2.1.

Wide-Field IPM

Figure 1 shows an off-axis Linnik interferometer illuminated by a Ti:Sapph laser (Micra-5, Coherent), tuned to a central wavelength of 780 nm, and a full width half maximum spectral bandwidth of 40 nm, defining a coherence length of 6.7 μm.¹⁷ This coherence length allows elimination of erroneous interferences originated from back-reflections from optical elements in the system or from unfocused layers inside the sample, yielding low noise levels.

As shown in Fig. 1, the illumination beam from the Ti:Sapph laser is expanded using lenses L1 and L2 in order to match the beam spot size upon the sample plane to the sensor size. After lens L3, positioned in a 4-f configuration with respect to the microscope objectives, the beam is split into reference and sample beams by beam splitter (BS). On the sample arm, a mirror directs the beam into microscope objective ${\mathrm{MO}}_{\text{SAMPLE}}$ ($\mathrm{M}\text{-}40\times$, Newport) and the sample is illuminated. Light back-reflected from the sample is magnified by ${\mathrm{MO}}_{\text{SAMPLE}}$ and projected through BS and lens L4 onto the CMOS digital camera (DCC1545M, Thorlabs, Newton, New Jersey). The reference beam, from the other output of the BS, goes through a compensating microscope objective ${\mathrm{MO}}_{\mathrm{REF}}$, and is back-reflected by a slightly tilted mirror. Then, the reference beam is projected onto the camera with a small angle with respect to the sample beam, creating an off-axis interference pattern on the digital camera.

2.2.

PT Excitation

Small diameter (10 to 90 nm) spherical AuNPs have peak or near-peak absorption coefficient at a green wavelength of 532 nm.⁶ To excite the AuNPs at the peak absorption wavelength, we used a diode pump solid state (DPSS) laser (LRS-0532-PFM-00200-03, Laserglow, Toronto, Canada) connected to a signal generator (DSOX 2002A, Agilent, Santa Clara, California), which can modulate the beam at wide ranges of frequencies (we have used 30 Hz for cellular imaging). The green excitation beam is incorporated into the interferometric sample arm by dichroic mirror (DM) (FF593-Di03, Semrock, Rochester, New York) and passes through ${\mathrm{MO}}_{\text{SAMPLE}}$. A long-pass filter (LPF) (FF01-593/LP, Semrock) prevents green light back-reflection residues from reaching the camera. For some experiments, we have added a green beam narrower/expander and/or a focusing lens, in order to adjust the green spot size upon the sample and consequently the power density ($\mathrm{W}/{\mathrm{cm}}^2$) to the excitation levels needed.

3.1.

Phase Acquisition

Assuming that the off-axis interference pattern on the camera contains straight fringes with respect to the $x$-direction, the off-axis interferogram of the sample can be expressed as follows:^18,19

The recorded interferogram is digitally two-dimensional (2-D) Fourier-transformed using MATLAB to create the image spectrum which consisted of a zero-order term and two high-order cross-correlation terms, spatially separated from the zero-order term.¹⁹ Then one of the cross-correlation terms is spatially filtered, centered, and padded with zeros. After an inverse Fourier-transform, we obtain a complex matrix, for which the argument represents the wrapped phase of the sample, $\varphi (x,y)$. This phase profile is then digitally unwrapped to remove $2\pi$ ambiguities.²⁰

3.2.

PT Signal Acquisition

We first used spherical AuNPs with diameters of 70 nm (A11-70, Nanopartz, Loveland, Colorado), which strongly absorb light at a wavelength of 532 nm.^6,21 This local light absorption produces a localized temperature rise and subsequent local variations of the refractive index, whereas the phase signal recorded by the camera explicitly depends on the refractive index. We modulated the excitation beam to illuminate the sample with an excitation frequency of $f_{\mathrm{ex}}=30\mathrm{Hz}$ and, as can be seen in Fig. 2(a), obtained local phase oscillations on the spatial point representing the nanoparticle position. Recording the image sequence at a frame rate higher than twice $f_{\mathrm{ex}}$ enabled us to analyze each pixel in the FOV, and de-modulate the temporal Fourier signal shown in Fig. 2(b) by filtering $f_{\mathrm{ex}}$. We define the PT phase signal as the peak-to-peak magnitude of the local PT-induced phase oscillations $\mathrm{\Delta }{\varphi }_{\mathrm{p}-\mathrm{p}}$, as obtained by an inverse Fourier-transform of the spectrum filtered around the excitation frequency.

Fig. 2

PT phase signal: (a) Phase oscillations in the AuNPs location caused by the PT effect, originated by a 30-Hz excitation of AuNPs at their plasmonic resonance peak wavelength. (b) The corresponding Fourier signal, showing a peak at the modulation frequency.

Analytical analysis for the absorption and heat conductivity was developed elsewhere.⁷ We present the spatial profile of the peak to peak phase shift at a distance ${\overrightarrow{r}}_0$ from the center of the AuNPs as follows:⁹

4.1.

PT Imaging of Aqueous AuNPs Solution and AuNPs on a Coverslip

To demonstrate wide-field PT imaging, we prepared a 250-μm thick chamber filled with an aqueous solution of the 70-nm AuNPs mentioned above (70 ppm diluted to 25%). The sample was excited at various illumination powers, modulated at 50 Hz in time. As described earlier, the PT phase signal for each pixel was acquired by analysis of the time-resolved change of the sample phase, $\mathrm{\Delta }{\varphi }_{\mathrm{p}-\mathrm{p}}$. For this experiment, we define the mean PT signal as the average signal over a small region of $5\times 5{\mu \mathrm{m}}^2$, overlapping with the excitation beam. Equation (2) determines the PT signal dependency on the excitation beam power, among other factors. By keeping all other parameters constant, the signal should rise linearly with an increase in the excitation beam power, which gives us an indication that we do measure a PT signal originated from the AuNPs. Figure 3 demonstrates this linear relation in our case. Indeed, since the thermal parameters and the concentration (estimated by $N=10$, the number of AuNPs contributing to the temperature rise in the excited volume) were invariant, the sole contributor to the rise of $\mathrm{\Delta }{\varphi }_{\mathrm{p}-\mathrm{p}}$ was $I({\overrightarrow{r}}_0)$. This suggests that the measured PT signal was linearly dependent on excitation power, which supports the theoretical analysis.

Fig. 3

Spatially averaged PT phase signal as a function of the excitation power for an aqueous solution of 70 nm spherical AuNPs, as measured by wide-field IPM. Dashed red line is a linear fit, $R^2=0.993$. Green line represents the theoretical response.

Next, we used wide-field IPM to image a coverslip covered with AuNPs. To prepare the sample, a drop of aqueous solution of AuNPs was carefully smeared upon a glass coverslip. Figures 4(a) and 4(b) show the phase signature of a $10\times 10{\mu \mathrm{m}}^2$ FOV, containing three AuNP conjugates. This demonstrates the system ability to obtain wide-field interferometric PT signals and image several AuNP conjugates simultaneously, without the need for lateral scanning. Possible explanations for the nonuniform PT signal strength of the different aggregates are the relatively small beam diameter, which caused excitation by different power densities across the FOV, as well as the different dimensions of the aggregates and the number of nanoparticles creating each of them.²² AuNPs aggregation changes the absorption coefficient of the nanospheres,²³ or $\sigma$, as defined in Eq. (2). This degrades our ability to conduct a thorough investigation of the PT signal as a function of the aggregate size. For this reason, we have also created a test target containing groups of isolated AuNPs using lithography, as presented in the next section.

Fig. 4

Wide-field PT phase signal of the AuNPs on a glass coverslip. (a) PT phase signal image. The white arrow indicates on the AuNPs located in the middle of the excitation beam. Video 1 shows the filtered phase oscillations in time (Video 1, MPEG, 5.3 MB) [URL: http://dx.doi.org/10.1117/1.JBO.18.11.111414.1. (b) Three-dimensional view of the same field of view shown in (a). (c) PT phase signal cross-section along the dashed white line shown in (a). Red-dashed graph: Gaussian fitted curve. (d) Blue dots: measured PT phase signal maxima as a function of the excitation power. Red-dashed line: linear fit. Green-solid line: theoretical graph.

For the case of AuNPs on a coverslip, in order to examine the PT phase signal response as a function of the excitation power, the excitation beam diameter was reduced to 5 μm and its power was gradually changed from 5 to 40 mW on the sample plane by the DPSS laser power control. The solid black line in Fig. 4(c) shows the PT phase signal cross section along the path indicated by the dotted white line in Fig. 4(a). The red-dashed line in Fig. 4(c) is a Gaussian fitted curve, with a good correlation to the measured PT phase signal ($R^2$ of 0.989). This result is in agreement with previous works, as well as with theoretical analysis of plasmon resonance.^9,24

The PT signal as a function of laser power is shown in Fig. 4(d) at the peak of the AuNPs aggregate marked by a white arrow in Fig. 4(a). Here too, the signal is linearly dependent on the laser power, as suggested by Eq. (2). Moreover, the actual measured PT signal is in good agreement with the theoretical values [solid line in Fig. 4(d)].

To assess the quality of the measurements, we define two types of signal-to-noise ratios (SNRs). Temporal SNR is defined as the PT phase signal at the center of the AuNPs when the PT excitation is active, divided by the phase value at the same spatial point, but without excitation. Spatial SNR is defined as the PT phase signal at the center of the AuNPs when the PT excitation is active, divided by the average phase value when the PT excitation is active but at an area where there are no AuNPs.

In our case, the peak level of the PT phase signal is 0.420 rad [see green-dashed line in Fig. 4(c)], whereas the average background signal, which defines the spatial noise level for this measurement, is 0.023 [see dotted black line in Fig. 4(c)], producing a spatial SNR of 18.1.

4.2.

PT Imaging of Isolated AuNPs Test Target

To test the PT wide-field IPM system sensitivity with an aggregate-free test target, we have used electron-beam lithography to create a target containing groups of different numbers of isolated AuNPs, 40- to 100-nm wide and 40-nm high. In this assessment, we investigated isolated AuNPs groups starting from 1 AuNP, through $2\times 2$, $2\times 3$, $3\times 3$, $3\times 4$, $4\times 4$, $4\times 5$, $5\times 5$, $6\times 6$, $7\times 7$, and up to $8\times 8$ AuNPs. Figure 5(a) presents a part of the AuNPs groups in this test target as imaged with environmental scanning electron microscope (eSEM), and Fig. 5(b) presents a magnified image of one of the groups ($5\times 5$ AuNPs). As verified by eSEM, due to the lithography process limitation on glass substrate, a small part of the AuNPs was lithographed smaller than others. With the elimination of AuNPs aggregates, this approach enabled us to make a more accurate assessment of our detection threshold compliance to the theoretical model.

Fig. 5

Imaging of a test target created by electron-beam lithography, containing groups of isolated AuNPs, with a different number of AuNPs in each group. (a) An eSEM image of several of the AuNPs groups from the test target. Image contrast is inverted for a better presentation. (b) Enlarged eSEM image of the AuNPs group marked by a circle in (a), containing $5\times 5$ isolated AuNPs. (c) Black line: PT phase signal as a function of the AuNPs number in the group. Blue-dashed line: the effective the effective detection threshold. Orange-dotted line: the measured detection threshold. Green triangles: the averaged temporal SNR of the PT phase signal as a function of AuNPs number.

To analyze the results, we define two types of thresholds. The measured threshold is defined as the PT phase signal $\mathrm{\Delta }{\varphi }_{\mathrm{p}-\mathrm{p}}$ (as defined in Sec. 3.2) in an area where there are no AuNPs and when the PT excitation is active. When using PT excitation with power density of $1\mathrm{MW}/{\mathrm{cm}}^2$ on the lithographed test target, the measured threshold is 9 mrad, as marked by the dotted orange line in Fig. 5(c). As shown in this figure, according to this threshold definition we have experimentally confirmed that groups of down to 4 AuNPs can be detected. However, since only groups containing 16 AuNPs and more produced a linear relation between the AuNPs number and the PT signal strength, as predicted by Eq. (2), we define another threshold, the effective detection threshold. This threshold is defined as the phase level above which the signal is linearly dependent on the AuNPs number according to Eq. (2). As shown in Fig. 5(c), the effective detection threshold in our case coincides with temporal SNR of 5, and all the groups containing $>16$ AuNPs are above this threshold.

5.1.

AuNP Synthesis and Biofunctionalization

Colloidal AuNPs were prepared by citrate reduction of ${\mathrm{HAuCl}}_4$ in solution.²⁷ About 1.0 mL of 12.7 mM chloroauric acid solution was added to 49.0 mL of deionized water (18 MΩ) in 100.0-mL flask equipped with reflux condenser. The resulting solution was brought to boil while stirring constantly, and then 0.94 mL of 38.8 mM trisodium citrate solution was added quickly. The boiling was continued for several more minutes after the solution developed a rich red color. The nanoparticles have an absorption maximum at 521 nm, hydrodynamic radius of 21.9 nm, as we measured by dynamic light scattering, and average size of $16.3\pm 0.9\mathrm{nm}$, as we measured by transmission electron microscopy. The suspensions were stored at 4°C in the dark in order to keep them from interacting with light.

Next, AuNPs were conjugated to the anti-EGFR monoclonal antibodies (host mouse).²⁶ AuNPs and anti-EGFR antibodies were both diluted with 20 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid buffer separately, and then mixed for 20 min. Poly(ethylene glycol) solution (1%) was then added to the AuNPs dispersion and centrifuged at 6000 rpm for 30 min. After centrifuging, the anti-EGFR/AuNPs pellet was dispersed in phosphate-buffered saline (PBS) buffer and used immediately for immunolabeling, as described in the next section.

5.2.

Cell Sample Preparation

We used MDA-MB 468 cell lines, which are known as EGFR-overexpressing ($\mathrm{EGFR}+$) cells. RPMI-1640 (ATCC, Catalog No. 30-2001) medium was supplemented with fetal bovine serum (10%), L-glutamine (2 mM), and penicillin/streptomycin (0.1%). Cells were grown in an incubator keeping an atmosphere with 95% relative humidity, 5% ${\mathrm{CO}}_2$ and temperature of 37°C. Cells to be immunolabeled and imaged were grown on glass coverslips in CoverWell™ imaging chambers to approximately 30% confluence.

Next, the solution of AuNPs conjugated to $\mathrm{EGFR}+$ was added to the cancer cell culture. The cell culture was incubated with the anti-EGFR/AuNPs solution for 20 min and then rinsed with PBS buffer several times in order to keep AuNPs conjugated upon the cell membrane and to avoid their uptake inside the cells.

Afterwards, to enable long measurements of the same cells using both IPM and confocal microscopy, the cells were fixated. To prepare the fixated cell sample, the labeled cells were first rinsed three times with PBS buffer, then fixated with 4% paraformaldehyde solution for 8 min, rinsed twice with PBS buffer, and finally mounted with Fluoromount medium. To ease finding the area of cells on two different imaging systems, a selected area was marked by a grid painted on the coverglass with a permanent pen.

5.3.

Experimental Setups

For PT IPM imaging of AuNP-conjugated cells, we used a $63\times$ oil-immersion objective (Plan-Apochromat with numerical aperture of 1.40, Zeiss). In practice, relatively weak excitation intensities were sufficient to produce significant PT signals. Therefore, the plane-wave excitation beam diameter was set to be $\sim 30\mu \mathrm{m}$ and its power was 10 mW, producing a power density of approximately $350\mathrm{W}/{\mathrm{cm}}^2$ everywhere upon the imaged cell and its surroundings. Similar power densities were used on the previous PT imaging methods.^9,10 Theoretically, in order to sample a 30-Hz signal, we should capture at least $60\text{frames}/\mathrm{s}$. We took 1000 full-field frames at a frame rate of $150\text{frames}/\mathrm{s}$, which improved signal clarity.

Next, we used a confocal microscope (Leica TCS SP5 II) with a $63\times$ oil-immersion objective to image six layers in the same cell by lateral and axial scanning with movements of $\mathrm{\Delta }z=0.5\mu \mathrm{m}$. Acquisition time for the entire FOV and all six sections was 50 s.

Figure 6 compares the different operation modes of PT wide-field IPM and confocal microscopy, illustrating that the PT IPM signal is additive over the entire sample thickness, whereas confocal microscopy can produce separated slices, but might miss a part of the AuNPs due to great dependency of its axial resolution on the scanning parameters. This will be further discussed in the next section.

Fig. 6

Schematic comparison between the different operation modes of: (a) PT wide-field IPM and (b) confocal microscopy. MO, microscope objective. On the bottom, schematic illustration of possible AuNP images obtained from the setups. Discrete $z$-scanning layers are denoted by $z1-z4$. Each “$z$” image illustrates an $x-y$ scanning of the beam focus across the layer. Blue and red arrows shown in (a) indicate on particles missed by confocal microscopy due to discrete $z$-scanning, but are imaged by PT wide-field IPM, which requires no scanning.

5.4.

Experimental Results and Discussion

Several cells were imaged with both techniques (PT wide-field IPM and confocal microscopy) for comparison. All measurements showed a good correlation between signal strength and location relative to the cell topography. Figure 7(a) shows several confocal slices of AuNPs attached to EGFRs on a cancer cell, by which we validated that the AuNPs are mainly conjugated to the membrane itself, so that AuNPs inner-cellular uptake is negligible. The cumulative confocal image obtained by summing these slices is shown in Fig. 7(b) in color, overlaying the DIC image of the cell shown in gray scale. Figure 7(c) shows the PT phase image of the AuNPs in the same cell shown in color, overlaying the regular quantitative phase map of the cell shown in gray scale. As shown in these figures, PT wide-field IPM is able to simultaneously yield the regular quantitative phase profile, which is proportional to the optical thickness of the cell per each of the cell spatial points in addition to the AuNPs PT phase signal. On the other hand, the confocal microscope we used was able to yield the qualitative DIC image of the cell in addition to the AuNPs reflection signal. Note that while the confocal reflection-based signal shown in Fig. 7(b) has good lateral resolution, the PT signal from the IPM system shown in Fig. 7(c) seems more spread-out.

Fig. 7

Imaging of MDA-MB 468 ($\mathrm{EGFR}+$) cancer cell with conjugated AuNPs: (a) Confocal microscopy: six consecutive confocal images of the AuNPs in the cell, from top-left image, representing the bottom of the cell, to bottom-right image, representing the top of the cell, in axial steps of 0.5 μm apart. (b) Confocal plus DIC microscopy: cumulative confocal image of the AuNPs (colored) as obtained by summing the six consecutive confocal images shown in (a), overlaying a DIC image of the same cell (gray scale). (c) PT IPM plus regular IPM: PT phase image of the AuNPs (colored), overlaying a regular wide-field IPM phase image of the same cell (gray scale). (c) Comparison of the phase fluctuations at the excitation frequency with active excitation (green, dashed), representing the PT phase signal, which is almost 2 orders of magnitude larger than the sample ambient fluctuations at the excitation frequency, but when no excitation applied (blue, solid). Graphs shown are along the cell cross-section marked by a white-dashed line in (c). Vertical-dashed lines in (d) represent the cell borders.

Comparison of the relative normalized amplitudes of the two AuNPs-originated signals (PT wide-field IPM versus confocal) yielded 75.6% agreement. These results are explained either by the accumulative phase-signal nature and lack of good sectioning capabilities of PT IPM imaging, and also by the PT signal origin.²⁴ The PT signal is originated in heat transfer from the AuNPs to the environment and as such is more likely to spread unless thermally confined by relatively higher excitation frequencies⁷ which were far beyond our CMOS camera capabilities, having an acquisition rate of $150\text{frames}/\mathrm{s}$. On the other hand, the confocal signal is originated in a point-like back-reflection source.²⁸ This gives the back-reflection approach a lateral resolution advantage.

On the other hand, as shown in Fig. 7(c), IPM generates phase signals which are cumulative along the $z$-axis of the sample. The 2-D signal is therefore comprised of projections of AuNPs PT contributions from the entire three-dimensional volume of the cell at each spatial point. PT IPM has the advantages of acquisition rates and flexibility of objective selection, as it would be able to image thicker transparent samples.

We next examined the phase fluctuations at the excitation frequency originated from the sample itself when the excitation is not active. This indicates on our ability to discriminate between the PT phase signal and the ambient cell phase fluctuations at the excitation frequency. Figure 7(d) shows a quantitative comparison between the phase fluctuation levels at the excitation frequency, with excitation (dashed green) and without excitation (blue). As shown in this figure, the fixated cell ambient phase fluctuations at the excitation frequency do not exceed 5 mrad, which is almost 2 orders of magnitude lower than the average of the PT phase signal, 0.22 rad. Therefore, these two signals are easily discriminated. In general, if there are cell biological phenomena which induce significant phase fluctuations at a frequency overlapping with the excitation frequency, the AuNPs excitation frequency can be adjusted to avoid overlap with these phenomena.

Note that the discussion above was concerning the fluctuations of the cell at the excitation frequency. In any case, the system is capable of producing PT phase signals that are approximately independent of the actual phase of the sample. This is possible due to the phase signal response to $\partial n/\partial T$ according to Eq. (2), which confines the dependence of the signal on the sample thermal properties, applicable at the excitation frequency only and not on its physical static phase defining the optical thickness of the cell.