Photoemission electron microscopy (PEEM) is a unique surface-sensitive instrument capable of providing real-time images with high spatial resolution. While similar to the more common electron microscopies, scanning electron microscopy and transmission electron microscopy, the imaging technology relies on the photogeneration of electrons emitted from a sample through light excitation. This imaging technique has found prominence in surface and materials sciences, being well suited for imaging flat surfaces, and changes that occur to that surface as various parameters are changed (e.g. temperature, exposure to reactive gases). Biologically focused PEEM received significant attention in the 1970s, but was not aggressively advanced since that pioneering work. PEEM is capable of providing important insights into biological systems that extend beyond simple imaging. In this article, we identify and establish important issues that affect the acquisition and analysis of biological samples with PEEM. We will briefly review the biological impact and importance of PEEM with respect to our work. The article also concludes with a discussion of some of the current challenges that must be addressed to enable PEEM to achieve its maximum potential with biological samples.
Photoemission electron microscopy (PEEM) is a unique, surface-sensitive technique capable of providing real-time images of the surface of a sample under high spatial resolution. The physical principle underlying this technique is the photoelectric effect. Simply, electrons are emitted from the surface of the sample when the associated photon energy of the incident light is above the photoionization threshold value characteristic of the sample of interest. The generated photoelectrons are accelerated through a series of electron optics and the surface of the sample is imaged at a high magnification. The first PEEM images were published in 1933 (1). Following these preliminary images tremendous development in the understanding and instrumentation of PEEM has been achieved (2,3). PEEM and related imaging techniques are used extensively in the field of materials science, and there is a dedicated biannual meeting and associated conference proceedings that address scientific and instrumental advances in the field. Information about the latest meeting can be found at http://www.leem-user.com/.
From a historical perspective, O. Hayes Griffith (4) was one of the pioneers of PEEM technology and applied the technique to biological samples in 1972 (4), reporting first PEEM images of rat epididymis. Griffith and his group realized the capabilities of PEEM as the electron-optics analog of fluorescence microscopy and used colloidal gold and silver enhanced colloidal gold particles to selectively label portions of a biologic with a higher spatial resolution (5–7). PEEM images for cells (8–10), viruses (11,12) and DNA (11,13–15), among others (16–18), were reported using both labeling and nonlabeling techniques. In addition to the experimental evidence, reviews were written establishing PEEM as a powerful tool for biological imaging (19–21).
Examples of the sensitivity, topographic contrast and spatial resolution achieved by Griffith’s group using biologically focused PEEM are reproduced in Fig. 1. The sensitivity of the instrument is shown in Fig. 1a with an image of a mixture of T-4 bacteriophages and tobacco mosaic viruses (TMV) (12). In this image, the head of the bacteriophages is significantly brighter than TMV. A single helical strand of RNA characterizes TMV while highly condensed DNA characterizes the head of a bacteriophage. PEEM is sensitive to the density of the nucleic acid because it has been suggested that there are not large differences in photoemission between DNA and RNA and the photoemission of nucleic acids is greater than the protein components of the virus (12). The topographic contrast is illustrated in Fig. 1b with a PEEM image of MCF-7 human breast carcinoma cells (8). Visible among the textured surface of these cells are the protrusions of the large nuclei and nucleoli with the sides of these cellular protrusions appearing darker than the top portions. An example of the spatial resolution achieved by Griffith’s group is displayed by his PEEM image of recA-DNA (11) reproduced in Fig. 1c. These three images, along with the multitude of others that Griffith and coworkers have produced, exemplify the strides his group accomplished in pioneering biologically focused PEEM.
Despite the capabilities demonstrated by Griffith and others, the technique did not flourish in the biological sciences and applications of PEEM to biological systems have lagged behind complementary electron imaging techniques, such as scanning electron microscopy (SEM) and transmission electron microscopy (TEM). Instead, as an ultra high vacuum (UHV) instrument suited for relatively flat samples, PEEM found prominence in the fields of surface and materials sciences (2,22). The applications of PEEM in surface and materials sciences go beyond imaging. For instance, surface dynamics can be imaged during in situ experiments (e.g. heating, cooling, exposure to reactive gases, etc.) by monitoring the changes in the electronic properties of the surface (22–28). If the resulting work function of the modified surface is below (above) the photon energy of the incident light an image will (will not) be observed. Subsequently, and more importantly for the purposes of this review, tunable monochromatic light sources make it possible to scan a series of photon energies and obtain the work function or selectively turn on or off a particular region of the sample (29–31). Thus, PEEM has become a critical tool in surface and materials sciences that is capable of achieving valuable and unique information.
More recently, the determination of electrochemical properties (work functions) through PEEM has emerged as a significant application for biological systems (32–38). This particular technique is widely applicable to those systems whose electrochemical properties are central to understanding their function, but are impossible to determine using conventional approaches. The electrochemical properties are obtained by fitting the threshold data with functional fits based on theory rigorously defined for photoelectric curves of semiconductor and metallic surfaces. A series of concerns arise when applying techniques (analysis and instrumentation) that are rigorously established and developed for metallurgic, surface and materials sciences to biological sciences. The aim of this review was to identify and establish the key issues that affect the acquisition and analysis of photoionization thresholds for biological samples and briefly review the biological impact and importance of threshold-PEEM with respect to our work with human melanosomes. We will conclude with a discussion of the current issues of PEEM that need to be addressed in order to derive a deeper molecular understanding from the data obtained using this technique.
A schematic of the photoelectron emission microscope is shown in Fig. 2. The instrument consists of a main chamber, a lens column and a detection center, all of which are held under UHV. Emission of photoelectrons occurs when light of an appropriate wavelength is illuminated on the surface of a sample. A variety of photon sources are available to excite the electrons (X-ray, synchrotron radiation, or UV arc lamps and lasers). UV excitation is generally required to photoionize the valence electrons of metals and biological materials, and so this spectral region is the focus of this review. As a consequence of the instrumental design, shown in Fig. 2, the sample of the surface must be illuminated with a photon beam at a high incident angle. This is necessary because the maximal resolution of the instrument is achieved only when there is a short working distance between the anode and the sample (∼4 mm).
Typically, the sample is held at a negative potential and an electric field is created between it and a grounded anode. This potential accelerates the generated photoelectrons through a series of electron optics. After passing through the lens column, the emitted electrons are amplified by a microchannel plate and subsequently imaged onto a phosphor screen. A CCD camera captures the data from the phosphor screen. The electron optics project the image plane of the sample onto the microchannel plate, thus enabling the direct collection of a spatially resolved image.
For a geometrically flat sample, the current lateral instrumental resolution is about 10 nm (39). Samples with significant topography introduce spherical aberrations as the emitted electrons are ejected in various angles. Similarly, according to Einstein’s equation,
samples of multiple chemical components with different characteristic photoionization energies, φ, introduce chromatic aberrations as the emitted electrons are ejected at different kinetic energies, KE, upon illumination with constant photon energy, hv. In addition to the sample-dependent accelerating field aberrations, there is an objective lens spherical aberration for UV illumination that also affects the resolution. Reports calculating the theoretical resolution for PEEM systems indicate that it could achieve 5 nm and propose that implementation of aberration-corrected lenses can improve the resolution (40,41). Currently, there are aberration correction techniques that have been developed and implemented in PEEM systems to enhance the resolution for UV (42) and X-ray (43–45) photon excitations.
In addition to discussing the lateral resolution, it is necessary to briefly address the depth of information of PEEM. The depth of information is the distance between the surface and a point at which information from the sample contributes to the image at a specified resolution (46). To generate photoelectrons, light that penetrates the sample is absorbed, induces photoionization and the photoionized electrons are transported to the surface where they then escape from the sample to the vacuum (8). The depth-dependent yield of photogenerated electrons is not known. Estimates for organic materials suggest that electrons are generated within the first 10 nm (46). However, it is reasonable to assert that the yield will decrease with depth from the surface and so while light may be absorbed deep in the sample, the volume relatively near the surface will comprise the dominant contribution to the signal (5).
Spatial dependence of the signal
As stated above, the incident light beam illuminates the sample at a high incident angle with respect to the surface normal. Consequently, if the sample possesses significant morphology, only a portion of the sample surface is illuminated, and accordingly imaged. The curved side of the specimen facing the incident beam will have a high photoelectron flux whereas the opposite side will be shadowed and dark. Given the geometrical considerations described above, spherical samples will have an asymmetric photoemission image. As shown in Fig. 3 the intensity contour plot of the PEEM image of a 150 nm gold nanosphere reveals the photoemission image to be ellipsoidal. The long axis is the axis that is not deformed by the shadowing that results from the geometrical constraints of the PEEM and the incident light source (47).
The above point is important because the three-dimensional topography of biological specimens will also result in photoelectron emission at various angles. Thus, “shapes” revealed in the PEEM image may reflect the interplay between the illumination angle and the surface geometry, not simply the shape of the biomolecule. If the incident photon energy used to generate the photoelectrons is near the sample’s photoionization threshold, then the emitted electrons will possess a low kinetic energy. The large accelerating potential enables these electrons to be collected with a minimum loss of lateral resolution—the electrons are emitted at various angles for a curved substrate and would then tend to spread laterally (3). However, the situation is certainly more complicated by a complex three-dimensional surface, which will present topographic contrast as the trajectories of the emitted electrons are influenced by the high electric field of the specimen (48). Spreading or bunching of the electrons can also occur from the topographical electric field variations. As a result, the image may have darker or brighter regions according to the topographic features of the sample. An aperture located in the lens column is used to prevent deflected electrons from reaching the detection area. Deflection of the electrons typically occurs from the sides of a protrusion whereas the electrons arising from surrounding flat regions and the top regions of protrusions pass through the electron-optics system undeflected. Thus, the top regions of protrusions and surrounding flat areas appear brighter whereas the sides of protrusions are darker.
Finally, a two-dimensional image detection (the CCD serves as a two-dimensional image plane, with each pixel having the same area) creates a nonlinear distance scale for the image of a three-dimensional object. Consider the projection image of a sphere (Fig. 4a). The region of a curved three-dimensional surface imaged onto an individual pixel increases with the angle from the surface normal. Therefore, a curved three-dimensional shape will affect the resolution achieved by the instrument. With increasing angle from the surface normal, the photoejected electrons will initially have increased velocity projections in the plane of the substrate, and therefore will fan out as they undergo different degrees of curvature to be accelerated toward the lens column. As a result, if the accelerating potential is reduced, the size of the image will increase. Figure 4b displays the length of the long axis of the image of a ∼150 nm spherical Sepia eumelanin granule (recall that spherical structures appear as ellipsoids) with respect to the accelerating voltage. A linear dependence is observed for the acceleration potential examined. Nepijko et al. (47) developed a theoretical analysis of such an experiment and provided functional relationships for determining the actual size of the three-dimensional spherical particle from such data.
All of the above-mentioned spatial considerations are important when imaging samples with significant structure, as is the case with biological specimens. The interpretation of the generated images may be more complicated than what one would expect with flat surfaces as in surface and materials sciences because the resulting image is dependent on the morphology and topography of the specimen. While a challenge to disentangle these features, it also presents the opportunity and potential power of PEEM to biological systems; the data will reflect the three-dimensional properties of the system being imaged.
Polarization dependence of the signal
As a consequence of the PEEM geometry the photon beam used to illuminate the sample must come in at a high incident angle (∼77° from the surface normal). This angle of incidence is critical because for many materials it is quite close to Brewster’s angle. Recall that Brewster’s angle is defined as the angle, measured from the surface normal, at which s-polarized light is reflected and p-polarized light is absorbed (49). (The direction of polarization is perpendicular and parallel to the plane of incidence for s- and p-polarized light, respectively.) Considering the PEEM geometry the incident plane is described by the propagation direction of the incident photon beam and the surface normal. Due to the high incident angle of the light source, the linearly polarized vectors are oriented in a plane that is tilted 13° from the surface normal. Therefore under PEEM geometry, s-polarized light is parallel to the plane of the substrate, whereas p-polarized light is 13° from the surface normal, as shown in Fig. 5.
PEEM imaging requires a conductive substrate for continual electron replenishment during photoemission. Most of the conductive substrates are reflective and reflect light from the incident photon beam. For a biological specimen characterized with a curved three-dimensional geometry, the total electric field at the sample surface will not just include the incoming light but also include the light that is reflected off the substrate (Fig. 6a). A curved biological specimen is susceptible to this interaction because the reflected beam is able to strike the surface along the curved region. However, a disk or a box-like structure has vertical edges that block the whole object from this reflection (Fig. 6b), whereas a film covers the substrate and prevents it entirely. It is important to understand this substrate effect for the analysis of three-dimensional biological samples in PEEM.
The reflection coefficients that define the reflectivity of the substrate for a specific polarization are calculated through Fresnel equations (49), Eq. (2)
where nsil and nvac are the refractive indices of silicon and vacuum, respectively, and θi is the angle of incidence. The refractive index of a vacuum is 1, but the refractive index of the substrate is dependent on the wavelength of the incident light and the temperature of the substrate. For the purposes of this review, we will focus on silicon as the conductive substrate, as it is a common substrate, and reasonably assume a temperature of 25°C. As for the wavelength, we will examine 207.0, 243.5 and 310.5 nm (5.99, 5.09 and 3.99 eV, respectively), because typical threshold scans are obtained by scanning over the region from 4 to 6 eV. At 25°C and these incident wavelengths, the refractive index of silicon is 1.2215, 1.7385 and 4.5401 (50), respectively. Figure 7 shows the reflection coefficients for s- and p-polarized light as a function of the angle of incidence, measured from the surface normal, of the incoming light. The Brewster angle of silicon for 207.0, 243.5 and 310.5 nm is 50.7°, 60.1° and 77.6°, respectively. Consequently, considering the incident angle of PEEM, ∼77°, polarization-dependent reflection effects must be considered in analyzing the images.
The total electric field, for each polarization, at a point on the surface of a curved biological sample consists of the incoming wave plus that of the reflected wave, Eq. (3).
In this equation, R is the reflection coefficient characteristic of the substrate (Eq. 2) q is the vector location and kinc and kref are the incident and reflected k-vectors. For simplicity, the electric field contribution from the laser, E0, is factored out and normalized to 1—this value only affects the magnitude of the total electric field. The transverse component of the k-vectors is conserved after reflection and therefore can also be factored out. Equation (4) then describes the propagation of the light wave in the direction of the surface normal for both p- and s-polarized light.
In the above expression, z is the distance from the substrate and the z-component of the k-vector is defined as with λ the wavelength of the incident light. Thus, as shown in Eq. (4), the total electric field at a particular wavelength is dependent upon the angle of incidence and the distance above the substrate. Figure 8 shows the total electric field as a function of height at the PEEM incident angle of 77° and λ = 207.0, 243.5 and 310.5 nm. The curves shown in Fig. 8 are for pure s- and p-polarized light sources, but combinations of these two light polarizations can describe virtually any light source. For biologically relevant heights, nm range, the total electric field between s- and p-polarized light varies significantly at the relative wavelengths used in a threshold scan. Nonuniform emission resulting from variations in the total electric field will be present along the curved surface of a biological specimen with significant height due to this effect at all wavelengths.
It is important to note that in addition to the fluctuation of the total electric field along a curved surface (due to the substrate reflection), the sample specimen is also characterized with its own unique dielectric properties and utilization of a particular polarization could enhance the emission of photoelectrons. Materials with larger dielectric constants are more sensitive to polarization (39) and signals could be greatly enhanced or inhibited by the polarized light source. Thus, disk or box-like samples will exhibit a fluctuation of the electric field that is a consequence of the dielectrics of the sample whereas curved samples will experience a fluctuation due to the dielectrics of both the sample and the substrate. Experimental fits of polarization curves for a gold nanosphere and a gold nanodisk are shown in Fig. 9. The curves are obtained by fitting the measured integrated brightness of the PEEM image as a function of polarization. Despite being of the same composition, the curve for the nanosphere is significantly shifted from the curve of the nanodisk. This is expected to be primarily a result of the added substrate effect for the gold nanosphere. Hence, when analyzing biological materials with larger dielectric constants and curved surfaces, polarized photon beams should be an important consideration.
Determination of threshold potentials
To acquire the photoionization threshold of a specimen, UV wavelengths are scanned and the integrated brightness of the image is analyzed as a function of the excitation energy. The measured brightness of the photoelectron image is taken to be proportional to the photocurrent as the emission current cannot be accessed in a standard PEEM. An image is obtained for each wavelength and its integrated brightness is measured by a summation of all of the pixel gray scale depths in the selected region. The minimum and maximum gray scale values are regarded as black or white saturated, respectively. Images with a large fraction of the pixel counts located at these values result in a loss of the information regarding the brightness of a granule and the photoelectron current is no longer proportional to the integrated brightness.
In addition to the previously discussed considerations of spatial and polarized signal dependence for three-dimensional biological materials, other standard data corrections are needed for threshold measurements. The voltage on the multichannel plate and/or the software gain of the imaging program is adjusted to ensure the intensity histogram of the image spans a distribution across the pixel gray scale depths to avoid black and white saturation. Also because of adsorption, the photon flux of the incident wavelength changes as one scans through the UV region and hence it is necessary to measure and compensate for this fluctuation.
To obtain the photoionization potential of the sample, threshold curves are functionally fit with early theoretical models proposed to explain photoelectric curves. In 1931, Fowler (51) described the photoelectric curve of solid metallic surfaces with an expression that is a function of temperature and the frequency of the incident light. The integrated brightness of the sample, S, proportional to the photocurrent, I, is in proportion to the product of the square of the temperature of the sample, T, and a function f(u) given by:
In the above expressions, , v is the frequency of the incident light, kB is the Boltzmann constant, h is Planck’s constant, χ0 is the threshold potential and χ is the thermionic work function (χ = χ0 − ɛ*, where ɛ* is the energy of the highest occupied molecular orbital). The threshold can thus be determined by functionally fitting the threshold curves of S/T2vs hv/kBT. The Fowler equation has been used to determine the surface photoionization threshold potentials of a variety of human pigments (33–38) and will be discussed later.
A simplified model, the Fowler-Nordheim law (52), is also used to determine photoionization potentials. The Fowler-Nordheim law describes the photoelectric current by:
where C is a constant dependent on the sample, χ is the threshold photoionization potential and hv is the photon energy. The photoionization potential is obtained by extrapolating the square root of the integrated brightness of the PEEM image, S1/2, to zero on a plot of S1/2vs the photon energy. This model may be simpler, but it implicitly assumes the measurement is being made at extremely low temperatures as it cannot account for the “tail” that appears in the photoionization threshold curve of room-temperature (or heated) samples, which originates from a thermalized (Boltzmann) density of states serving as the ground state from which photoionization occurs. This can lead to misinterpretation of data collected at room temperature. Although the temperature-related emission has a much smaller contribution than the actual threshold emission at room temperature, uncertainty from the extrapolation method may still exist due to the thermal energies of the electrons (53).
In 1962, Kane (54) proposed a model for the photoelectric emission of semiconductors that involved direct and indirect processes with yields proportional to E − χ and (E − χ)5/2, respectively. Following Kane’s theory for semiconductors, Kochi et al. (55) established a cubic power model to explain photoemission from organic crystals. These models, like the Fowler-Nordheim model, do not include the effect of temperature and therefore do not account for the thermal tail of the threshold curve. The above theoretical models are based on photoelectric curves of metallic, semiconductor or organic surfaces, and such systems differ from biological specimens. As discussed below, the applicability of the Fowler analysis to biological and molecular systems remains to be rigorously validated.
Wilson et al. (56) obtained the ionization energy of biological nanoparticles by analyzing the photoelectron kinetic energies with use of a velocity-map imaging photoelectron spectrometer. Extrapolation to zero of the linear relationship between the maximum kinetic energy release and the incident photon energy provides the ionization energy according to Einstein’s equation. Wilson et al. confirmed the ionization energy of the biological nanoparticles with the Fowler-Nordheim model and a plot of the total electron yield vs the incident photon energy. The same value was obtained upon utilization of both techniques. Additionally, studies of black human hair eumelanosomes using PEEM in the absence of heating revealed similar results, within experimental uncertainty, between the Fowler model (33,36) and the temperature-independent Fowler-Nordheim model (37,38). Recently, threshold measurements of the biological protein fibrinogen were analyzed according to the Fowler model, the Fowler-Nordheim model and the cubic power model (32). A comparison of the photoionization potentials obtained from all three theoretical models revealed no significant deviations and a conclusion that all can be equally used to analyze a threshold curve (in the absence of heating) was drawn. From these results, photoionization potentials of biological materials can be obtained with models established on nonbiological systems.
The connection between the vacuum threshold potential and the electrochemical potential vs NHE has been explored in great depth (57–61). The relationship between the two scales has largely been derived by comparing electrochemical and work function data on Hg. The IUPAC-recommended relationship is that 0 V vs NHE corresponds to −4.44 V in vacuum. However, it is worth mentioning that the studies cited above report values corresponding to 0 V vs NHE between −4.4 V and ∼−4.8 V. Generally, measurements incorporating the contact potential difference of the uncharged metal and the work function of the clean surface result in higher values in this range whereas work on immersed electrodes, which are believed to retain their interfacial region, result in lower values. The analysis of wavelength-dependent PEEM images is uniquely capable of providing electrochemical data for the surface of the intact human pigments, which cannot be determined by conventional approaches.
PEEM studies of human pigments
In this section, the application of PEEM to biological problems is exemplified by examining some of our recent work on human melanosomes. Melanosomes are organelles found in several regions of the human body (skin, eye, hair, inner ear and a related structure in the brain), whose functions include photoprotection, mitigation of the effects of reactive oxygen species and/or metal chelation. The dominant constituent of the melanosome is the pigment melanin. In general, the functions ascribed to melanosomes are enabled by the melanins they contain. Melanin is generally classified into two major types, eumelanin and pheomelanin, reflecting different molecular precursors in melanogenesis. In the last decade, there has been significant progress in understanding melanins and their impact on human health. Studies designed to probe the surface properties of melanosomes are few, yet the physical and chemical properties of the surface of the melanosome are inextricably linked to function (62).
First, we examine work demonstrating how PEEM uniquely enables determination of the surface electrochemical properties of different types of human melanosomes. The relevance of the threshold potential data collected under high vacuum conditions to that exhibited by pigments in physiologically relevant buffer solutions is established by comparing results from different experimental methodologies. Figure 10 shows the experimental data for melanosomes isolated from human black and red hair (36,37). The fit of these data to the Fowler equation reveals that black and red hair melanosomes both have a surface photoionization threshold of 4.4 ± 0.2 eV (282 nm). To accurately fit the data for red hair melanosomes, a second threshold potential of lower energy 3.8 ± 0.2 eV (326 nm) is required.
A common ionization potential for the two pigments is not surprising. The human red hair melanosomes are not “pure” pheomelanin; chemical analysis reveals they are ∼25% pheomelanin, 75% eumelanin (63). Thus, they should reveal the threshold signature exhibited by the black hair melanosomes, which are pure eumelanin. The additional lower ionization threshold observed for red hair melanosomes is therefore attributed to pheomelanin.
It is important to establish that the threshold potential of melanosomes observed in isolation under high vacuum (the conditions of the PEEM experiment) are relevant to how the organelle behaves under physiologically relevant conditions. While the intact organelle is not amenable to traditional electrochemical techniques, there are related experiments that can be performed to address this issue. Specifically, we examined femtosecond transient absorption spectroscopy and electron paramagnetic resonance oximetry experiments of synthetic pheomelanins to determine the threshold for photoionization of pure pheomelanin in solution at physiological pH (36). Figure 11 shows these data. These model experiments were designed based on the report by Chedekel et al. that UV excitation of synthetic pheomelanin in solution resulted in the formation of the superoxide radical anion, being formed by a mechanism where the first step is photoionization of the pigment (64–66):
Figure 11 shows the absorption spectrum of pheomelanin and the action spectrum of Chedekel et al. for the formation of the superoxide radical anion (64). We attribute the constant rate of superoxide formation reported by Chedekel et al. in the region of λ > 400 nm to a dark reaction (36), and so once corrected for this dark reaction, these data would represent the action spectrum represented by the dashed line, where photogeneration of superoxide now becomes appreciable when λ < 330 nm. Data from our studies using femtosecond absorption, electron paramagnetic resonance oximetry and PEEM are also indicated in this plot. The femtosecond transient spectroscopy experiments were designed to detect the presence of solvated electrons following photoexcitation. Such signals are observed following excitation at 303 nm, but are not found for excitation at 350 nm, revealing that the threshold lies between these two wavelengths (36). A detailed action spectrum using this approach was not collected due to the complexity of the experimental technique. Electron spin resonance oximetry experiments narrow this range, and were significantly easier to implement as excitation is achieved using a conventional lamp and narrow bandpass optical filters. These data show that the threshold for the synthetic pigment lies between 338 and 323 nm (36). Finally, we recall that PEEM reveals a threshold potential of 326 nm (36,64). There is remarkable consistency between these experiments. More importantly, there is excellent agreement between the synthetic pigment in buffer solution and the intact human melanosome under the isolated high vacuum conditions characteristic of PEEM.
Having now established that the threshold potential determined by PEEM is relevant to pheomelanin under physiological conditions, it is of interest to consider the implications of the value obtained. Specifically, consider the solar radiation impinging on the surface of the Earth. Figure 12 shows the solar irradiance at the Earth’s surface for several different solar zenith angles (36,67), clearly revealing that humans are exposed to the wavelengths of light needed to ionize pheomelanin, but not those needed to photoionize eumelanin. As a result, the lower ionization potential observed for pheomelanin could be a part of the explanation for the greater incidence rate of UV-induced skin cancers in the populations whose melanins contain increased concentrations of pheomelanin (68,69).
As a second example of the application of PEEM to human pigments, we consider the spatially resolved images on human neuromelanin. Neuromelanin is composed of both eumelanin and pheomelanin, with pheomelanin constituting about 25% of the total pigment (70,71). The spatial distribution of these two different pigments within the organelle remains an important problem to solve, especially because pheomelanin has been shown to be able to induce oxidative stress, whereas eumelanin mitigates such stress (72). The above discussion demonstrates that the threshold potentials of pheomelanin and eumelanin can be distinguished by PEEM, and so the spatial analysis of data on human neuromelanin granules would provide information on the pigment present on the surface of the granule.
Figure 13 shows a plot of the integrated intensity of the PEEM image of pigment granules isolated from the substantia nigra of a human brain as a function of wavelength, and the associated fit of the Fowler equation (34). The fit of the theoretical expression to the experimental data is excellent, and reveals a photoionization threshold potential of 4.5 ± 0.2 eV (282 nm) (34), which is within experimental error of the oxidation potential measured for human eumelanosomes. There is no indication in the data that pheomelanin is on or near the surface of the granule. Detailed studies of the kinetic studies on the early chemical steps of melanogenesis show that in the case of pigments containing a mixture of pheomelanin and eumelanin, pheomelanin formation occurs first with eumelanin formation predominantly occurring only after cysteine levels are depleted (73). Such a kinetic model predicts a structural motif with pheomelanin at the core and eumelanin at the surface. This “casing” model was originally proposed in 1982 by Rorsman (74) based on biochemical evidence and has received indirect support from several other studies (75–77). The PEEM data provide definitive data in support of such a structural model.
This structural model has significant implications for the role of neuromelanin. Dopamine, the precursor to neuromelanin, is cytotoxic to neuronal cells through oxidation to dopaminequinone (78). The formation of cysteinyldopamines is hypothesized to be a protective mechanism, preventing formation of the neurotoxic dopaminequinone. A casing model for the structure of neuromelanin therefore provides a mechanism to capture dopaminequinone as part of the pheomelanic core, removing this neurotoxic precursor, and then shielding it from the neuron by encasing it in a eumelanin coat.
It is also interesting to consider this result in terms of the pathology of Parkinson’s disease (PD). In PD, there is a selective loss of pigmented neurons in the substantia nigra (79). Unlike the case with pheomelanin, a surface composed of eumelanin is not sufficiently reductive to generate a high level of oxidative stress. Therefore, neuromelanin’s surface electrochemical potential is less reductive than might be anticipated given neuromelanin’s complicated but unresolved role in the selective loss of pigmented neurons of the substantia nigra associated with PD. However, degradation of neuromelanin would expose the pheomelanic core, which would then cause oxidative stress. Thus a slow loss of pigment could actually result in a constant level of inflammation.
Challenges in elucidating molecular information on biological systems using PEEM
The previous section, along with the earlier work by Griffith, clearly reveals that PEEM provides important insights into biological systems. The technology has matured and there are instruments available for use at most of the light source facilities around the world (for example, see user list at http://www.elmitec-gmbh.com/). Advancements in the applications of this technique require interplay between the development of theoretical analyses and models and experiments designed to understand the effect of morphology on the images obtained. Different classes of materials (e.g. lipids, proteins) have different ionization properties. In addition, because the photocurrent is describable theoretically by a Golden rule formula, the photocurrent is dependent on the overlap of the directions of the incident electric field and the transition moment of the molecule. This latter effect affords the ability to look at molecular orientation and alignment, which contributes to the uniqueness of PEEM among electron microscopies. While such experiments are currently feasible—in principle any system that can be studied by SEM can be examined by PEEM—there are challenging theoretical issues in extracting detailed molecular information from the images obtained. The remainder of this section examines some of these issues, and research in these areas is required for PEEM to achieve its maximal potential as an imaging method.
Establishing the validity of the Fowler and related theoretical models to “molecular” systems or developing molecular-based theories
As stated above, the current models for elucidating threshold potentials, and hence oxidation potentials, rely on fitting wavelength-dependent PEEM intensities to either the original Fowler equation, or an expression derived from a similar set of starting principles. In these theoretical derivations, the number of electrons available for photoionization are modeled in accord with the distribution law of Sommerfeld’s theory of metals. Whether this model can be applied to accurately determine thresholds for biological materials has not been established. However, in order to assess the surface oxidative processes of biological organelles relative to other oxidation/reduction potentials of cellular processes, the accuracy of the potentials derived from the wavelength-dependent analysis needs to be addressed.
Methods for determining the actual imaged region for a three-dimensional object
As exemplified by gold spheres, the actual region imaged in the PEEM depends on the experimental geometry and the shape of the material being imaged. While the analysis is straightforward in the case of thin films, the situation is much more complex when there is significant three-dimensional topography. This also affects experimental protocols for obtaining images. When considering the imaging of spherical objects in the PEEM, the image is not expected to be round because of the incident angle to the light beam. However, because of an expectation that a spherical object should give a round image, the experimentalist may adjust the focusing conditions to make the image round, which actually introduces a distortion to the image.
Depth profiling—electrons can originate from below the surface
While PEEM is a “surface-sensitive” microscopy, it is clear that ionization occurs from atoms (in metallic films) and molecules (in organic and biological systems) that lie below the surface. The depths from which electrons are generated depend on many variables. The penetration of light into the material depends on the oscillator strength. The escape yield depends on how rapidly the surrounding material dissipates the kinetic energy of the photogenerated electron and the rate constants for various electron-scavenging processes that can occur, especially in complex media such as biological systems. As a limiting example, one could image a complex system in which a molecule (or material) is encased by materials of higher ionization potential, which are transparent to the wavelength of light corresponding to the threshold potential of the encased molecule (or material). Such situations could include specific types of layered quantum dots, or the neuromelanin granules present in the human brain (34) where pheomelanin (threshold potential of 3.8 eV) is encased in the black eumelanin (threshold potential of 4.4 eV). In these cases, one may observe the lower threshold material in the PEEM images for the corresponding wavelengths, but the signals are generated from within the systems being imaged, not from their surface. Thus, with the application of this technique to complex media that contain molecules with varying oxidation potentials, there is the need to develop methods for distinguishing surface processes from those occurring inside the samples of interest.
Understanding the factors that control the spatial intensity distribution in a PEEM image
Because of the size of biological systems, the substrate effects discussed above contribute to the spatial intensity of the PEEM image. In addition, the magnitude of these effects changes with wavelength because both the reflection from the surface and the height dependence of the resulting optical interference effects are wavelength dependent. Thus, in order to develop insights about the biological (or materials) systems of interest from the spatial PEEM image, the substrate effects must be well understood so that attention is focused on the molecular properties of the system being imaged. In addition, there are instrumental effects that must be considered. The instrument is designed to image the plane of the sample directly onto the microchannel plate/CCD detector with constant pixel dimension over the image. However, if there is curvature in the sample, the actual surface area of the material of interest can be quite different from pixel-to-pixel. Thus, caution must be exercised in interpreting the effects of spatially dependent intensity maps. Reliable methods for disaggregating all of these effects are required, and this necessitates the development of a standard set of control experiments for studying biological assemblies.
Acknowledgements— This work was supported by Duke University. We especially thank Professor Robert Nemanich, who has collaborated with our group on many projects, for allowing us to continue to use his PEEM instrument located at the Free Electron Laser Lab at Duke. We thank Professor David Smith, Professor David Brady, Dr. Jack Mock, Jonah Gollub, Professor Jack Rowe, Xianhua Kong, Professor Glenn Edwards and Professor Ying Wu for helpful discussions on various aspects of this work. We further acknowledge those researchers that have been involved in our PEEM studies on biological systems: Alexander Samokhvalov, Yan Liu, Lian Hong, Jacob Garguilo, William Bush and Luigi Zecca.