Slide-free virtual histochemistry ( Part II ) : detection of field cancerization

Tumor-adjacent “normal” tissue constitutes a peri-tumoral field that affects early cancer detection, risk assessment, surgical decision, and postoperative surveillance. Modern genetic analysis has revealed valuable information from this field, but without the spatial resolution of optical microscopy to understand the vital microenvironments that surround individual cells. Rapidly advanced optical imaging techniques free of labor-intensive sample preparation, despite great promise to perform slide-free imaging of cell structure and shift the histology-centered cancer diagnostic paradigm, have lacked compatible and complementary histochemical imaging of cell function or phenotype to interrogate the peri-tumoral field. In the first part (Part I) of this two-part series study, we developed a technique of slide-free virtual histochemistry to phenotype various cells in in vivo animal and ex vivo human tissue. Here, in the second part (Part II) of this two-part series study, we employ this technique to examine various peri-tumoral fields and produce the volumetric histochemical evidence of field cancerization consistent with the structural changes at larger spatial scales. We also link the field cancerization with cancer dormancy in a significant portion of breast cancer patients. © 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement


Introduction
Standard hematoxylin-and-eosin histology (H&E) and subsequent immunohistochemistry (IHC) in preoperative diagnosis of small biopsies and postoperative prognosis or surveillance of tissue samples, along with fast frozen section analysis (FSA) of surgical specimens during intraoperative assessment, form the gold standard for cancer diagnostic management that has not undergone a paradigmatic shift since 1950s.Recently, a wave of novel optical imaging technologies have emerged as volumetric alternatives [1][2][3][4][5][6] (Table 1) to typically twodimensional-based H&E and FSA, and have been used to directly visualize fresh or formalinfixed tissue in a digital form without labor-intensive (and artifact-prone) sample preparation on an H&E (or FSA) microscope slide [7].The digital image data can be converted into virtual H&E images understandable by pathologists, which structurally highlight cell nuclei from the cytoplasm and the extracellular matrix as if the images were acquired by standard H&E [1][2][3][4].The resulting slide-free virtual H&E imaging (SF-vH&E) appears nonperturbative to the sample, even with external fluorescent labeling [1,3], and may thus conserve the valuable tissue specimen as a resource for subsequent molecular/genetic analysis and future use [2].Additional advantages associated with volumetric imaging [7] and label-free intraoperative imaging in situ (in vivo inside the surgical cavity without removing the tissue) [8] reinforce the view that this paradigmatic shift is emergent [9].
However, remaining obstacles in both the non-intraoperative (preoperative or postoperative) assessment and intraoperative assessment must be overcome to enable a full paradigmatic shift.Outside of the intraoperative setting, imaging techniques for slide-free stain-free functional/metabolic assessment are currently lacking.Without this imaging, the usefulness of SF-vH&E would be considerably limited, e.g., to fast but rudimentary cancer diagnosis or determining adequacy of biopsy sampling [1].If conventional IHC or tissueclearing-assisted volumetric IHC imaging [10] (Table 1) is required to follow SF-vH&E, the exact limitation of histology (labor-intensive sample/slide preparation) that SF-vH&E has avoided would be largely retained.In the intraoperative assessment, the benefit to replace FSA with faster SF-vH&E is compromised by a lack of information about the peri-tumoral field with field cancerization [11][12][13][14], which may define surgical tumor margins differently from sporadic tumorigenesis.Specifically, a negative surgical margin validated by either FSA or SF-vH&E may not prevent the local recurrence of a second field tumor in an unresected cancerized field [12].Although it is possible to detect field cancerization in a peri-tumoral field by SF-vH&E, just like Slaughter and associates did in 1953 by H&E [11], the associated interobserver variability to discern subtle precancerous morphologies (hyperplasia, metaplasia, dysplasia, etc.) has fundamentally limited this practice [13,14].There is a need for a quantitative imaging technique that could objectively assess the metabolic and structural properties of the complex microenvironment.
It is then clear that in order to enable a full paradigmatic shift, a slide-free histochemical imaging technique free of labor-intensive sample preparation is needed to phenotype not only the primary tumor cells or tumor-associated cells for prognosis, but also other cells in the surrounding peri-tumoral field for detecting field cancerization.The definition (detection) of field cancerization by cell phenotypical changes has gained recognition over the more conventional definition (detection) by cell genetic or epigenetic alterations, due to the challenge to distinguish cancerized lineages against abundant mutant lineages [14].Also, volumetric imaging is preferred to obtain spatial information, which is absent from genetic analysis, to discern cell niches, vasculature, and layer/duct formations [10].
In the first part (Part I) [15] of this two-part series study, we demonstrated a technique of label-free tetra-modal multiphoton microscopy for slide-free volumetric histochemical imaging (i.e.slide-free virtual histochemistry) compatible with label-free SF-vH&E (Table 1).This technique has no labeling-associated disadvantages [9], and can in principle realize in situ intraoperative imaging [8].Here, in the second part (Part II) of this two-part series study, we demonstrate the diagnostic value of this technique in breast cancer not only for the primary tumor but also for the peri-tumoral field that has not routinely provided imagingbased diagnostic value previously.ping cancer-a otyping of vari lso be applica ering its abili red cancer-as urce, multiphot s study [15]

Detecting field cancerization in the peri-tumoral field
In our breast cancer rat model, we frequently observed macroscopic (~1 cm) multifocal tumors by gross examination.A natural question arises whether grossly invisible microscopic (<1 mm) tumor foci, often termed as lesions, patches, and clusters, occur independently as the result of field cancerization [12][13][14].According to the classical paper by Slaughter and associates [11], field cancerization in breast cancer can be characterized by three unique structures (Fig. 1): (i) emergence of both macroscopic and microscopic tumor foci in a field of precancerous tissue with abnormality of hyperplasia, metaplasia, or dysplasia; (ii) growth of multiple separate (independent) microscopic foci of DCIS and/or invasive breast cancer (IBC) near grossly visible tumor boundary or surgical margin; and (iii) coalescence of multiple (contiguous) tumor foci at the tumor center.
To answer this question, we identified a carcinogen-injected rat with macroscopic multifocal tumors, and performed wide-field (~2 mm 2 with stitched adjacent high-resolution field-of-views collected by scanning the specimen with a mechanical stage) tetra-modal imaging on one of them near a grossly detected tumor boundary (Fig. 4(a)), at the tumor center (Fig. 4(b)), and in the more remote peri-tumoral field (Fig. 4(c)).Noticeably, two separate egg-shaped microscopic (~300 µm) tumor foci emerge in one field of view, with similar tumor cells visible by their cyan-colored nuclei (Fig. 4(a)).Also, similar cyan-colored nuclei are found inside several mammary ducts in the adjacent peri-tumoral field (arrowheads, Fig. 4(a)), revealing the cancer-associated rat mammary epithelium (Fig. 3(c) vs. Fig.3(h)) that appears normal or precancerous (hyperplasia, metaplasia, or dysplasia) by H&E histology (Fig. 4(a), inset) but could have suffered genetic alterations [17].Moreover, the imaging site near the tumor center exhibits the apparent coalescence of two larger (~1 mm) tumor foci (Fig. 4(b)).Thus, all three unique structures of field cancerization are present and imaged in this rat tumor.4 and 3 in F ored field canc 3(f)-3j), consid s (Fig. 3).For another carcinogen-injected rat exhibits a similar tumor microenvironment and peri-tumoral field (Fig. 4(d)), but no cyan-colored field cancerization is present to expand the corresponding phenotype of tumor cells with cyan-colored cytoplasm (box 4, Fig. 4(d)).This phenotype of tumor cells reproduces that of the reported tumor cells in a transgenic mouse breast cancer model [18].Interestingly, all observed tumor foci are surrounded by a similar microenvironment of neo-vasculature, cancer-associated collagen structure, and infiltrating FAD-rich NADH-poor (i.e., yellow-colored) macrophages [18] (bidirectional arrows, Figs.4(a), 4(d)), independent of the breast cancer model selected and the presence of field cancerization.
From the perspective of this colored field cancerization, our rat mammary tumor model recapitulates human breast cancer development, with no interference from xenografting, implantation, immune suppression, genetic modification, or fluorescence protein expression.

Observing signatures of cancer dormancy
In the cyan-colored field cancerization, one H&E-benign (normal or precancerous) but histochemically cancerous rat mammary duct clearly intertwines with surrounding cancerassociated blood vessels at a cooperative juncture (arrow, Fig. 4(a)), allowing possible intravasation of the epithelial cells with cyan-colored nuclei into systemic blood circulation.
Because the corresponding peri-tumoral field approximates a precancerous field before (or after) the emergence (or resection) of the primary tumor, this intravasation may function as the early dissemination of precancerous (dormant tumor) cells [21] during cancer dormancy [22] that dominates overt metastasis (cancer mortality) [23].It is thus important to investigate whether the observed field cancerization possesses the three characteristic signatures of this cancer dormancy [21][22][23]: (i) dormant tumor cells are stem-like cells resistant to therapy that targets primary tumor cells; (ii) early dissemination of the dormant tumor cells requires a hijacked program of branched mammary tubulogenesis; and (iii) therapeutic suppression (natural progression) of tumor proliferation rejuvenates (impedes) this early dissemination.
There is strong evidence that indicates this stem-like origin of the cyan-colored field cancerization.First, the dominant optical phenotype of cyan-colored nuclei corresponds to concentrated unbound nuclear NADH that has been associated with progenitor (stem-like) cells [24].Second, the observed alveoli/ducts in the peri-tumoral field remarkably resemble the mammary gland tubulogenesis by normal stem cells [25] (Figs.3(b)-3(d) vs. Figure 1 in ref. 25).Third, stem-like gene signatures have been associated with the early dissemination of low-burden cancerous tissue that approximates a precancerous/dormant peri-tumoral field [26].This plausible stem-like origin suggests that the dominant optical phenotype (cyancolored nuclei) might originate from one stem cell progenitor, which passed this optical phenotype and the underlying genotype to all its differentiated descendants of epithelial cells, endothelial cells, stromal cells, and tumor cells that share the optical phenotype (monoclonal tumor expansion) [12].The magenta-colored field cancerization observed in the human tumors might originate from stem cell progenitors with a magenta-colored optical phenotype.Further investigations will be performed in the future to systematically establish the link between the optical phenotypes and different cell types.Regardless of the color (dominant optical phenotype) of field cancerization, a common stem-like origin may link it with cancer dormancy.
To critically test this plausible link to cancer dormancy, we recognize the recently observed reversible transition between binary states of tumor cell proliferation vs. dissemination in breast cancer [23].According to this transition, preoperative suppression of the proliferation by chemotherapy rejuvenates the early dissemination, and would enhance (suppress) the magenta-colored field cancerization if such link does (not) exist.We thus performed imaging of an ex vivo specimen from a cancer patient after preoperative neoadjuvant chemotherapy followed by mastectomy.The specimen was dissected from the grossly normal tissue between two multifocal primary tumors (diagnosed by core-needle biopsy) separated by ~5 cm.The two tumors were 1.1-cm and 0.3-cm in size on x-ray mammography, but underwent a pathologic complete response after chemotherapy [27] (no tumor mass was grossly found or histologically identified around the coiled biopsy marker clip).In contrast to this seemingly favorable prognosis by H&E histology, our tetra-modal imaging of the specimen in the peri-tumoral field pointed to a rather different picture (Fig. 6(a)).The chemotherapy seemed to activate branched mammary tubulogenesis [21,23] (Fig. 6(b); Visualization 3) in comparison to its counterpart without preoperative therapy (Visualization 1).This mammary tubulogenesis was intertwined with irregular blood capillaries for plausible dissemination (Fig. 6(b)).One capillary was undergoing active development by assembling intracellular vacuoles of single endothelial cells [28] (arrowhead, Fig. 6(c)), while the connecting larger blood vessel was undergoing active sprouting development (Fig. 6(c); Visualization 4) in comparison to its therapy-free counterpart (Visualization 2).This enhanced picture of magenta-colored field cancerization (Fig. 6) over its therapy-free counterpart (Fig. 5(a)) indirectly validates the above three characteristic signatures, and thus supports the link between field cancerization and cancer dormancy.

Field canceriz
The emergen different from detection, int assessment of sporadic tumo an organs.amentally rly cancer stological ting from nt cancers without primary tumor proliferation may dominate cancer mortality in comparison to proliferative cancers at primary tumor sites.The majority of solid tumors undergo years or decades of latency, which allows for asymptomatic minimal residual disease to ultimately evolve into local recurrences or recurrent metastases.However, proliferation has remained as a hallmark for cancer possibly due to the difficulty in identifying cancer dormancy from circulating blood, bone marrow, and primary tumors.The peri-tumoral field is relevant to field cancerization before the emergence of the primary tumor, and to cancer dormancy after the resection of the primary tumor, and is thus an ideal site to study both phenomena that have been pursued by separate research communities.Unfortunately, the investigation of this field has been largely restricted to genetic analysis without spatial information afforded by H&E histology.We have attempted an extensive investigation of this field by histochemically identifying common cell phenotypes in an organ of interest (breast).This context, along with millimeter-scale wide-field imaging that representatively samples a large surgical specimen from the primary tumor deep into the peri-tumoral field, allows us to detect field cancerization in the peri-tumoral field.Our high-content histochemical imaging of the peritumoral field seems to open up new research frontiers for investigation of local invasion and distant metastasis.Clinically, the magenta-colored field cancerization might extend >2 cm beyond the tumor boundary in regions routinely labeled as "normal" by pathologists, necessitating more detailed study on how it changes geographically at a "margin" and the relevance of this to intraoperative assessment.
In the emerging paradigmatic shift in histology from conventional H&E and IHC to SF-vH&E, our imaging platform fills the critical gap to realize slide-free virtual IHC (SF-vIHC) complementary with SF-vH&E.One highly ideal tandem is to perform SF-vH&E by SRS [4] followed by SF-vIHC with our programmable multiphoton imaging, using one robust fiber laser-based optical source suitable for portable application and clinical translation.This tandem, in comparison to an alternative tandem that conducts light-sheet SF-vH&E microscopy [2] followed by the tissue-clearing-assisted volumetric IHC [10], would gain advantage in in vivo applications without the tissue clearing and unexpected labeling-induced artifacts, at the moderate cost of lowered imaging speed.The optical source for perspective implementation of this tandem will likely favor our fiber laser-induced supercontinuum over the dedicated fiber laser optimized for SRS [4].The flexibility in our supercontinuum source to arbitrarily program the excitation pulses (wavelength, bandwidth, chirp, intensity, etc.) has been demonstrated in the first part (Part I) of this study [15], allowing perspective incorporation of coherent Raman scattering microscopy (including SRS).The resulting slidefree virtual histochemistry will advance rapid diagnosis of small biopsies, real-time intraoperative assessment of the surgical tumor margin with plausible field cancerization, imaging-based stratification of cancer patients at an early-stage for precision medicine, and accurate postoperative prognosis or surveillance of minimal residual disease with possible cancer dormancy.

Fig. 1 .
Fig. 1.Selecti Fig. 2 rat.(a arrow of the with l a prec mamm mamm This illum phenotypical are clearly ne We identified in control spe

Fig. 3
Fig. 3(i)).T xtensive imagin rowheads, Fig the duct in Fig ssue.Interestin having cyanor cells (Fig. 3( Part I of this ape as well as t which highly co tures could be indicate high Endothelial cell lls.
Fig. 3 colore (b) Ra in a m vessel vessel (cross view) visible Huma sectio duct ( vessel endoth Huma epithe Huma Conne blood of-vie

Table 2 .
3. Phenotypical ch ed field cancerizat at epithelial cells i mammary duct wi ls with elongated ls.(f) Rat stroma s-sectional view) l lacking visible ep e endothelial cells an stromal cells in onal view) with ye (lateral view) with l lacking visible helial cells.(p) H an epithelial cells elial cells in a ma an endothelial cell ection of a magen vessel (Visualiza ew.Scale bars: 50 e in peri-tumo ar to matched p served as elong developed bloo wever, in the gated cyan-col d vessels (or c vessels (or con (t))].hanges of rat mam tion, respectively.n a mammary alve ith cyan-colored n cyan-colored nuc al cells in remote lacking visible ep pithelial cells.(i) s. (j) Developed r n remote peri-tum ellow-magenta-col h yellow-magentaendothelial cells.Human primary t s in a tumor regi ammary duct with s of developing bl nta-colored blood c ation 2).Images (b µm.oral fields (arro primary tumor gated cyan-col od vessels wit cancer-associ lored nuclei (o capillaries) [Fi nnect with larg mmary and human (a) Rat primary tu eolus with cyan-co nuclei.(d) Rat end clei.(e) Similar r e peri-tumoral fiel pithelial cells.(h) Developing rat ne rat blood vessel la moral field.(l) Nor lored epithelial ce -colored epithelial (o) Developed h tumor cells with ion with a mage h a magenta-color lood capillaries w capillary (broken b), (c), and (d) are owhead, Fig. 3 r cells (Fig. 3(p lored cells in P th yellow-colo iated rat (or h or a magenta-c ig.3(d) (or Fig ger magenta-co breast cells in cy umor cells with cy olored nuclei.(c) R dothelial cells of rat cells that line ld.(g) Normal ra Normal rat mamm eo-vasculature (bro acking visible end rmal human mam ells.(m) Normal cells.(n) Develop human blood vess a magenta-colore enta-colored cytop red cytoplasm (Vi with a magenta-col box) with a larger e from different de 3(r)) and tumo p)).Part I of this st ored erythrocyt human) blood colored cell cy g. 3(s))] (Tabl olored developi yan-and magentayan-colored nuclei Rat epithelial cells developing blood e developed blood at mammary duct mary duct (lateral oken line) lacking dothelial cells.(k) mmary duct (crosshuman mammary ping human blood sel lacking visible ed cytoplasm.(q) plasm.(r) Human isualization 1).(s) ored cell body.(t) r magenta-colored epths of one field-Classified vital cells in mammary tissue with distinct optical phenotypes.Rat cells specific to cyan-colored field cancerization (2 out of 16 carcinogen-injected rats) Huam cells specific to magenta-colored field cancerization (4 out of 12 breast cancer patients) Fig. 4 rats.I and th the vi (arrow numer colore cance 3(a), 3 Although vH&E structu independent a using the un dominates the Fig. 4(a)), th vasculature ( respectively), appears to be heterogeneity 4. Wide-field tetra-Images (a), (b), an he remote peri-tum cinity of the tumor wheads), endothel rous stromal cells ed nuclei with the rization beyond th 3(b)-3(d), and 3(e) this field canc ural imaging, and perhaps m ique common e constituent ce e epithelia (Fi (Figs.3(d), 3 and the vascu e a highly abn y of these tissu -modal imaging of nd (c) reflect the v moral field of a tu r boundary in anot lial cells lining y s in the peri-tumo primary tumor ce he primary tumor( ), respectively.Sca erization could our tetra-mod ore elegant wa optical phen ells of the tum igs.3(b), 3(c), 3(e), correspon ulature-free stro ormal event (F ue components f cyan-colored fiel vicinity of mamm mor in one rat, re ther rat for direct c yellow-colored bl oral field share th ells (broken box 1 (s).Dashed boxes ale bars: 100 µm.d be alternative dal histochemic ay to detect the otype of cyan mor foci (Fig. 3 , correspondin nding to brok oma.The resu Figs.3(a)-3(e) with widely v n carcinogen-inject dary, tumor center, image (d) reflects (a).Epithelial cells uble arrows), and henotype of cyancyan-colored field so plotted in Figs y H&E histolog ffers another r zation (Figs. 4 lei that simult ding to broken box 2 in Fig. Fig. 5 patien tumor (arrow pheno box 3 Dashe Scale By selecti analogous to diagnosis, we patients.This stratify breast largely shares blood vessels colored const different optic to molecular d is an importa 5. Wide-field tetra nt (Subject #388).r center, and the w) and endothelia otype of magenta-c 3), illuminating a ed boxes 1, 2, 3, bars: 100 µm.ive wide-field breast cancer e observed the s field canceriz t cancer patien s with the contr s (Figs.3(n), 3 tituent epitheli cal field charac differences bet ant finding th Fig. 6 colore cance root-li modal showi myoep magen volum (arrow (d) m differe