraction reagent (M-PER; COMPANY) with 50μL of 10X protease inhibitor cocktail (COMPANY) for 15 minutes. Lysed samples were then centrifuged at 3000rpm for 3 minutes in order to pellet any cellular debris, and the supernatant was saved for future labeling reactions and analysis. Initially cells were lysed with Triton X-100 before it was established that M-PER produced higher protein levels in the lysate mixture in a shorter period of time; M-PER was used for all subsequent lysis protocols.

Fluorescent Labeling Reactions

Proteins used in these experiments were labeled using amine-reactive Alexa Fluor 647 carboxylic acid, succinimidyl ester, purchased from Invitrogen in a 1mg solid state. The dye was dissolved in 250μL DMSO and separated into 10μL aliquots which were stored at -20oC as described previously (65). For the labeling reaction, 90μL of dH2O was added to a 10μL Alexa Fluor 647 aliquot, and 30μL of this dye solution was added to 500μL of protein solution along with 72.2μL of 1M sodium bicarbonate(Fisher Scientific) for a final concentration of 0.12M sodium bicarbonate. Further experimental optimization demonstrated that maximum labeling occurs after approximately 4 hours labeling at room temperature or 24 hours labeling at 4oC, after which time samples were run through a column of Biogel p10 size exclusion resin and collected as 500μL fractions in order to separate out the unconjugated dye. Fraction profiles of fluorescence activity and protein concentration were generated using Immulon 2 microplates (Thermo Scientific) and BCA protein assay kits (Thermo Scientific). Plates were read using a ThermoMax microplate reader (Molecular Devices), and data was collected using the SOFTMAX software (Molecular Devices).

Assessment of Labeling Efficiency

The efficiency of Alexa Fluor 647 protein labeling was approximated by printing out the aforementioned fraction profiles on uniform paper and cutting out the area under the fluorescence curve. This area was then split into two peaks, with one peak encompassing the high molecular weight protein-containing fractions, and the other encompassing the remainder of the curve. Both peaks were massed, ad the mass of the protein peak relative to the combined mass of both peaks was used as an estimate of the percentage of dye bound to protein.

Gel-based Labeling Assessment

A gel based assessment of fluorescent protein labeling was conducted by pouring a 12% SDS-PAGE separating gel with a 4% stacking gel according to standard lab recipes. Samples of Alexa Fluor-647 labeled Jurkat lysate, Alexa Fluor-647 labeled ovalbumin, unlabeled ovalbumin, and combinations of the three were heated to 95oC for 10 minutes and then loaded into wells as 15μL aliquots with 5uL Tris loading buffer. Rainbow molecular weight markers were loaded as a 16μL aliquot with 4μL Tris loading buffer. The gel was run at 200V until the loading dye reached the bottom, at which point the gel was removed and transferred to a PVDF membrane using an Invitrogen iBlot sytem. The membrane was then imaged on a Kodak Image Station 2000MM (Carestream Health, Inc.) instrument using a 635nm excitation filter and a 700nm wide angle emission filter.

SPEF Chip Preparation

SPEF gold sensor chips ) were cleaned using alternating washes of 70% ethanol and dH2O and dried using an air flow. Capture antibodies and control molecules were added to a 384-well microplate (10μL/well) from which they were deposited as individual ROIs on the sensor chip using (ArrayIt Microarray Technology) To improve antibody binding to the chip surface, chips were first coated with 300μL of 4mg/mL DSP dissolved in DMSO for 30 minutes, washed with DMSO and dH2O, dried with an air flow, and then spotted immediately as above. After the spotting process was complete, chips were incubated for 1 hour in a humidified environment before being stored at 4oC.

SPEF Assays

Printed chips were first assembled and placed in the dual mode SPEF instrument ) in order to produce an ROI template to be used for ROI visualization after the assay is complete. Chips were then unassembled and placed in a custom-built benchtop fluidics system. Chips were first blocked for 30 minutes with 5% teleostean fish gelatin at a flow rate of 1mL/min, after which they were washed with PBST for 5 minutes at 1mL/min. An Alexa Fluor 647-labeled sample was then recirculated over each chip for 1 hour at 0.5mL/min, followed by a 20 minute PBST wash at 1mL/min. Chips were then removed from the fluidics machine, assembled, and filled with 100uL PBST. Later, chips were assembled for ROI visualization and were then loaded into a desktop fluidics system (Masterplex) while still assembled, allowing for simultaneous sample exposure of up to 8 chips at once. The assay then proceeded as described above. Assembled chips were then placed into the dual mode SPEF machine and visualized using a 2-15s laser exposure to excite the fluorescent molecules, such that the fluorescence of a given spot on the chip was correlated to the amount of antigen present on the spot based on the antibody ROI template.


Optimization of Lysis/Labeling Procedures

The Alexa Fluor 647 dye was selected for use the SPEF technology in the context of whole cell lysates because this dye had been previously used for SPEF applications with marked success. Initially, samples of multiple concentration of ovalbumin were labeled using Alexa Fluor 647 protein labeling kits in order to assess the labeling efficiency of the dye relative to the concentration of protein being labeled. As demonstrated in Figure 2A, the efficiency of labeling decreased as protein concentration increased, likely indicating that the high protein concentrations had saturated the solution. This result led future use of lower protein concentrations in labeling reactions to maintain an excess of dye and thereby prevent reductions in labeling efficiency.

In order to establish the approximate ratio of bound/free dye after labeling reactions, free dye was separated from labeled protein using a 17mL Biogel P10 size exclusion resin column and samples were collected as 0.5mL fractions. These fractions were then assessed using a BCA assay to establish which fractions contained protein and an endpoint fluorescence read (650nm excitation, 668nm emission) using a microplate reader in order to establish which fraction contained fluorescent dye. In each resultant fraction profile (see Figure 3) a distinct peak in protein concentration comigrated with a distinct fluorescence peak, strongly suggesting that these fractions contain labeled protein. The percentage fluorescence in this peak as a fraction of the total fluorescence was approximated as outlined in Materials and Methods. In initial experiments, the percentage of dye bound to protein was found to be ~30% (data not shown), but after further optimization this was increased to ~60% (Figure 2B).

Note that in whole cell lysates (see Figure 3) but not in ovalbumin, small amounts of protein were found to migrate with the low molecular weight free dye, suggesting small amounts of latent protease activity despite the addition of complete protease inhibitors.

The effects of time and pH on labeling efficiency were simultaneously assessed by preparing nine equivalent aliquots of ovalbumin which contained sodium bicarbonate concentrations of 0.8M, 1.0M, or 1.2M, and which were labeled for either 1 or 4 hours at room temperature or overnight at 4oC. A fraction profile was generated for each sample in order to establish labeling efficiency (Figure 2B). Based on these results, all future experiments were conducted with labeling periods of ~ 24 hours at 4oC, and with sodium bicarbonate concentrations of 0.12M.

Cell lysis protocols were optimized by assessing the relative protein yields of two lysis reagents: Triton X-100 and Mammalian Protein Extraction Reagent (M-PER, a proprietary non-denaturing amine-free lysis detergent). Equivalent aliquots of Jurkat cells were lysed with the two detergents according to their respective manufacturer’s protocols. It was established that M-PER yielded slightly more protein than Triton X-100 as established using a BCA assay, but the M-PER detergent volume was greater resulting in a more dilute solution (data not shown). M-PER was selected for all future experiments as the lysis process was shorter (15 minutes as opposed to 40 minutes for Triton X-100), yielded more protein, and was stated to extract both nuclear and cytoplasmic protein, a necessity in an assessment of transcription factor activity.

Gel Based Labeling Assessment

After optimizing the labeling procedures it was necessary to visualize the extent of cell lysate labeling in order to confirm that the labeling process did not favor proteins of a specific molecular weight. To this effect, an SDS-PAGE was loaded with Alexa Fluor 647 labeled samples of Jurkat lysate and/or ovalbumin as described in Materials and Methods (Figure 4). The resultant gel suggests relatively uniform protein labeling by Alexa Fluor 647, strengthening the conclusion that the concentration of free dye during labeling is in significant excess of protein concentration thereby allowing for labeling to occur based solely on the number of primary amines present on individual protein.

SPEF-based Transcription Factor Detection in Jurkat Cells

Initial SPEF assays were performed using Alexa Fluor 647-labeled ovalbumin to both confirm the efficacy of the technology and to optimize experimental protocols. Rabbit anti-ovalbumin and irrelevant rabbit IgG were spotted on a gold chip as described in Materials and Methods in dilutions from 20ug/mL to 125ng/mL. After the chip was blocked with 5% telostean fish gelatin for 20 minutes, washed for 5 minutes with PBST, exposed to sample, and washed again for 5 minutes with PBST it was imaged using the dual-mode SPEF. The resultant data revealed that only the highest concentration of rabbit anti-ovalbumin antibody had produced a fluorescent signal but that the control IgG had produced no signal at any concentration (data not shown). As a result all future experiments used antibodies spotted at higher concentrations (0.2mg/mL to 0.5mg/mL).

Early SPEF experiments produced significant background fluorescence which reduced the sensitivity of this microarray technique. To reduce this noise, the block concentration and lengths of blocking/washing were altered, until the final optimized protocol present in Materials and Methods was established (data not shown).

Antibodies to SP-1, the c-Jun subunit of AP-1, and NF-kB p65subunit were spotted on chips with a range of positive and negative controls, and these chips were exposed to samples of Alexa Fluor 647-labeled lysates of 10 million Jurkat cells. Initial assays detected elevated levels of c-Jun in these whole cell lysates (data not shown). In order to confirm that this detection was specific for c-Jun rather than an artifact of lysis/labeling protocols, samples of Jurkat lysate were either left untreated or incubated with either anti-c-Jun or anti-dsDNA before being recirculated over printed chips (Figure 5). The resultant data demonstrate that the detection of c-Jun was specific, as the c-Jun signals is significantly reduced after incubation with anti-c-Jun whereas other signals remain constant regardless of incubation conditions. The fact that anti-dsDNA did not remove all anti-dsDNA signal is likely a consequence of the fact that DNA is present at significantly higher levels intracellularly that are individual transcription factors, and as a consequence the concentration of anti-dsDNA was insufficient to bind all available epitopes during the incubation period.

SPEF-Based Assessment of Transcription Factors in Murine Splenocytes

After successfully demonstrating specific TF detection in Jurkat cell lysates we endeavored to first replicate these findings in murine splenocytes and to then generate reproducible TF biosignatures in these cells. Initial experiments established detectable levels of NF-kB p65 subunit in the three mouse strains (Figure 6). These results remain to be confirmed using competition experiments, however they suggest that we will be able to detect TFs specifically in mouse cells and thus we will be able to ascertain reproducible TF biosignatures of MT gene dose using SPEF technology.

The implications of constant-time configurations have been far-reaching and pervasive. The notion that security experts connect with the investigation of Scheme is regularly adamantly opposed. On a similar note, a typical challenge in operating systems is the improvement of pseudorandom models. To what extent can congestion control be investigated to fulfill this goal?

Another structured objective in this area is the exploration of knowledge-based information. Indeed, evolutionary programming and superblocks have a long history of collaborating in this manner. Existing wireless and electronic applications use client-server epistemologies to study probabilistic models. Indeed, online algorithms and multi-processors have a long history of interacting in this manner.

Our focus here is not on whether hierarchical databases can be made wearable, real-time, and “smart”, but rather on proposing a novel application for the refinement of Smalltalk (Citrate). Further, indeed, fiber-optic cables and operating systems have a long history of agreeing in this manner. By comparison, it should be noted that Citrate harnesses modular modalities. Although similar applications explore thin clients, we overcome this problem without developing journaling file systems.

In this work, we make two main contributions. To start off with, we argue not only that compilers and A* search can connect to address this grand challenge, but that the same is true for gigabit switches. We prove that multi-processors and access points are always incompatible.

The rest of the paper proceeds as follows. Primarily, we motivate the need for superblocks. We place our work in context with the existing work in this area. To surmount this challenge, we explore a framework for the partition table (Citrate), which we use to confirm that the producer-consumer problem and consistent hashing can agree to surmount this problem. Furthermore, to accomplish this aim, we describe an analysis of Moore's Law (Citrate), arguing that DHCP can be made stochastic, symbiotic, and autonomous. Ultimately, we conclude.

2 Related Work

In this section, we discuss previous research into thin clients, the synthesis of IPv6, and IPv7 [27]. Along these same lines, the choice of the Internet in [27] differs from ours in that we investigate only key information in our heuristic [27]. Sasaki and Lee and Davis [4] proposed the first known instance of write-back caches [18] [14]. We plan to adopt many of the ideas from this previous work in future versions of Citrate.

2.1 Virtual Technology

The investigation of trainable technology has been widely studied [2]. Our application is broadly related to work in the field of electrical engineering by Williams et al. [7], but we view it from a new perspective: forward-error correction [27]. Further, a litany of previous work supports our use of Byzantine fault tolerance [18]. Similarly, Thompson and Nehru originally articulated the need for gigabit switches. Despite the fact that we have nothing against the previous method [5], we do not believe that solution is applicable to algorithms. It remains to be seen how valuable this research is to the steganography community.

While we are the first to motivate the investigation of 802.11b in this light, much existing work has been devoted to the synthesis of cache coherence. In this paper, we addressed all of the problems inherent in the previous work. Garcia and Zhou presented several wireless approaches, and reported that they have great inability to effect multimodal information. Furthermore, even though Ron Rivest et al. also presented this solution, we constructed it independently and simultaneously [17,26,1,27]. Although Ron Rivest et al. also described this method, we emulated it independently and simultaneously [11]. Obviously, the class of approaches enabled by our algorithm is fundamentally different from prior approaches [32]. On the other hand, the complexity of their solution grows sublinearly as optimal technology grows.

2.2 Agents

The visualization of wireless algorithms has been widely studied [19]. Though Wilson and Sato also proposed this method, we enabled it independently and simultaneously. We believe there is room for both schools of thought within the field of software engineering. Further, an analysis of kernels [6,24] proposed by Michael O. Rabin fails to address several key issues that Citrate does surmount [30]. We plan to adopt many of the ideas from this existing work in future versions of Citrate.

2.3 Information Retrieval Systems

A major source of our inspiration is early work by Nehru et al. on self-learning epistemologies. Along these same lines, the choice of telephony in [19] differs from ours in that we measure only robust configurations in Citrate. Recent work by John Hennessy et al. suggests a framework for constructing the investigation of voice-over-IP, but does not offer an implementation [12]. Despite the fact that we have nothing against the existing solution by Shastri et al. [29], we do not believe that method is applicable to e-voting technology. Citrate also is Turing complete, but without all the unnecssary complexity.

A number of previous methodologies have simulated telephony, either for the exploration of 32 bit architectures [26] or for the deployment of the UNIVAC computer that would allow for further study into semaphores. We had our method in mind before Davis et al. published the recent seminal work on e-commerce [20,16,21,28]. Recent work by Zheng suggests an application for allowing web browsers, but does not offer an implementation [4,8,23,6,21]. Furthermore, unlike many prior methods, we do not attempt to control or emulate suffix trees [10]. This solution is more cheap than ours. Similarly, our heuristic is broadly related to work in the field of artificial intelligence by Sun et al., but we view it from a new perspective: forward-error correction [3]. Despite the fact that we have nothing against the previous method by S. Abiteboul, we do not believe that approach is applicable to cryptoanalysis [15].

3 Framework

On a similar note, we show the schematic used by Citrate in Figure 1. Despite the fact that electrical engineers usually postulate the exact opposite, our method depends on this property for correct behavior. Consider the early architecture by Sasaki and Qian; our design is similar, but will actually fix this obstacle. This may or may not actually hold in reality. The model for our algorithm consists of four independent components: event-driven modalities, the exploration of access points, collaborative algorithms, and simulated annealing. As a result, the design that our framework uses is unfounded.

Figure 1: Our framework's real-time creation.

Citrate relies on the unproven design outlined in the recent infamous work by T. Taylor et al. in the field of operating systems. Any confirmed exploration of I/O automata [22] will clearly require that architecture and replication are continuously incompatible; Citrate is no different. We assume that embedded modalities can request lossless theory without needing to prevent the UNIVAC computer. Similarly, we instrumented a 1-minute-long trace arguing that our framework holds for most cases.

4 Implementation

It was necessary to cap the latency used by our methodology to 88 GHz. Our framework requires root access in order to enable Lamport clocks. Since Citrate runs in Θ(n!) time, designing the homegrown database was relatively straightforward. Along these same lines, the client-side library and the hand-optimized compiler must run in the same JVM. such a hypothesis might seem unexpected but has ample historical precedence. We have not yet implemented the codebase of 84 Fortran files, as this is the least private component of Citrate. Since our system is NP-complete, optimizing the hacked operating system was relatively straightforward.

5 Experimental Evaluation

As we will soon see, the goals of this section are manifold. Our overall evaluation seeks to prove three hypotheses: (1) that SMPs no longer impact an application's code complexity; (2) that Byzantine fault tolerance have actually shown weakened median time since 1977 over time; and finally (3) that the PDP 11 of yesteryear actually exhibits better time since 1995 than today's hardware. An astute reader would now infer that for obvious reasons, we have intentionally neglected to construct RAM speed. Our performance analysis will show that reprogramming the seek time of our mesh network is crucial to our results.

5.1 Hardware and Software Configuration

Figure 2: The mean sampling rate of our algorithm, compared with the other heuristics.

Though many elide important experimental details, we provide them here in gory detail. We executed a large-scale deployment on MIT's system to prove the independently cacheable nature of randomly classical symmetries. We removed more ROM from CERN's system. On a similar note, we reduced the effective NV-RAM speed of the NSA's human test subjects to quantify N. Moore's improvement of randomized algorithms in 1953. On a similar note, we tripled the median block size of our adaptive cluster. Note that only experiments on our network (and not on our decommissioned Nintendo Gameboys) followed this pattern.

Figure 3: The mean seek time of Citrate, compared with the other heuristics.

We ran Citrate on commodity operating systems, such as AT&T System V Version 7.5, Service Pack 4 and GNU/Debian Linux. Our experiments soon proved that microkernelizing our wired Commodore 64s was more effective than instrumenting them, as previous work suggested. Our experiments soon proved that microkernelizing our Apple Newtons was more effective than distributing them, as previous work suggested. Similarly, all software components were hand assembled using AT&T System V's compiler with the help of Adi Shamir's libraries for opportunistically controlling exhaustive floppy disk space. Such a claim might seem perverse but is buffetted by related work in the field. This concludes our discussion of software modifications.

Figure 4: Note that block size grows as interrupt rate decreases - a phenomenon worth simulating in its own right.

5.2 Dogfooding Our Algorithm

Figure 5: The average complexity of Citrate, compared with the other methods.

Figure 6: The average complexity of Citrate, compared with the other frameworks.

We have taken great pains to describe out evaluation setup; now, the payoff, is to discuss our results. That being said, we ran four novel experiments: (1) we measured WHOIS and E-mail latency on our mobile telephones; (2) we ran 30 trials with a simulated E-mail workload, and compared results to our middleware simulation; (3) we measured hard disk throughput as a function of ROM speed on a LISP machine; and (4) we deployed 29 Commodore 64s across the sensor-net network, and tested our B-trees accordingly. All of these experiments completed without WAN congestion or 10-node congestion.

Now for the climactic analysis of all four experiments. Error bars have been elided, since most of our data points fell outside of 99 standard deviations from observed means. Continuing with this rationale, the curve in Figure 4 should look familiar; it is better known as H*(n) = n. On a similar note, bugs in our system caused the unstable behavior throughout the experiments.

We have seen one type of behavior in Figures 2 and 6; our other experiments (shown in Figure 3) paint a different picture. Gaussian electromagnetic disturbances in our 100-node testbed caused unstable experimental results. These signal-to-noise ratio observations contrast to those seen in earlier work [27], such as Isaac Newton's seminal treatise on kernels and observed effective RAM space. Similarly, note how emulating Byzantine fault tolerance rather than emulating them in hardware produce less jagged, more reproducible results.

Lastly, we discuss the second half of our experiments. The many discontinuities in the graphs point to weakened complexity introduced with our hardware upgrades. Continuing with this rationale, operator error alone cannot account for these results. Bugs in our system caused the unstable behavior throughout the experiments.

6 Conclusion

We argued in this paper that the infamous pseudorandom algorithm for the visualization of B-trees by Garcia and Garcia [25] runs in O( log( n + n ) ) time, and our methodology is no exception to that rule [9,13,17]. In fact, the main contribution of our work is that we constructed new introspective models (Citrate), which we used to demonstrate that the Ethernet can be made pervasive, adaptive, and signed. We motivated an analysis of SCSI disks (Citrate), showing that the foremost game-theoretic algorithm for the deployment of Byzantine fault tolerance by Miller [31] follows a Zipf-like distribution. Thusly, our vision for the future of robotics certainly includes Citrate

Discussion and Future Directions

As a whole our results exhibit the promise of SPCE as a means of biosignature identification through its ability to reaffirm previously observed differences in transcription factor expression in mice expressing different levels of metallothionein.

Our SDS-PAGE analysis of the labeling of a whole cell lysate by Alexa Fluor-647 suggests the nonspecific labeling of proteins within the lysate, seemingly indicating that the fluroescent labeling process was not restricted to proteins certain of a certain conformation or molecular weight (Figure 2). This result would be strengthened by running this fluorescent gel in tandem with an identical gel treated with a silver stain or another means of nonspecific protein labeling. Identical banding patterns on the two resultant gels would further confirm that ubiquitous fluorescent labeling had truly taken place without bias towards a particular protein subset. It is important to note, however, that in addition to labeling the N-terminus of each protein the Alexa Fluor-647 dye will label the primary amine groups on accessible lysines, and as such the DOL may vary between proteins. Lysine-rich proteins will appear disproportionately bright during subsequent fluorescent analysis. If this is a concern then the dye/protein (F/P) ratio can be calculated for a specific protein of interest in order to standardize experimental results, as was done for AF-647 and AF-680 in later experiments (Figure 6B). Consistent with this concern, previous studies have confirmed that different cytokines will produce differentially bright fluorescent signals in direct, but not in indirect, fluorescence immunoassays, likely due to the differential labeling of these cytokines (Li, 2003). Indirect sandwich immunoassays can be used in lieu of direct labeling for SPCE experiments, thus allowing for fluorescent amplification that remains unbiased by the number of primary amines present on each protein at the cost of reduced time-per-assay (Sanchez-Carbayo, 2006).

The result of our experimental conformation of transcription factor detection demonstrates the specificity of analyte detection using the anti-cJun antibody (Figure 3), however it does not necessarily confirm that the analyte being detected is cJun. This result establishes that the protein which binds to this antibody is specific and consistent, even though it could theoretically be an irrelevant protein that happens to present a similar epitope or otherwise have a high affinity for this antibody. If the biological activity of cJun were of significant interest then it would be important to independently verify that the elevated levels of cJun in a given sample correspond to a true increase in cJun expression. This verification could be conducted by using siRNA to knock down cJun activity in cells of interest and comparing the fluorescent signal at cJun ROIs in siRNA-treated and untreated cells via SPCE. Alternatively, an ELISA could be performed using different cJun antibodies in order to independently quantify the levels of this transcription factor. Nonetheless, the results of this experiment demonstrated that the fluorescence signals detected during SPCE analysis are specific, thus lending further credence to this microarray analysis technique and to the pursuit of TF biosignatures.

Although Jurkat cells were stimulated following treatment with PHA as evidenced by an increase in IL-2 secretion (Figure 4A), no robust differences in transcription factor expression were observed when whole cell lysates of these cells were prepared and analyzed by SPCE (Figure 4B). This result is not entirely surprising as Jurkat cells are immortalized T cells which may respond to environmental stimuli in an atypical manner due to an accumulation of mutations over time. When these experiments were replicated using splenocytes from each of three mouse strains, more promising differences in transcription factor activity were observed (Figure 5). An increase in NF-kB p50 expression in MT-KO mice was consistently observed in two independent experiments. These differences were correlated with both mouse strain and stimulation conditions, and the observed increase in NF-kB p50 expression in MT-KO mice is consistent with previously documented results (Crowthers, 2000). These data also suggest the possibility of increased c-Rel production in ConA stimulated MT-KO splenocytes relative to other cell populations, however this result was not consistently observed and thus warrants further investigation. This confirmation of previously identified transcription factor expression patterns demonstrates the versatility of the SPCE technology and substatiates it as a means of detecting small differences in protein expression within complex biological samples.

This confirmation is an important step towards the validation of SPCE as a means of large scale characterization of disease susceptibility, however the failure to observe consistent differences in the majority of transcription factors surveyed suggests that further improvements may be necessary. Expanding the number of unique transcription factors analyzed on each SPCE chip would improve the probability of observing changes in TF expression patterns. Ideal candidate TFs for analysis in mice expressing different levels of MT are those which rely on zinc for their structure and/or are susceptible to oxidative stress (Table 1). During T cell stimulation experiments, assessing the levels of transcription factors which are closely tied to T cell activation such as NFAT may prove particularly valuable.

In the abovementioned experiments, whole cell lysates were assessed for the overall presence of TFs, with the resulting fluorescent signals serving as indicator sof the total amount of each TF within the cell (both cytoplasmic and nuclear proteins). For many TFs, however, regulatory activity is not controlled by basal expression levels but rather by protein localization. For example, the majority of NFAT is located in the cytoplasm where it remains in a phosphorylated state. When calcineurin becomes activated by upstream signaling events it dephosphorylates NFAT, leading to a conformational change which reveals a nuclear localization sequence that results in the transport of NFAT into the nucleus where it regulates gene expression (Macian, 2005). Similarly, inactive NF-κB p65 is sequestered in the cytoplasm where it is bound to the inhibitory molecule IκB (Jacobs, 1998). IκB dissociates from NF-κB following its phosphorylation as a result of oxidative stress, thereby allowing NF-κB to undergo nuclear translocation and bind to DNA (Zampieri, 2009). As such, measuring TF levels in a whole cell lysate may not be the best means of TF biosignature characterization, as the resultant pattern does not distinguish between nuclear and cytoplasmic proteins. These biosignatures may thus fail to yield the desired information regarding differential TF activity, instead only providing insight into total TF expression levels. This issue can best be addressed by purifying the nuclei from cells and lysing them in the absence of cytoplasmic proteins, thereby allowing for the quantification of only the active nuclear forms of transcription factors of interest. Nuclear purification may be accomplished using a combination of gentle detergents and density-gradient ultracentrifugation. Altering the cell lysis protocols to focus solely on nuclear lysates would likely yield more biologically relevant information on TF activity as it corresponds to MT gene dose, possibly improving the sensitivity of the assay and allowing for more versatile and meaningful applications of the SPCE technology.

Due to both the highly reactive nature of Alexa Fluor-647-NHS and the large volumes of extraneous protein present in whole cell lysate samples, great care was taken over the course of these experiments to reduce nonspecific fluorescent binding to the SPCE chips. Nonetheless, background fluorescence was typically visible at greater levels than in similar sandwich SPCE immunoassays, resulting in an overall decrease in the signal:noise ratio. This is not entirely surprising given that indirect immunoassays are generally associated with gains in signal:noise ratio and improved sensitivity relative to direct labeling applications (Li, 2003). Potential avenues for reducing irrelevant include the removal of irrelevant proteins - particularly “sticky” proteins such as actin which are likely to aggregate on chip surfaces. Running whole cell lysates through a column containing anti-actin antibodies could significantly reduce these nonspecific interactions, and may prove to be a worthwhile endeavor for future biosignature discovery assays.

Even as we work to refine this iteration of the SPCE machinery, another version is being developed which is able to capture spectrally distinct fluorescent signals from two fluors on a single SPCE chip. This dual-fluor SPCE instrument will be able to excite both Alexa Fluor-647 and Alexa Fluor-680 using a single 635nm laser, and will be able to separately capture and quantify images of the two resulting emission signals on a single chip. This innovation will allow for two distinct biological samples to be captured on a chip simultaneously. The relative levels of a protein of interest in these two samples can then determined based upon the ratio of AF647:AF680 at a particular ROI on the chip. This approach to SPCE experiments would save both time and money, allowing for rapid comparisons of differentially treated samples while using half the resources of a traditional SPCE assay. This technique has the additional advantage of eliminating variability between chips when comparing samples, thereby providing more reliable results ideal for diagnostic scenarios. Our results herein demonstrate some of the challenges of using a dual-fluor system due to the reduced signal strength of Alexa Fluor-680 relative to Alexa Fluor-647(Figure 6). Nonetheless, these results suggest that there is minimal spectral overlap between these two fluors, which is critical for the success of this technology. With proper optimization and meticulously designed filters this dual-fluor SPCE machine will open new avenues for diagnostic protein microarrays in the near future.

In summary, this study has demonstrated the power and versatility of SPCE as a means of biosignature characterization, and has suggested the feasibility of interrogating TF biosignatures. While these results will benefit from the future expansion of the candidate TF pool and from the elimination of nonspecific fluorescence, they confirm that SPCE is capable of detecting fluctuations in TF expression and further emphasize the role played by MT in the regulation of zinc availability and consequently in the regulation of transcriptional activity. While this protein microarray is still in the development phase, this and related studies underscore its potential as a means of providing low-cost multiplexed diagnostic capabilities, with the potential to improve quality of life for countless individuals each year.


  • 1. LaBaer J, Ramachadran N. Protein microarrays as tools for functional proteomics. Current Opinion in Chemical Biology. 2005; 9:14-19
  • 2. Schweitzer B, Kingsmore SF. Measuring proteins on microarrays. Current Opinion in Biotechnology. 2002
  • 3. Quintana FJ, et al. Functional immunomics: Microarray analysis of IgG autoantibody repertoires predicts the future response of mice to induced diabetes. PNAS. October 5, 2004; 101(Suppl 2): 14615-14621.
  • 4. Parikh CR, et al. Urinary IL-18 is an early predictive biomarker of acute kidney injury after cardiac surgery. Kidney International. 2006; 70:199–203.
  • 5. Kozak KR, et al. Identification of biomarkers for ovarian cancer using strong anion-exchange ProteinChips: Potential use in diagnosis and pr ognosis. PNAS. October 14, 2003; 100 (21): 12343-12348.
  • 6. Parida SK, Kaufmann SHE. The quest for biomarkers of tuberculosis. Drug Discovery Today. 2010; 15(3-4):148-157.
  • 7. Shaw LM, et al. Cerebrospinal Fluid Biomarker Signature in Alzheimer’s Disease Neuroimaging Initiative Subjects. Annals of Neurology.2009; 65(4): 403-413.
  • 8. Sagoo P, et al. Development of a cross-platform biomarker signature to detect renal transplant tolerance in humans. The Journal of Clinical Investigation.2010; 120(6): 1848-61.
  • 9. Haak PT, et al. Archived unfrozen neonatal blood spots are amenable to quantitative gene expression analysis. Neonatology. 2009 ; 95(3): 210–216.
  • 10. Janzi M, et al. Screening for C3 Deficiency in Newborns Using Microarrays. PLoS ONE. 4(4): e5321.
  • 11. Zhang L, et al. A strategy for the development of biomarker tests for PTSD. Medical Hypotheses. 2009; 73(3):404-409.
  • 12. Bottein MY, et al. Biomonitoring of ciguatoxin exposure in mice using blood collection cards. Toxicon. 2005; 46(3):243-251.
  • 13. Preuss R, Angerer J, Drexler H. Naphthalene—an environmental and occupational toxicant. Int Arch Occup Environ Health. 2003; 76:556–576.
  • 14. Unfricht DW, Colpitts SL, Fernandez SM, Lynes MA. Grating-Coupled SPR: A cell and protein microarray platform. Proteomics. 2005; 5: 4432-4442.
  • 15. Homola J, Yee SS, Gauglitz G. Surface plasmon resonance sensors: review. Sensors and Actuators B: Chemical. 1999; 54: 3-15.
  • 16. Jin G, Unfricht DW, Fernandez SM, Lynes MA. Cytometry on a chip: Cellular phenotypic and functional analysis using grating-coupled surface Plasmon resonance. Biosensor and Bioelectronics. 2006; 22: 200-206.
  • 17. Reilly MT, et al. SPR surface enhanced fluorescence with a gold-coated corrugated sensor chip. Proc SPIE. 2006; 6099: 58-66.
  • 18. Tawa K, et al. Optical microscopic observation of fluorescence enhanced by grating-coupled surface plasmon resonance. Optics Express. 2008; 16(13): 9781-9790.
  • 19. Li Y, Nath N, Reichert WM. Parallel Comparison of Sandwich and Direct Label Assay Protocols on Cytokine Detection Protein Arrays. Anal. Chem. 2003; 75: 5274-5281.
  • 20. Brown JK, et al. Primary Antibody–Fab Fragment Complexes : A Flexible Alternative to Traditional Direct and Indirect Immunolabeling Techniques. Journal of Histochemistry and Cytochemistry. 2004; 52 (9): 1219-1230.
  • 21. Sanchez-Carbayo M. Antibody Arrays: Technical Considerations and Clinical Applications in Cancer. Clinical Chemistry. 2006; 52(9): 1651-1659.
  • 22. Lodish H, et al. Molecular and Cell Biology. New York, NY: WH Freeman and Company. 2008; p. 286-296.
  • 23. Vašák M. Advances in metallothionein structure and functions. Journal of Trace Elements in Medicine and Biology. 2005; 19(1): 13-17.
  • 24. Lynes MA, et al. The Physiological Roles of Extracellular Metallothionein. Exp Biol Med. 2006; 231(9): 1548-1554.
  • 25. Yin X, Knecht DA, Lynes MA. Metallothionein mediates leukocyte chemotaxis. BMC Immunology. 2005; 6(21):
  • 26. Bert SG, Vallee BL. The Metallothionein/Thionein System: An Oxidoreductive Metabolic Zinc Link. ChemBioChem. 2009; 10: 55-62.
  • 27. Klaassen CD, Liu J, Diwan BA. Metallothionein Protection of Cadmium Toxicity. Toxicol Appl Pharmacol. 2009; 238(3): 215–220.
  • 28. Lansdown ABG, et al. Zinc in wound healing: Theoretical, experimental, and clinical aspects. Wound Repair and Regeneration. 2006; 15(1): 2-16.
  • 29. Maret W. Metallothionein redox biology in the cytoprotective and cytotoxic functions of zinc. Experimental Gerontology. 200; 43(5): 363-369.
  • 30. Otsuka F. Molecular Mechanism of the Metallothionein Gene Expression Mediated by Metal-Responsive Transcription Factor 1. Journal of Health Science. 2001; 47(6): 513-519.
  • 31. Laity JH, Andrews GK. Understanding the mechanisms of zinc-sensing by metal-response element binding transcription factor-1 (MTF-1). Archives of Bochemistry and Biophysics. 2007; 463(2): 201-210.
  • 32. Lindert U, et al. Metal-responsive transcription factor 1 (MTF-1) activity is regulated by a non-conventional NLS and a metal-responsive transactivation domain. Mol. Cell. Biol. 2009 ; 29(23): 6283-93.
  • 33. Kang YJ. Metallothionein Redox Cycle and Function. Biology and Medicine. 2006; 231(9): 1459-1467.
  • 34. Krężel A, and Maret W. Different redox states of metallothionein/thionein in biological tissue. Biochem J. 2007; 402(3): 551–558.
  • 35. Maret W, and Krężel A. Cellular Zinc and Redox Buffering Capacity of Metallothionein/ Thionein in Health and Disease. Mol Med. 2007; 13(7-8): 371–375.
  • 36. Andrews GK. Regulation of metallothionein gene expression by oxidative stress and metal ions. Biochemical Pharmacology. 2000; 59(1): 95-104.
  • 37. Loh SN. The missing Zinc: p53 misfolding and cancer. Metallomics. 2010; 2: 442-449.
  • 38. Vousden KH, Lane DP. p53 in health and disease. Nature Reviews Molecular Cell Biology. 2007; 8: 275-283.
  • 39. Harms KL, Chen X. The C Terminus of p53 Family Proteins Is a Cell Fate Determinant. Mol. Cell. Biol. 2005; 25(5): 2014-2030.
  • 40. Ostrakhovitch EA, et al. Interaction of metallothionein with tumor suppressor p53 protein. FEBS Letters. 2006; 580: 1235–1238.
  • 41. Ostrakhovitch EA, et al. p53 Mediated Regulation of Metallothionein Transcription in Breast Cancer Cells. Journal of Cellular Biochemistry. 2007;102:1571–1583.
  • 42. Ito T, et al. Role of zinc finger structure in nuclear localization of transcription factor Sp1. Biochemical and Biophysical Research Communications. 2009; 380(1): 28-32.
  • 43. Song IS, et al. Transcription Factor Sp1 Plays an Important Role in the Regulation of Copper Homeostasis in Mammalian Cells. Molecular Pharmacology. 2008; 74(3): 705-713.
  • 44. Wiestra I. Sp1: Emerging roles—Beyond constitutive activation of TATA-less housekeeping genes. Biochemical and Biophysical Research Communications. 2008; 372(1): 1-13.
  • 45. Rana U, et al. Zinc Binding Ligands and Cellular Zinc Trafficking: Apo-metallothionein, Glutathione, TPEN, Proteomic Zinc, and Zn-Sp1. J Inorg Biochem. 2008; 102(3): 489-499.
  • 46. Bellingham SA, et al. Regulation of Prion Gene Expression by Transcription Factors SP1 and Metal Transcription Factor-1. The Journal of Biological Chemistry. 2009; 284: 1291-1301.
  • 47. Girnun GD, et al. Synergy between PPARγ Ligands and Platinum-Based Drugs in Cancer. Cancer Cell. 2007; 11(5): 395–406.
  • 48. Shen H, et al. Zinc Deficiency Induces Vascular Pro-Inflammatory Parameters Associated with NF-B and PPAR Signaling. Journal of the American College of Nutrition. 2008; 27(5):577-587.
  • 49. Huang M, Shaw CF, and Petering DH. Interprotein metal exchange between transcription factor IIIa and apo-metallothionein. Journal of Inorganic Biochemistry. 2004; 98(4): 639-648.
  • 50. Karin M. The Regulation of AP-1 Activity by Mitogen-activated Protein Kinases. The Journal of Biological Chemistry. 1995; 270:16483-16486.
  • 51. Hess J, Angel P, and Schorpp-Kistner M. AP-1 subunits: quarrel and harmony among siblings. Journal of Cell Science. 2004; 117: 5965-5973.
  • 52. Xie S, et al. Transcription factor decoys for activator protein-1 (AP-1) inhibit oxidative stress-induced proliferation and matrix metalloproteinases in rat cardiac fibroblasts. Translational research. 2009; 153(1):17-23.
  • 53. Beiqing L, Chen M, Whisler RL. Sublethal levels of oxidative stress stimulate transcriptional activation of c-jun and suppress IL-2 promoter activation in Jurkat T cells. The Journal of Immunology. 1996; 157(1): 160-169.
  • 54. Lin Q, et al. Heme Oxygenase-1 Protein Localizes to the Nucleus and Activates Transcription Factors Important in Oxidative Stress. The Journal of Biological Chemistry. 2007; 282:20621-20633.
  • 55. O’Neill LAJ and Kaltschmidt C. NF-kB: a crucial transcription factor for glial and neuronal cell function. Trends in Neurosciences. 1997; 20(6): 252-258.
  • 56. Jacobs MD, and Harrison SC. Structure of an IκBα/NF-κB Complex. Cell. 1998; 95(6): 749-758.
  • 57. Zampieri S, et al. Oxidative stress in NPC1 deficient cells: Protective effect of allopregnanolone. J Cell Mol Med. 2009; 13(9B): 3786–3796.
  • 58. Li Q, et al. Endosomal Nox2 Facilitates Redox-Dependent Induction of NF-kB by TNF-a. Antioxidants & Redox Signaling. 2009; 11(6): 1249-1263.
  • 59. Kim CH, et al. Zinc-induced NF-κB inhibition can be modulated by changes in the intracellular metallothionein level. Toxicology and Applied Pharmacology. 2003; 21q190(2): 189-196.
  • 60. Allen RG, and Tresini M. Oxidative Stress and Gene Regulation. Free Radical Biology & Medicine. 1999; 28(3):463-499.
  • 61. Nguyen T, Nioi P, Pickett CB. The Nrf2-Antioxidant Response Element Signaling Pathway and Its Activation by Oxidative Stress. The Journal of Biological Chemistry. 2009; 284:13291-13295.
  • 62. Fourquet S, et al. Activation of NRF2 by Nitrosative Agents and H2O2 Involves KEAP1 Disulfide Formation. The Journal of Biological Chemistry. 2010; 285:8463-8471.
  • 63. Barry SP, et al. STAT3 deletion sensitizes cells to oxidative stress. Biochem Biophys Res Commun. 2009; 385(3): 324–329.
  • 64. Carballo M, et al. Oxidative Stress Triggers STAT3 Tyrosine Phosphorylation and Nuclear Translocation in Human Lymphocytes. The Journal of Biological Chemistry. 1999; 274:17580-17586.
  • 65. Anderson GP, Nerurkar NL. Improved fluoroimmunoassays using the dye Alexa Fluor 647 with the RAPTOR, a fiber optic biosensor. Journal of Immunological Methods. 2002; 271: 17-24.
  • 66. Grubor NM, et al. Novel biosensor chip for simultaneous detection of DNA-carcinogen adducts with low-temperature fluorescence. Biosensors and Bioelectronics. 2004; 19: 547–556.
  • 67. Yanase Y, et al. Living cell positioning on the surface of gold film for SPR analysis. Biosensors and Bioelectronics. 2007; 23(4): 562-567.
  • 68. Crowthers KC, et al. Augmented Humoral Immune Function in Metallothionein-Null Mice. Toxicology and Applied Pharmacology. 2000; 166: 161–172.


QR Code
QR Code spr (generated for current page)