I have been fortunate to work with excellent collaborators through Jim's Nanosystems Biology Cancer Center (NSBCC) between Caltech, UCLA, and the Institute for Systems Biology (ISB). We further demonstrate that the sensitivity and dynamic range of the sensor can be modulated by using different surface chemistry. This principle provides a quantitative prediction of the role of disorder and enables the characterization of the protein-protein interaction network.

Therefore, we also propose the use of microbubbles as a model system to investigate the physical transformations of the PS system when exposed to environmental challenges. 61 Figure 2.15 High-resolution XPS spectra of surfaces 2 and 3 illustrating the coupling of an unnatural azide-containing amino acid and then a FLAG peptide.

M ICRO - AND NANOTECHNOLOGIES IN BIOLOGY
H OW TO STUDY BIOLOGY : T OP - DOWN VS . BOTTOM - UP
C OMPLEXITY OF BIOLOGY AND MULTI - PARAMETER ANALYSIS
T HESIS OVERVIEW
R EFERENCES

We are already seeing the success of single-cell-based bottom-up approaches in biology and medicine.20–23 Lahav et al. Multi-parameter analysis will be one of the main topics discussed during this thesis. One of the challenges we noticed from the approach presented in Chapter 4 is how to analyze the large amount of information.

We also present studies of the heterogeneous ozonolysis of a mixture of saturated and unsaturated phospholipids at the air-liquid interface. A mixture of the saturated phospholipid 1,2-dipalmitoyl-sn-phosphatidylglycerol (DPPG) and unsaturated POPG is examined in negative ion mode using FIDI-MS, while a mixture of 1,2-dipalmitoyl-sn-phosphatidylcholine (DPPC) and 1-stearoyl-2-oleoyl-sn-phosphatidylcholine (SOPC).

I NTRODUCTION

E XPERIMENTAL M ETHODS

Nanowire sensor fabrication
Surface functionalization and characterization for DNA sensing and antibody-
Surface functionalization and characterization for peptide-based protein sensing
SPR and electronic measurements

R ESULTS AND D ISCUSSION

DNA sensing
Protein sensing with antibodies
Protein sensing with peptide

C ONCLUSIONS

R EFERENCES

I NTRODUCTION

Microfluidic chip fabrication for DNA patterning
Patterning of DNA barcode arrays
Microfluidic chip fabrication for multi-protein detection
Cell culture
Multi-protein detection
On-chip cell lysis and mulplexed intracellular protein profiling from single cells
Data analysis
Molecular dynamic simulation
Modeling of electrostatic adsorption of DNA to poly-L-lysine (PLL) surface

C ONCLUSIONS

R EFERENCES

I NTRODUCTION

Patterning of DNA barcode arrays
Microfluidic chip fabrication for the detection of protein secretion
Cell culture
On-chip secretion profiling
Data analysis

Multiple-protein secretion profiling from single cells and small cell colonies
Secretion profile from integrated barcode chip vs. bulk experiment
Single-cell protein secretion profiling of GBM cell line: U87 cells
Cell–cell communication effect
PTEN activity on GBM cell line: U87EGFRvIII vs. U87EGFRvIII/PTEN
Toward clinical sample: analysis on GBM primary tumor cells

C ONCLUSIONS

R EFERENCES

I NTRODUCTION

Protein signaling pathways play important roles in tissue processes ranging from tumorigenesis to wound healing.1-5 Elucidation of these signaling pathways is challenging, in large part, due to the heterogeneous nature of tissues.6 A such heterogeneity makes it difficult to separate cell-autonomous changes in function from changes caused by paracrine signaling and may mask the cellular origin of paracrine signaling. Intracellular signaling pathways can be resolved through measurements of multiple proteins at the single-cell level.7 For secreted protein signaling, there are additional experimental challenges. Intracellular staining flow cytometry (ICS-FC) requires the use of protein transport inhibitors, which can affect measurements.8 Furthermore, the largest number of cytokines analyzed simultaneously in single cells by ICS-FC is only 5.9 Finally, some biological concerns, such as as the influence of one cell on another, are difficult to decipher using ICS-FC.

Here, we describe an experimental/theoretical approach aimed at unraveling the coordinated relationships between secreted proteins and understanding how molecular and cellular perturbations can affect these relationships. We characterize the secretome at the single-cell level using a microarray platform in which single stimulated macrophage cells are isolated into microchambers of 3 nanoliter (nl) volume, with ~1000 microchambers per chip. Each microchamber allows duplicate assays for each of the dozen proteins secreted during the several-hour incubation period following cell stimulation.

We demonstrate that the observed spread in protein levels is dominated by cellular behavior (biological fluctuations), rather than experimental error. This matrix is analyzed at both coarse and fine levels to extract protein-protein interactions. We demonstrate that our system has the stability properties required for the application of a quantitative version of a Le Chatelier-like principle, which allows a description of the system's response to a perturbation.

The fluctuations, as assessed from the multiple protein assays of undisturbed single cells, are used to predict the results when the cells are perturbed by the presence of other cells, or by molecular (antibody) perturbations.

M ETHODS

Experimental Methods
Theoretical Methods

One contribution to the experimental error arises from the variability of the flow pattern antibody barcodes. For a given table, each row represents the copy numbers of the twelve proteins for a single cell or small cell colony. Even for one cell, there may be deviations from the bell-shaped theoretical functional form in the higher tail of the histogram due to autocrine signaling.

However, we have numerical indications that the unperturbed state of the single cell may be unstable in the presence of many other cells. These are analogs of the chemical potentials as introduced in the thermodynamics of systems of more than one component. To theoretically characterize the response of the cellular secretion to a perturbation, we first calculate the change in the distribution for the special case in which a perturbation changes the potential of protein i from , where there is a small increase.

We show (equation S5.2 in appendix B, section 5.7.3) that the distribution changes to first order in the change of the potential with One is that a perturbation will distort the shape of the copy number distribution of a given protein. It is thus the high end of the distribution that is most strongly affected by the disturbance (see, for example, Fig. 5.9).

The other immediate implication of the change in distribution is that the mean values will change. Therefore, the smaller the fluctuations (ie, the narrower the histogram), the more resilient the distribution is to changes. Therefore there is some disturbance through autocrine signaling, as seen in the bump in the higher tail of the histogram.

Figure 5.3 Cross-reactivity check and calibration curves. (A) Scanned image showing cross- cross-reactivity check for all 12 proteins

R ESULTS AND D ISCUSSION

Computing the covariance matrix
The network
The composite networks
The number-based network

The conversion from fluorescence intensity to number of molecules does not change the coefficient of variation when we are in the linear regime of the calibration curve (see Fig. 5.3). It is these interdependencies, as revealed by the columns of the covariance matrix, that provide the prediction of network connectivity (part B). Entries are the covariance of the indicated protein with other proteins listed on the abscissa.

This interval is chosen to dampen the high reading of the autocorrelations in the covariance matrix. The large and positive magnitude of the covariance of MIF and IL-8 is shown as a double-headed arrow. In the diagram, inhibition is indicated as usual with a bar at the end of the connector.

In the second stage in our analysis of the covariance matrix we aim to show a more resolved structure and in this way to notice features that are revealed in the global network of Fig. matrix on the number of cells in the sample. Similar to the single-cell case (Fig. 5.11), the level inputs are scaled by the magnitude of the eigenvalues.

The range is fixed in order to mitigate the effect of eigenterms in the covariance matrix. We quantify the magnitude of that reduction by looking to reproduce the copy number reduction of the perturbed protein directly. The quality of prediction in IL-8 and VEGF perturbation experiments is excellent, as shown in Fig.

Figure 5.13 Heat map of the covariance matrix (left) and of the contributions to the first two tiers of the network (right) for measurements on chambers containing three cells

C ONCLUSIONS

R EFERENCES

H.; Ferrari, G.; Janetzki, S., Measurement of cytokine release at the single cell level using the ELISPOT assay. Wang, J.; Ahmad, H.; Ma, C.; Shi, Q.; Vermes, O.; Vermeš, U.; Heath, J., An automated one-step chip for rapid, quantitative and multiplexed detection of proteins from whole blood pricks.

A PPENDIX A: S UPPLEMENTARY E XPERIMENTAL M ETHODS (SI.I)

Experimental procedure
Experimental data analysis methods

Cells were first treated with 100 ng/mL phorbol 12-myristate 13-acetate (PMA) for 12 hours during which a characteristic morphological change was observed as an indication of macrophage induction (Fig. 5.15). Since the signal range is highly dependent on antibody activities as well as cell biology, it is necessary to decide whether the signal is real and reliable. The ratio of mean values over all single-cell experiments (IL-2-specific protein) gives an S/N value.

One of the most important characteristics of SCBC analysis is the heterogeneous cellular behavior at the single-cell level. The experimental variability of the SCBC platform, which reflects the system error as well as the biological variability due to the cellular heterogeneity, contributes to the fluctuation of the overall signal. The former can be estimated by the histogram of the fluorescence intensity from the calibration experiment with recombinant proteins.

As a result, the fluorescence intensity distribution of a specific recombinant reflects the detection profile of the DNA barcode. The protein signal is thus dependent on diffusion and therefore cell location may be a source of the variation. The detection variation of the MIF protein due to the DNA uniformity obtained from the histogram of the calibration data set was included in the analysis.

To account for the worst case, we used a barcode variability of 10% for the remainder of the analysis. The CVs from this simulation represent the distribution of our measurements for single-cell chambers without accounting for cellular heterogeneity, i.e. system errors. We can see that the biological variation dominates the total error of the test.

Figure 5.15 Morphology change of THP-1 cells upon PMA/LPS activation for 24 h. (A) monocytic THP-1 cells without induction, (B) macrophage-like THP-1 cells after PMA/LPS treatment

A PPENDIX B: S UPPLEMENTARY T HEORY M ETHODS (SI.II)

Introduction to theoretical supplementary methods
The ensemble: A basis for making predictions
Fluctuations describe the response to small perturbations
The principle of Le Chatelier
The equation for the direction of change
Tiers of the network are eigenvectors of the correlation matrix
The spectral representation of the covariance matrix
The role of the number of cells in the sample
Antibody perturbations

The system we consider is many independent replicas of a compartment containing a single cell in a nutrient solution in thermal equilibrium. Since the system is not large, its different replicas may differ in the number, Ni, of type i proteins secreted. The solution is well known because if we measure many compartments, the required distribution is the one whose entropy is the largest.

It is possible to show 37 that this approach does not require the system to be macroscopic in size. It is sufficient if we measure enough replicates so that the distribution of proteins does not change significantly when we add more measurements. If each replica is small, we can observe the fluctuations, which is the experiment described in the main text.

The main point is that even if the fluctuations are not small, it is possible to make predictions. We discuss three types of predictions in the paper, with more details given in this section. We emphasize that the prediction is made strictly independent of the experiment with which it is compared.

The numerical value of these multipliers is determined at the final stage by setting the condition that the distribution (Eq. S5.1) reproduces the given values of the averages. If our system were macroscopic in size, we would call µi 'the chemical potential of protein i'. Ξ is a function of all the Lagrange multipliers and its role is to ensure that the sum of the probability over all possible combinations yields one.

A PPENDIX C: S UPPLEMENTARY T ABLES

I NTRODUCTION