Chemical imaging of aggressive basal cell carcinoma using time-of-flight secondary ion mass spectrometry Open Access

Basal cell carcinoma (BCC) is the most common cancer in humans.¹ The BCC incidence is reported to be between 77 and 158 per 100 000 person years in Europe.² The incidence in Sweden has increased almost tenfold over the last 30 years. Estimations indicate that 15% of the Swedish population will suffer from a BCC by the age of 75 years.¹ Exposure to ultraviolet radiation is considered the main causative risk factor.² Thus, the majority of BCCs are seen in patients older than 50 years and on sun-exposed parts of the skin (e.g., the head and neck area). Clinically, a BCC can present as a shiny bump, red patch, or sometimes as a scar.³ Although not a lethal tumor in the majority of cases, aggressive BCC types with destructive growth can cause remarkable morbidity to the affected person, especially when they grow around the eyes, nose, mouth, and ears.² 

Although the cell of origin is still controversial, BCCs have been thought to originate in the skin's basal keratinocytes, which are located just above the dermoepidermal junction and in the hair follicles.² Histologically, BCCs form a group of intradermal epithelial tumors which can be classified by their morphological growth pattern, into indolent (nodular and superficial) or aggressive (micronodular, morphoeic, infiltrative, and metatypical or basosquamous) subtypes, while some cancers show mixtures of these types.⁴ However, histology does not provide information of the biomolecular background of the tumor.

New insights in understanding the genetic and molecular changes behind BCCs have arisen from studies of patients with genetic syndromes associated with higher risk of BCC development. The Hedgehog (Hh) receptor Patched 1 (PTCH1) was first found as the diseased gene in Gorlin's syndrome, a syndrome in which patients develop multiple BCCs starting at an early age. Later research has shown that almost all BCCs have an atypical activity in the Hh receptor.² 

The burden of BCC leads to high societal costs secondary to treating the tumor.⁵ The current gold standard when treating aggressive BCC subtypes is Mohs micrographic surgery (MMS), a technique offering 100% margin control with stepwise histopathological evaluation of frozen sections between each session of surgery until all cancer cells have been removed.¹ For patients with Gorlin's syndrome and patients with locally advanced or metastatic tumors that are unsuitable for surgical treatment or radiotherapy, expensive oral Hh inhibitors can provide partial to complete responses in 30%–60% of cases.² Nevertheless, other alternatives for these patients may be developed with a better understanding of the biochemical and molecular nature of BCCs.¹ 

There is a growing interest to better understand the role that alterations in lipid metabolic enzymes and their pathways have on oncogenesis.⁶ In regards to BCC, studies have shown that there are significantly increased levels of total lipids and phospholipids in these tumors when compared to healthy skin tissue.^7,8 Interestingly, the activation of the Hh pathway has also been shown to depend on the levels of specific phospholipids.⁹ 

One way of examining and characterizing lipids in cancer tissues is secondary ion mass spectrometry (SIMS). Advances in SIMS technology have provided significant improvements in analytical capabilities in recent years including new variants of time-of flight (ToF) SIMS instruments, sometimes utilizing quadrupole-ToF and Fourier transform (FT) based MS analyzer technology,^10–12 and new ion beams to improve signal levels from intact molecular species. Specifically gas cluster ion beams (GCIBs) that are now routinely used for low damage etching of samples for SIMS and x-ray photoelectron spectroscopy analysis can also be used as analysis beams.¹³ Although the ionization efficiency of secondary species is low when used at low energy, GCIBs provided a ×30–50 increase in intact lipid signal per primary ion on rodent brain compared with C₆₀⁺ when both beams were accelerated through 40 kV.¹⁴ These technological advances have greatly enhanced the capabilities of SIMS for probing complex biological systems such as drug effects on Drosophila brain, chemical changes in breast cancer biopsy tissue, bacteria, and cardiolipins in rodent brain.^15–18 

In this study, a 40 keV (CO₂)_6k⁺ GCIB was used to analyze the lipid distributions in aggressive BCCs compared to surrounding normal skin by using ToF-SIMS. Differences between high and low aggressive tumors were also identified.

In this study, nine BCCs verified as aggressive during MMS were included (patients #1–8, and #10). For comparison one punch biopsy of a less aggressive BCC (patient #9) was added.

Fresh frozen BCC tissue was obtained during MMS performed at Sahlgrenska University Hospital in Gothenburg, Sweden, after approval from the Regional Ethical Review Board of Gothenburg. Prior to surgery all patients gave informed consent. MMS was performed routinely using an incision at a 45° angle enabling flattening of the specimen for sectioning. The specimen was oriented by using orientation cuts dyed with different colors. The specimen was frozen using Histolab^® cryospray (Histolab AB, Sweden) and was thereafter fixed using the optimal cutting temperature compound Histolab OCT Cryomount (Histolab AB, Sweden). Horizontal sectioning was performed using a cryostat (Leica CM1520) at −20 °C with a thickness of 10 μm per slice. The thin tissue slices were thaw mounted on to indium tin oxide glass (ITO-glass), which is conductive and transparent. Hematoxylin and eosin (H & E) staining of consecutive sections was performed. The sectioning technique allowed for a horizontal view of the whole skin thickness including epidermis, dermis, and subcutaneous structures (Fig. 1).

The extra tissue specimens used for analysis with ToF-SIMS did not cause the patient any harm or discomfort and did not delay the surgical procedure.

As the sectioning and pathological analysis of these samples was performed in parallel to the surgery to decide if more tissue should be excised, tissue handling (mounting, slicing, and thaw mounting) was performed by the clinical staff at the hospital and not in the MS lab. The samples were transferred to the MS lab on dry ice and stored at −80 °C before being placed in a vacuum desiccator for 1 h prior to SIMS analysis. To ensure that sample preparation artifacts would not appear due to chemical relocation, a preliminary SIMS analysis was performed to check what chemical signals were present on and just below the tissue surface.

ToF-SIMS analysis was performed using a J105—3D Chemical Imager (Ionoptika, Ltd., UK) ToF-SIMS instrument described in detail elsewhere.¹² Briefly, the instrument uses a quasicontinuous primary ion beam to produce a stream of secondary ions that are bunched to a time focus at the entrance to the reflectron mass analyzer. In this study, a 40 keV GCIB was used to bombard the sample with (CO₂)_n⁺ cluster ions. A Wien filter was used to select a cluster peak cluster size of n = 6000 (approximately ±2000). Throughout the manuscript, the beam will be referred to as (CO₂)_6k⁺. Mixed Ar/CO₂ GCIBs have been shown to produce a larger mean cluster size at equivalent operating pressures compared to pure Ar beams. Additionally improved spot sizes, most likely due to a narrower cluster size distribution, have also been reported using CO₂ cluster beams compared to Ar on the same ion optical column. CO₂ is also cheaper.^14,19 An initial sample was used to examine the optimum analysis conditions before a further nine biopsy samples were analyzed in both positive and negative ion mode. One biopsy is not included in the results due to poor quality of the tissue preparation that distorted the histology and also resulted in increased sample charging during SIMS analysis (patient #8). A final section (patient #6) was not analyzed and will be used for further studies beyond this manuscript.

The SIMS images shown in Figs. 3–6 were acquired at 25 μm/pixel resolution with a primary ion dose density of 3.2 × 10¹¹ ions/cm² based on optimization described in Sec. III A. The m/z range 100–2000 was selected and mass resolution (m/Δm) was approximately 5000–8000 in the intact phospholipid (ca. m/z 700) range. Low energy (12 eV) electron flooding was employed in negative ion mode. Internal mass calibration was performed using commonly observed biomolecular peaks at m/z 369.35, 551.50, 734.57, {positive ion mode: Cholesterol, diacylglyceride (DAG)/triacylglyceride (TAG) C32:0 fragment, and phosphatidylcholine PC(32:0) [M+H]⁺, respectively} and m/z 465.30, 581.31, 885.55 {negative ion mode: Cholesterol sulfate [M−H]⁺, phosphatidylinositol (PI)-lipid fragment, and PI(38:4) [M−H]⁻}.

Maximum autocorrelation factor (MAF) analysis was performed in matlab (version 2017a, TheMathWorks). Data was loaded into matlab and the time resolution reduced by a factor of 4 (1 to 4 ns) to reduce the size of the dataset. MAF was performed as described by Henderson et al. on the m/z 100–1000 mass spectral range.²⁰ A large part of the analyzed area corresponded to the ITO coated glass substrate so the signal from these pixels was removed by thresholding the data based on the first MAF that classified tissue from substrate and the analysis was rerun on the reduced data set similar to the process previously applied to principal components analysis of breast cancer biopsy samples.¹⁶ MAF “scores” were displayed in images where positive scoring pixels are green and negative scoring pixels are red. Pixels with no variance in the specified MAF are black.

The use of GCIBs on bulk tissue samples such as these overcomes the traditional “static” surface analysis restrictions of organic SIMS experiments. It was therefore decided that the optimum ion dose density for analysis of the biopsy sections should be determined. Figure 2 shows the results from the negative mode analysis of the test sample where the dashed line [Fig. 2(a)] shows a smaller area that was analyzed 32 times to an accumulated dose density of 1.5 × 10¹³ ions/cm². The erosion rate, and hence analysis depth, was not measured for this sample but for comparison this would correspond to the removal of approximately 75 nm of Irganox 1010 (a common organic standard for ToF-SIMS that has been used in several interlaboratory depth profiling studies).^14,21,22 Several different series of signals were observed, some of which can be seen in the individual ion images [Figs. 2(c)–2(l)]. While signals from many chemical species showed little or no change under increasing ion bombardment several showed a signal transient at low ion dose density with signal levels either decreasing or increasing before quickly stabilizing. It was apparent that the GCIB was not causing any significant damage to the intact lipids within this dose density range. The surface/subsurface variation was mainly observed in very lipid-rich regions of the tissue such as sebaceous glands. This was attributed to possible differential surface migration/segregation or minor smearing during the sectioning process while additional matrix effects may also play a role.

Based on this test, an experimental design was adopted for the further analysis of biopsy tissue specimens where the whole, or a large area, of the tissue sample would be imaged in the sequence of positive–negative–positive–negative ion mode with a primary ion dose density of 3.2 × 10¹¹ ions/cm² for each analysis. While the first positive and negative images would be kept as a reference, further data analysis would be performed on the second set of images ensuring that surface transient signals and any migrating/smearing species did not distort the interpretation.

The structure of the skin biopsy is complex and contains many different structures that are heterogeneously distributed laterally and vertically in the biopsy sample. MAF analysis highlighted many chemically different regions in the tissue (down to MAF16) with the most prominent areas of interest captured in MAFs 2, 3, and 7 as shown in Fig. 3 for patient #7. Positive scoring pixels associated with each factor are colored green while the negative scoring pixels are red. MAF2 clearly distinguishes between the “normal” and cancerous tissue (green and red, respectively) where the healthy tissue shows strong highly localized signal from sebaceous glands and hair follicles with weaker signal (still colored green) between these features in the dermis and stromal tissue [Fig. 3(a)]. The cancerous tissue is displayed in red and surrounds some of the hair follicles in the upper part of the image while the lower part of the image shows little normal skin structure (glands/follicles), which correlates with the more purple regions in the image of the H & E-stained consecutive tissue slice [Fig. 3(d)]. MAF3 highlights an area in the center of the specimen [shown in red in Fig. 3(b)]. Pathological assessment of the H & E-stained consecutive section indicated that this region contained fibrotic stroma possibly associated with earlier surgery. Particularly localized chemical differences associated with the epidermis and hair follicles were highlighted in factor 7 [green pixels in Fig. 3(c)].

The loadings for MAF2 [Fig. 3(e)] show a series of negative loading peaks that can be attributed to the cancerous parts of the tissue. The five strongly negative loading peaks [Fig. 3(e)] were measured at m/z 184.07, 369.35, 542.50, 666.48, and 725.55. The peak at m/z 184.07 is almost ubiquitous in SIMS analysis of biological samples, corresponding to the PC lipid head group, and is formed through fragmentation of PC and sphingomyelin (SM) lipids. The peak m/z 369.35 is assigned to the [M+H-H₂O]⁺ ion of cholesterol. Negative ion mode analysis of skin tissue shows extremely high signal levels for cholesterol sulfate localized to the epidermis, hair follicles, and sebaceous glands [see m/z 465.3 signal in Figs. 2(a) and 2(b)]. Although changes in the apparent distribution of these species due to matrix effects cannot be ruled out, the different distributions of the cholesterol (cancerous tissue in positive ion mode) and cholesterol sulfate (healthy tissue in negative ion mode) suggests that the cancerous signal is from a different form of cholesterol (free cholesterol or cholesterol esters) and not a fragment of cholesterol sulfate. The lipid maps database suggests several possible assignments for the m/z 725.55 with the most likely candidates being a SM lipid or a phosphatidylethanolamine-ceramide. The coloading peaks at m/z 666.48 and m/z 542.50 strongly support the SM assignment and represent the following series of ions: [M+Na]⁺ [SM(34:1), m/z 725.55], [M+Na-TMA]⁺ [where TMA is trimethyl amine (59 Da), m/z 666.48], and [M+Na-PC]⁺ (m/z 542.50). Loading plots and score images for the first 16 factors in both positive and negative ion modes can be found in Figs. S1–S4 in the supplementary material.²⁶ 

The individual ion images of the four highest loading peaks associated with the cancerous tissue, healthy tissue/sebaceous glands and the fibrotic stroma are presented in Figure 4 showing the SM(34:1) ions [Figs. 4(b)–4(e)] detailed above and m/z 571.48, 754.54, 198.09, and 146.98 [Figs. 4(g)–4(j)] where the m/z 754.54 and 571.48 peaks are attributed to PC(32:1) as [M+Na]⁺ and [M+Na-PC]⁺, respectively. The m/z 146.98 peak is assigned as the rearrangement of the PC head group with Na adduct (phosphonoacetaldehyde cationized by sodium) the peak at m/z 198.09 appears to be correlated with PC possibly [PC+CH₂]⁺. The fibrotic stroma and adipocytes are characterized by peaks at m/z 577.52, 603.53, 575.50, and 601.52. These peaks are commonly associated with DAG type ions although their assignment is ambiguous as they may represent the [M+H-H₂O]⁺ ions of DAGs or the [M-RCOO]⁺ ions of the generally more abundant TAGs which is in good agreement with the observation of adipocytes in the H & E image.

Further inspection of the raw mass spectral data showed a second series of [M+Na]⁺, [M+Na-TMA]⁺, and [M+Na−PC]⁺ localized to the cancerous area that could also be attributed to increased SM content this time corresponding to SM(42:2).

The negative ion mode data were dominated by several very intense signals originating from the hair follicles and sebaceous glands including fatty acid (FA) [RCOO]⁻ ions and cholesterol sulfate [M−H]⁻ at m/z 465.3. While MAF3 clearly discriminated between the glands and the hair follicle/epidermis (see supplementary material), due to the dominance of the signals from these normal skin structures, the MAF analysis failed to produce any clear scores images that correlated with the cancerous part of the tissue. As an alternative approach, lower intensity peaks were manually selected and imaged.

Due to the heterogeneity of the sample set (aggressive BCC, mixed type BCCs with aggressive component and different locations on head area) comparison of different potential cancer-related ions across the patient cohort is critical. Figure 5 shows the H & E images along with single ion images of species that are consistently distributed in the cancerous areas in different patients. Specifically, the [M+Na]⁺ ions of the SM(34:1) and SM(42:2) discussed above along with ions selected from negative ion mode images at m/z 687.54, 788.54, and 885.55. The m/z 687.54 ion has previously been detected in studies of muscular dystrophy using ToF-SIMS and is assigned as a negative ion fragment of the SM(34:1) observed in positive ion mode.²³ The peak at m/z 788.54 is assigned to the phosphatidylserine, PS(36:1), while the m/z 885.55 ion is assigned to PI(38:4).

While the identified species correlate well with tissue regions showing aggressive BCC growth pattern several additional observations can be made regarding the different patients. Patient #10 showed particularly high lipid signal for all species in the normal tissue regions compared to the cancerous areas so while the peaks imaged in Fig. 5 were relatively more intense than other species in the cancerous region the clear delineation of the cancer by the sphingomyelins was not possible in this case. Patients #3 and #4 showed a mixed cancerous growth pattern with regions of aggressive and more indolent tumor growth. In patient #3, all of the tissue in the SIMS image was cancerous and the hot spots of signal are from larger, less aggressive, tumor growths that are more clearly resolved from the stroma. Pathological assessment of the patient #3 sample also identified a less aggressive tumor region to the top left of the image that does not show increased signal from all of the peaks imaged in Fig. 5. A similar situation is present in patient #4 where the sphingomyelin peaks are localized to a band of tissue running diagonally from top left to bottom right of the SIMS image. This area was assessed as aggressive cancer growth while the upper right of the image showed normal pathology and the lower left showed a less aggressive growth pattern. The data from these mixed cancer type samples was compared to that from a punch biopsy of a nodular BCC that was taken from a patient undergoing Mohs surgery (Fig. 6). Red lines on the H & E images [Figs. 6(a) and 6(b)] of the mixed cancer samples show the approximate boundary between the cancerous areas that are less and more aggressive while the blue line indicates the approximate demarcation of the diseased and normal tissue. The less aggressive cancer also appears as a slightly darker blue/purple in these H & E images. As discussed above, and illustrated in Fig. 5, the less aggressive tumors did not show the increase in the sphingomyelin signals. Instead, phosphatidylcholine species were increased in these regions particularly PC(32:0) and PC(34:1) [Figs. 6(d)–6(o)]. Patient #3 showed a particular increase in the PC(32:0) in the less aggressive tumor while PC(34:1) was of greater intensity in the less aggressive tumor region in patient #4 and in the punch biopsy from patient #9. While the [M+Na]⁺ and [M+K]⁺ ions of each of these PC lipids showed similar distributions in the biopsy from patient #9, distinct differences can be observed in the mixed cancer Mohs samples of patients #3 and #4. In patient #4 the [M+Na]⁺ signal of PC(32:0) in the less aggressive tumor region is lower than that in the aggressive tumor region while still higher than in the normal tissue, whereas the [M+K]⁺ signal is of similar intensity in both tumor regions. The PC(34:1) shows a similar trend, but in this case, the [M+Na]⁺ signal is only slightly lower (on average) in the less aggressive region and the [M+K]⁺ signal is higher in the less aggressive region than in the aggressive tissue area. The difference in the relative distributions of the two adduct ion types may indicate a possible isobaric interference in the mass spectrum or a change in the local cation concentrations in the different tissue environments. Changes in Na⁺ and K⁺ adduct formation have previously been associated with changes in metabolic states of tissue samples in ischemic brain samples imaged with nanoDESI MS and in ToF-SIMS imaging of infarcted mouse hearts where the infarcted (oxygen depleted) region of the tissue produced predominantly [M+Na]⁺ PC lipid ions compared with the noninfarcted tissue where mainly [M+K]⁺ PC lipid ions were formed.^24,25 For PC(32:0), patient #3 shows an increase in signal in the top left of the image for both the [M+Na]⁺ and [M+K]⁺ ions while for PC(34:1) no increase in the [M+Na]⁺ signal is seen in this region and the [M+K]⁺ shows an increase in signal in a more localized area of the sample that correlates with the “bluest” feature in the H & E stained image.

Previous studies have shown that optimum sample preparation for tissue analysis involves slicing frozen tissue and performing the entire ToF-SIMS analysis without allowing the sample to warm up while also using minimal tissue embedding medium to attach the tissue sample to the holder in the microtome.¹³ It is common to adopt a compromised approach to sample handling in interdisciplinary studies and the analysis of vacuum desiccated tissue is common in matrix-assisted laser desorption ionization and SIMS studies.¹⁴ However, the surface sensitivity of SIMS can render the technique especially susceptible to sample preparation artefacts due to chemical relocation such as surface segregation/migration and smearing during sectioning. It is shown that the use of GCIBs allows detection of intact lipid species at high primary ion dose density so intact molecular imaging can be performed.

The lipid signals detected during the SIMS analysis provide direct anatomical correlation with conventional H & E-based pathology with clear chemical distinctions detected when comparing normal structures associated with sebaceous glands and hair follicles to cancerous and fibrotic areas of tissue.

While the total lipid signal obtained from the tissue samples was dominated by signals from the follicles and sebaceous glands, several chemical species were consistently seen to be increased in cancerous tissue, particularly sphingomyelins. In a previous cancer study with ToF-SIMS,¹⁶ sphingomyelin increase was associated with apoptosis in necrotic areas of tissue but that is not the case here. Sphingomyelin lipids are also associated with lipid rafts and the organization of membrane proteins. A change in these species may indicate a switch in cell signaling mechanisms in the cancerous cells in these samples although at the spatial resolution used for this initial study it was not possible to tell if the SM lipid change was associated with the aggressive cancer cells themselves or from the stromal microenvironment surrounding them. Data from tissue containing mixed cancer types was compared with a punch biopsy nonaggressive cancer lesion and showed that less aggressive tumors do not show the SM lipid increase and instead show increases in PC lipids, particularly PC(32:0) and PC(34:1). Increased PC lipid content has been previously associated with increased cell proliferation in tumors.¹⁶ A potential change in the local cation concentrations was also detected. Our pilot study shows that ToF-SIMS is feasible in detecting different lipids in healthy and cancerous regions of the skin. Furthermore, our findings can help to explain the heterogeneity of BCCs and their varying growth patters indicating that BCC is a group of tumors with different biochemical profiles.

The imaging mass spectrometry measurements were performed at Go:IMS at the Gothenburg University/Chalmers University of Technology Gothenburg, Sweden. The authors gratefully acknowledge financial support from the Swedish Research Council (VR), Västra Götalands Region (VGR).