We report a novel approach that generates a high resolution, three-dimensional (3D) fluorescent staining atlas of tissue microbiopsies in a microfluidic device without destroying the tissue. We demonstrate that this method preserves tissue architecture for multiple murine organs by comparing traditional 2D slices to an optically sectioned 3D H&E-mimic. The H&E-mimic slices show a close qualitative match to traditional H&E. The 3D spatial and molecular information obtainable from this method significantly increases the amount of data available for evaluating both tissue morphology and specific biomarkers in a wide range of both research and clinically driven applications and is amenable to automation.

Analysis of hematoxylin and eosin (H&E) stains is the cornerstone method for the diagnosis of cancer and other related tissue pathologies. H&E has been used for more than a century with few modifications.1 A main limitation of traditional H&E is that the requisite slicing step is destructive, limiting the information that can be extracted from valuable patient samples, especially for tissue biopsies that are often difficult, painful, or dangerous to obtain in large sizes (e.g., kidney, brain, or lung).2 Few studies have examined diagnostic error rates due to sample sizes available to clinicians; however, there are indications that relying solely on 2D histological information can lead to either an over-, under-, or mis-diagnosis of the patient's prognosis, as suggested by recent studies.3,4 Increasingly, there are attempts to develop characterization tools that reduce the required tissue volume by maximizing the utility of microbiopsies, either for tissue characterization or molecular profiling.5,6 While other 3D imaging modalities such as magnetic resonance imaging, computed tomography, and 3D ultrasound have enhanced patient diagnoses compared to their 2D imaging counterparts, they lack cellular resolution. Recent advances in the field of optical tissue clearing, in which lipids are removed from large tissue samples to render them optically transparent, in combination with modern 3D microscopy and microfluidics enable the possibility to extend traditional histology to 3D.7,8 Similarly, developing 3D histology approaches for microbiopsies9 with the potential to provide additional spatial information with molecular resolution for diagnostic biomarkers throughout the tissue microenvironment can combine the molecular profiling and traditional histology in the same sample.

Here, we describe a method that non-destructively processes multiple tissue samples in a microfluidic chip by labeling with fluorescent analogs of traditional H&E as well as other fluorescent markers. The 3D tissue architecture is kept intact throughout the process, thereby avoiding the loss of valuable small-volume tissue samples due to mechanical sectioning in conventional H&E. Additionally, the option to quench and restain tissue samples8 allows pathologists to perform as many tests as required to assure an accurate diagnosis. This circumvents a common problem with current histology methods where tissue sections have been exhausted, but additional analysis is desired. Further, automated imaging with fluorescence microscopy means that the full extent of the tissue is available in a convenient digital format for easy review. Although methods exist to create 3D reconstructions of whole tissue from stained 2D thin sections at lower spatial resolutions, these techniques still require laborious preparation of individual sections.10 Such methods also destroy sample integrity, precluding subsequent analysis. This novel approach has the potential to provide detailed 3D microenvironment information to the histologist, is amenable to automation, and is fully extensible to high-level molecular profiling in 3D. The use of simple microfluidic cartridges provides multiple benefits: protection of fragile tissue samples that are not embedded in paraffin, minimization of expensive reagent volumes, and integration into low-footprint automated platforms that interface directly with imaging systems. We demonstrate this approach by generating 3D H&E-mimic images, and comparing these to traditionally processed H&E sections from the same tissues. The proof-of-principle process described here can easily be extended to multiplexed labeling of virtually any tissue type.

Whole brain, kidney, and lung tissues were collected from an adult virgin female BALB/c Rag1-/- mouse. The animal was cardiacally perfused with 20 ml cold 1× phosphate-buffered saline (PBS) and followed with 20 ml cold 4% paraformaldehyde (Sigma Aldrich). Individual organs were further fixed in 10% neutral buffered formalin (Azer Scientific) for 2 h. Half of each organ was used for the proof-of-principle 3D H&E-mimic labeling and the other half reserved for traditional H&E comparison staining. 3D H&E-mimic tissue biopsy dimensions were 1 mm in diameter and 140 μm in thickness, similar in dimension to reported microbiopsy punches.11 Biopsies were cleared of lipids for 24 h in 1.5 ml microcentrifuge tubes in CUBIC-1 solution on a rotating platform followed by multiple washes in PBS.7 

Each microfluidic chip consisted of a 2 × 2 array of cylindrical sample chambers with individual fluid inlet and effluent lines and was made of polydimethylsiloxane (PDMS) using soft lithography12 (Fig. 1(a)). Both the chips and 24 × 60 mm glass coverslips (Ted Pella) were treated in parallel with corona discharge (Enercon Dyne-A-Mite HP) for 90 s. Immediately after corona discharge treatment, each sample chamber was loaded with a different cleared tissue biopsy. The treated surfaces of the chip and coverslip were immediately brought into contact to initiate a bond to seal the chip and then baked in an oven at 90 °C for 15 min to complete the bond. Subsequently, the chip chambers were flooded with PBS and stored in a humidified petri dish.

FIG. 1.

3D histology platform. (a) Microfluidic chip design showing an expanded view of a single sample chamber and a representative image of a physical chip with the channels labeled green with food coloring. A US dime is shown for size comparison. (b) and (c) Normal mammary gland tissue injected with MCF10-DCIS.com tumor cells. Duct was injected with trypan blue before dissection. Duct was optically cleared and stained on-chip with a combination of 5 μg/ml DAPI (blue), 100 μM Fluorescein Phalloidin (Sigma Aldrich) (green), and anti-Cytokeratin 8 (Abcam) (red) conjugated with Zenon Alexa Fluor 647 rabbit IgG labeling reagent (Life Technologies) per manufacturer directions. Scale bar is 100 μm. (c) Orthogonal slice showing the hollow lumen 3D structure (white rectangle in (b)). A clump of tumor cells is readily identifiable by optical sectioning. Scale bar is 50 μm.

FIG. 1.

3D histology platform. (a) Microfluidic chip design showing an expanded view of a single sample chamber and a representative image of a physical chip with the channels labeled green with food coloring. A US dime is shown for size comparison. (b) and (c) Normal mammary gland tissue injected with MCF10-DCIS.com tumor cells. Duct was injected with trypan blue before dissection. Duct was optically cleared and stained on-chip with a combination of 5 μg/ml DAPI (blue), 100 μM Fluorescein Phalloidin (Sigma Aldrich) (green), and anti-Cytokeratin 8 (Abcam) (red) conjugated with Zenon Alexa Fluor 647 rabbit IgG labeling reagent (Life Technologies) per manufacturer directions. Scale bar is 100 μm. (c) Orthogonal slice showing the hollow lumen 3D structure (white rectangle in (b)). A clump of tumor cells is readily identifiable by optical sectioning. Scale bar is 50 μm.

Close modal

In H&E, hematoxylin labels nucleic acids and basophilic compounds predominantly found in the nuclei. Eosin labels eosinophilic proteins, predominantly staining the cytoplasm.1 To provide a mimic of the morphological information in an H&E stain, we used a fluorescent staining solution that consists of 5 μg/ml 4′,6-diamidino-2-phenylindole (DAPI) to provide a fluorescent labeling of nucleic acids and 4 mg/ml Sulforhodamine B (SRB) in 1% glacial acetic acid to label proteins (Sigma Aldrich).13 The staining solution was perfused through the chip using a syringe pump (Harvard Apparatus) at a flow rate of 10 μl/h for 24 h. After staining, samples were washed with PBS for 1 h at 20 μl/h before being imaged on a Nikon Eclipse Ti spinning-disc confocal microscope or Olympus FV1000 2-photon confocal microscope. We developed a custom Matlab script (available upon request) that maps H&E-mimic color information drawn from true H&E reference images onto the optically sliced data to generate H&E-mimic slices with equivalent information. In this way, the confocal z-stack is convertible to a series of optically sectioned, H&E-mimic images of the entire tissue thickness that can serve as a morphological template for overlays of multiplexed fluorescent marker data. Unlike previous methods employed for creating H&E-mimic images from fluorescence data,14 this method also allows the H&E-mimic images to be matched to the specific range of colors expected for different tissue types.

Tissues specimens for traditional H&E comparison staining were transferred to 70% ethanol overnight prior to tissue processing for H&E. Tissue samples were processed in a Leica TP1020 automatic processor following the procedures in the instruction manual (Leica). Following processing, samples were paraffin-embedded and cut into 5–10 μm sections (Leica RM2125). Slides were placed on a 65 °C heat block for 2 h prior to H&E staining. Staining was performed using Gill's Hematoxylin (VWR) and Eosin (Sigma) based on a standard protocol.15 After staining, slides were mounted with Permount (VWR) and imaged at 20× magnification on a Leica microscope.

Representative optical slices of the 3D H&E-mimic are compared to traditional 2D H&E physical slices in Figs. 2(a)–2(c) for kidney, lung, and brain samples. The 3D H&E-mimic optical slices show a close qualitative match to traditional H&E (Figs. 2(e)–2(g)). Further, the entire optical stack is available to the histologist, enabling a much more in-depth analysis of the tissue microenvironment and cell-cell connectivity (Fig. 1(c)) by providing the third dimension, which is unavoidably lost by physical slicing (Figs. 2(d)–2(i)). Fig. 2(d) shows an orthogonal view through the entire tissue thickness for kidney, demonstrating stain penetration through the full thickness of the stack. Figs. 2(h) and 2(i) show two optical slices 50 μm apart in the stack. Calculation of normalized mutual information (NMI), a metric used to quantify image similarity that is equal to 0 with no image commonality and equal to 1 for identical data sets, between these two slices demonstrates that even over small tissue distances (<100 μm), dramatic morphological changes occur.16 

FIG. 2.

Comparison between the traditional H&E and 3D H&E-mimic. (a)–(c) Traditional H&E reference images from mouse kidney, brain, and lung tissues, respectively. Images were cropped to represent the same physical area as achieved on the confocal imaging setup. (d) Orthogonal view from a stained kidney sample imaged on a 2-photon microscope showing stain penetration through the full thickness of the stack. (e)–(g) Optical sections from the 3D H&E-mimic for the mouse kidney, brain, and lung tissues, respectively, transformed using the custom matlab script and an H&E color reference image for the same tissue. (h) and (i) Show two representative slices, denoted by the blue and yellow lines in (d), respectively, from a distance of 50 μm apart. Calculation of the NMI between the two images demonstrates the differences in morphology that can occur even at short depth scales and highlights the importance in preserving the full 3D architecture of the tissue sample for multiplex testing. All scale bars are 50 μm.

FIG. 2.

Comparison between the traditional H&E and 3D H&E-mimic. (a)–(c) Traditional H&E reference images from mouse kidney, brain, and lung tissues, respectively. Images were cropped to represent the same physical area as achieved on the confocal imaging setup. (d) Orthogonal view from a stained kidney sample imaged on a 2-photon microscope showing stain penetration through the full thickness of the stack. (e)–(g) Optical sections from the 3D H&E-mimic for the mouse kidney, brain, and lung tissues, respectively, transformed using the custom matlab script and an H&E color reference image for the same tissue. (h) and (i) Show two representative slices, denoted by the blue and yellow lines in (d), respectively, from a distance of 50 μm apart. Calculation of the NMI between the two images demonstrates the differences in morphology that can occur even at short depth scales and highlights the importance in preserving the full 3D architecture of the tissue sample for multiplex testing. All scale bars are 50 μm.

Close modal

This proof-of-principle study makes a clear demonstration of the potential nondestructive 3D immunostaining holds for providing a complete picture of microbiopsies. The device and method are also completely extensible toward next-generation immunohistochemical characterization of intact tissues and can be combined with recent advances to enhance transport of biomarkers into larger tissue pieces for enhanced staining kinetics.8,17,18 By not destroying the sample, a pathologist can archive and retrieve digital data for later review without precluding follow-up assays on the same sample. Importantly, ease of access to serial tissue sections, which is missing when just a few slices are selected from a sample, could improve interpretation and aid in challenging diagnoses. Combining the morphological information provided by the 3D H&E-mimics described here with subsequent rounds of immunohistochemical analysis will enable the co-registration of morphological data with high numbers of fluorescent antibody stainings8,19 in the same tissue sample in an easily interpretable format.

This work was supported in part by National Science Foundation Grant No. CBET 1403887 (C.N., J.Z., D.H., and P.B.) and NIH Grant Nos. UO1 HL116330 (P.B. and T.B.R.) and 1R01CA194697-01 (S.Z.), Notre Dame Advanced Diagnostics & Therapeutics (AD&T) seed grant and AD&T Berry Fellowship (C.N.), and the Harper Cancer Research Institute's Research Like a Champion award (C.N., K.C., J.Z., and S.Z.), and CTSI Core Facility Pilot Fund (S.Z.). Imaging was supported by the Notre Dame Integrated Imaging Facility and IUSM-SB Imaging Core Facility. The authors also thank Thomas Storey and Melinda Lake for technical assistance. Additionally, Laura Tarwater and the Notre Dame tissue bank for input on the H&E stain procedures. The authors thank Miranda Burnette, Dr. Bobbie Sutton, MD, and the Cleveland clinic for helpful feedback.

1.
A. H.
Fischer
,
K. A.
Jacobson
,
J.
Rose
, and
R.
Zeller
,
CSH Protoc.
pdb.prot4986
(
2008
).
2.
R.
Roth
,
S.
Parikh
,
D.
Makey
,
J.
Foster
,
G.
Rozenblit
,
A.
Satoskar
,
G.
Nadasdy
,
J.
Von Visger
,
L.
Hebert
,
B. H.
Rovin
 et al,
Am. J. Nephrol.
37
,
249
(
2013
).
3.
J. G.
Elmore
,
G. M.
Longton
,
P. A.
Carney
,
B. M.
Geller
,
T.
Onega
,
A. N. A.
Tosteson
,
H. D.
Nelson
,
M. S.
Pepe
,
K. H.
Allison
,
S. J.
Schnitt
,
F. P.
O'Malley
, and
D. L.
Weaver
,
J. Am. Med. Assoc.
313
,
1122
(
2015
).
4.
A.
Delides
,
M.
Maniadakis
, and
E.
Stathopoulos
,
Anticancer Res.
33
,
3213
(
2013
).
5.
T. W.
Prow
,
L. L.
Lin
, and
H. P.
Soyer
,
Future Oncol.
9
,
1241
(
2013
).
6.
J. R.
Townsend
,
J. R.
Hoffman
,
M. S.
Fragala
,
L. P.
Oliveira
,
A. R.
Jajtner
,
D. H.
Fukuda
, and
J. R.
Stout
,
Med. Sci. Sports Exercise
48
,
331
(
2015
).
7.
K.
Tainaka
,
S. I.
Kubota
,
T. Q.
Suyama
,
E. A.
Susaki
,
D.
Perrin
,
M.
Ukai-Tadenuma
,
H.
Ukai
, and
H. R.
Ueda
,
Cell
159
,
911
(
2014
).
8.
E.
Murray
,
J. H.
Cho
,
D.
Goodwin
,
T.
Ku
,
J.
Swaney
,
S.-Y.
Kim
,
H.
Choi
,
Y.-G.
Park
,
J.-Y.
Park
,
A.
Hubbert
 et al,
Cell
163
,
1500
(
2015
).
9.
L.
Lin
,
J.
Tan
,
H.
Soyer
, and
T.
Prow
,
J. Invest. Dermatol.
135
,
Supplement 2
,
S18
(
2015
).
10.
I.
Arganda-Carreras
,
R.
Fernández-González
,
A.
Muñoz-Barrutia
, and
C.
Ortiz-De-Solorzano
,
Microsc. Res. Tech.
73
,
1019
(
2010
).
11.
L. L.
Lin
,
T. W.
Prow
,
A. P.
Raphael
,
R. L.
Harrold
 III
,
C. A.
Primiero
,
A. B.
Ansaldo
, and
H. P.
Soyer
,
F1000Research
2
(
2013
).
12.
Y.
Xia
and
G. M.
Whitesides
,
Angew. Chem. Int. Ed.
37
,
550
(
1998
).
13.
V.
Vichai
and
K.
Kirtikara
,
Nat. Protoc.
1
,
1112
(
2006
).
14.
L.
Buchynska
,
E.
Kashuba
, and
L.
Szekely
,
Exp. Oncol.
30
,
327
(
2008
).
15.
K. L.
Schwertfeger
,
W.
Xian
,
A. M.
Kaplan
,
S. H.
Burnett
,
D. A.
Cohen
, and
J. M.
Rosen
,
Cancer Res.
66
,
5676
(
2006
).
16.
A. F.
McDaid
,
D.
Greene
, and
N.
Hurley
, e-print arXiv:1110.2515.
17.
S.-Y.
Kim
,
J. H.
Cho
,
E.
Murray
,
N.
Bakh
,
H.
Choi
,
K.
Ohn
,
L.
Ruelas
,
A.
Hubbert
,
M.
McCue
,
S. L.
Vassallo
,
P. J.
Keller
, and
K.
Chung
,
Proc. Natl. Acad. Sci. U. S. A.
112
,
E6274
(
2015
).
18.
J.
Li
,
D. M.
Czajkowsky
,
X.
Li
, and
Z.
Shao
,
Sci. Rep.
5
,
10640
(
2015
).
19.
C.
Van Poznak
,
M. R.
Somerfield
,
R. C.
Bast
,
M.
Cristofanilli
,
M. P.
Goetz
,
A. M.
Gonzalez-Angulo
,
D. G.
Hicks
,
E. G.
Hill
,
M. C.
Liu
,
W.
Lucas
,
I. A.
Mayer
,
R. G.
Mennel
,
W. F.
Symmans
,
D. F.
Hayes
, and
L. N.
Harris
,
J. Clin. Oncol.
33
,
2695
(
2015
).