Tissue fixation and the effect of molecular fixatives on downstream staining procedures.

It is impossible to underplay the importance of fixation in histopathology. Whether the scientist is interested in the extraction of information on lipids, proteins, RNA or DNA, fixation is critical to this extraction. This review aims to give a brief overview of the current "state of play" in fixation and focus on the effect fixation, and particularly the effect of the newer brand of "molecular fixatives" have on morphology, histochemistry, immunohistochemistry and RNA/DNA analysis. A methodology incorporating the creation of a fixation tissue microarray for the study of the effect of fixation on histochemistry is detailed.

Introduction

Fixation is the foundation step behind the study of pathology and essentially exists to prevent the autolysis and degradation of the tissue and tissue components such that they can be observed both anatomically and microscopically following sectioning. A number of fixatives exist, either having being in use for decades, or in the case of formaldehyde over a century, whilst others have only been created in the last 10 years. To attempt to classify this chaos, fixatives can be placed into two categories; denaturing fixatives and cross-linking (or addition) fixatives. Table 1 details some of the standard histological fixatives used and their classification.

Table 1

Details of the breakdown of the different fixatives.

Fixative	Method of fixation	Contents
B5	Denaturing	5.4% Mercuric Chloride (w/v), 1.1% Sodium Acetate (w/v), 4% Formaldehyde (v/v), Water
Bouin’s	Denaturing, cross-linking	25% of 37% formaldehyde solution, 70% picric acid, 5% acetic acid
Carnoy’s	Denaturing	60% ethanol, 30% chloroform, 10% Glacial acetic acid
Glutaraldehyde	Cross-linking	Generally, 2% v/v of glutaraldehyde to water/PBS
Methacarn	Denaturing	60% methanol, 30% chloroform, 10% Glacial acetic acid
Neutral buffered formalin (NBF)	Cross-linking	10% of 37% formaldehyde solution, in a neutral pH
Paraformaldehyde (PFA)	Cross-linking	Generally, 4% w/v of paraformaldehyde to Water/PBS
Zenker’s	Denaturing	5% Mercuric Chloride (w/v), 2.5% Potassium Dichromate (w/v), 5% Glacial acetic acid (v/v), Water

The alcohol-based fixatives, for example Carnoy’s and Methacarn, are denaturing fixatives. The action of the alcohol present in the solution acts to cause protein denaturation through the removal of water from the free carboxyl, hydroxyl, amino, amido and imino groups of the proteins [1] which results in protein coagulation and tissue shrinkage. Carnoy’s fixative adds chloroform and acetic acid to the mixture which counteracts the shrinkage effects of ethanol and engenders tissue fixation through hydrogen bonding of the constituents to the tissue [2]. Similarly, Methacarn, where ethanol in the Carnoy’s solution is replaced by methanol, appears to work by the same method [3].

The mercuric-containing fixatives, for example B-5 and Zenker’s, are little-used in current practice and are thought to act through binding to sulphydryl and amino groups in an additive reaction [1,4].

Bouin’s, like Carnoy’s, was first described in the late 19th Century by Pol Andre Bouin. Consisting of picric acid, acetic acid and formaldehyde, it has both a coagulative as well as cross-linking effect on proteins. In particular, the penetration of picric acid into the tissue is slow and it coagulates proteins but has no known chemical interaction with them. Whereas, acetic acid penetrates relatively quickly and opposes the tissue shrinkage caused by the picric acid. With formaldehyde present in a higher concentration than for formaldehyde alone, approximately 10% formaldehyde, the actions of formaldehyde over time cross-link the tissue, although this may be inhibited by the low pH of the solution, pH 1.3–1.6 [1].

Undoubtedly the most investigated fixative for its mechanism of action is formaldehyde. First discovered in 1859, its use in pathological applications was characterised by the work of Ferdinand Blum (for review see [5]). Formaldehyde is a small molecule (MW = 30) existing as a gas, which is commonly in the form of a 37% formaldehyde solution created by bubbling formaldehyde gas through water until saturation point. Its most common form in histological laboratories is as a 10% solution, thus ∼4% formaldehyde, either diluted in water (originally termed formalin) or in a buffered solution (termed neutral buffered formalin or NBF). However, while the principle component of formaldehyde solution is formaldehyde, the researcher should be aware that oxidisation of formaldehyde will produce an unknown amount of formic acid, hence the reason why unbuffered formalin is acidic, and that 10–15% methanol may be present in the solution and act as a stabilising agent.

The active ingredient in any formaldehyde solution is methylene glycol, the hydrated form of formaldehyde and the two chemicals co-exist within the solution in an equilibrium favouring methylene glycol. It has been proposed that the known paradox between the rate of penetration of formaldehyde and its rate of fixation may be due to the fast penetration speed of methylene glycol and the slow fixation rate of formaldehyde [5].

The mechanism of action of formaldehyde has been researched extensively and occurs through the formation of intra and inter-molecular cross-links. The principal cross-links occur between side chain amino group of lysine which over time results in the formation of methylene bridges [6]. However, cross-linking can also occur between the aminomethylol groups and phenol, indole and imidazole side chains by a form of the Mannich reaction [7]. Consequently, the variety of amino acids affected by formaldehyde includes lysine, arginine, tyrosine, asparagine, histidine, glutamine and serine [8]. Through work by Rait et al. [9] and Mason & O’Leary [10], who have modelled formaldehyde fixation using ribonuclease A, it has been determine that the secondary and tertiary structure of proteins are unaffected, and thus preserved, by formaldehyde fixation. Moreover, that structure can be revealed, and the activity of the enzyme recovered, by heating [9]. Following this work, Sompuram et al. [11] have sought to classify proteins, based on their staining using immunohistochemistry following fixation of a small peptide with formaldehyde vapour. Using this model, peptides fall into three groups, based on (1) the presence of tyrosine at the antibody-binding site and an arginine elsewhere, (2) a tyrosine, but no arginine and (3) no tyrosine present.

A final consideration applying to all fixatives is the rate of penetration, temperature and length of time in fixative, which are all interlinked and following from this, the tissue processing method. Medawar [12] initially proposed the formula , where d is depth, K is the coefficient of diffusion and t is time. As the coefficient of diffusion is different for each fixative, different fixatives and chemical mixtures will have different properties, as demonstrated by Dempster [13] where acetic acid shows the quickest penetration followed by formaldehyde. As noted above, this does not necessarily reflect the rate of fixation which will proceed at a slower rate and may reflect the often seen observation of differential staining on the outside of a large specimen, compared to the middle. A general rule of thumb to apply for penetration is 1 mm/h and a fixation time of 24 h is recommended for NBF-fixed specimens.

The temperature at which the sample in fixative is stored is relevant since basic chemical principles dictate that the speed of any reaction can be increased with heat, and consequently slowed when chilled. This can be demonstrated experimentally, as Fox et al. [5] showed that tissue sections fixed in formaldehyde reached equilibrium in 24 h at 25 °C and <18 h at 37 °C while Sompuram demonstrated loss of antigenicity, as a model for fixation, in peptides fixed at 42 °C compared to room temperature [11]. However, as fixation in a clinical environment is often through the processing unit being used, the method of processing is also important. In support of heating as a method for speeding tissue penetration and fixation, the introduction of automated microwave tissue processing has shown that using microwaves can shorten processing times without any effect on morphology or other downstream processes [14,15] presumably through heating. However, heating may not necessarily be required, as ultrasound methodologies have also been demonstrated to shorten fixation times, but heating is not implicated [16].

Molecular fixatives

It has been known for many years that formaldehyde fixation represents not only a biohazard for the laboratory [17] but also limits the quality of RNA & DNA available for extraction from the block. With the increasing use of molecular testing in the clinical arena, this limitation could be a severe restriction to the use of these tests. Thus, a number of “molecular fixatives” have been created and applied to histology with the aim to replace formaldehyde for experimental or health reasons.

The majority of molecular fixatives use alcohol or acetone as a base solution, with the addition of other stabilising agents to overcome the well-documented shrinkage effects of alcohol in a straight-forward replacement of formalin in the tissue processing procedure. However, in some cases, the procedure requires modification, for example, the HOPE technique, first introduced in 2001, which consists of Hepes-glutamic acid-buffer mediated Organic solvent Protection Effect combined with an acetone fixation step [18]. It is claimed that the amino acid components of the HOPE solution protect the tissue from the deleterious effects of the acetone. In other cases, specialised equipment is recommended to maximise the efficiency of the fixative. For example, UMFIX (Universal Molecular FIXative) first published in 2003 [19], which has been used extensively in the Department of Pathology, University of Miami. It is methanol-based with the addition of polyethylene glycol and is most effective in combination with a rapid tissue processing system [20]. Finally, not all solutions include conventional denaturing liquids. Z7 [21], and similar Zinc-based buffers such as ZBF [22,23], are a mixture of Zinc salts, thus are inexpensive to make, non-toxic and should slot into established processing protocols.

Effect of molecular fixatives on downstream processes

Morphology

The preservation of morphology is a central tenet of histopathology and therefore it follows that the manufacturers of the commercially-prepared molecular fixatives ensure that the morphological criteria of the staining is similar to formaldehyde fixation. However, some subtle differences do occur, which are principally associated with the ethanolic base of the fixatives used [19,24,25]. It has been reported that Paxgene-fixed tissue [26] increases eosinophilia, but not sufficiently to limit diagnosis. Shrinkage of tissues and swelling or lysing of erythrocytes has been extensively reported for molecular fixatives, occurring for FineFix, F-Solv, RCL2 [27,28] and Paxgene [26] and in addition, there has been a report of mast cell degranulation in Finefix tissue [29]. However, overall, the morphology of tissue following use of molecular fixatives is similar to formaldehyde-fixed tissue and suitable for most histological purposes.

Effects on RNA/DNA

Fixation in formaldehyde has profound effects on the extraction of RNA & DNA from tissues. Not only does cross-linking occur with the surrounding histones, but formaldehyde can react directly with nucleotides [30,31]. As part, or as a consequence, of this process, formaldehyde has not only been reported to block retrieval of nucleic acids [32], but also degrade nuclear material [33] and alter sequences [34]. For this reason, most researchers use fresh-frozen material as the basis of their molecular studies.

However, it has been known for many years that DNA collapses in ethanolic fixatives [3] and therefore both Carnoy’s and Methacarn are suitable for the preservation of nucleic acids [35] which thus prompts the advent of the molecular fixatives listed in Table 2.

Table 2

Molecular fixatives and publications.

Fixative	Publications
FineFix	[51,52,25]
Glyo-fix	[25]
Histochoice	[53]
HOPE	[51,54,55]
Neo-Fix	[25]
PAXgene	[24,41,56,26]
RCL2	[42,51,56–58]
Streck’s	[43]
UMFIX	[44,40,19,20]
Z7	[21,56]
ZBF	[25,23]

In separate publications, all of the molecular-friendly fixatives have been demonstrated to either preserve RNA or DNA through the measurement of 18S and 28S RNA, or through Ct values for qPCR and can be seen to be more similar to fresh-frozen material than to their formalin-fixed equivalent. Similarly, longer read lengths for DNA or cDNA products can also be demonstrated (Table 3), where it is assumed that formalin-fixation will average between 200 bp for PCR amplification [36]. However, it is also clear that the relative preservation of RNA & DNA is affected by factors other than fixation, including warm ischemia time [37], temperature of fixative [38] and methods of isolation [20]. Indeed, on this last point, it has been shown by Turashvili and colleagues that while use of a molecular fixative in tandem with rapid tissue processing can provide the reproducible amplification of DNA & RNA to read lengths of 1.4 kb and 816 bp, respectively, the use of a good in-house or kit-based extraction protocol can gain read lengths of 988 bp for DNA in 6/9 samples and 816 bp for RNA in 7/9 samples. Therefore, whilst molecular fixatives do increase the preservation of RNA & DNA in tissue preparations, the entire chain of fixation and storage needs to be considered when processing tissue through for RNA/DNA, a point made by Womack & Mager in this edition.

Table 3

RNA read lengths following tissue fixation for RTPCR.

Fixative	Read lengths DNA	Read lengths RNA
HOPE	600 bp [54]	300 bp [54]
Methacarn	1900 bp [35]	850 bp [35], 463 bp [57]
PAXgene	571 bp [41]	712 bp [41]
RCL2	600 bp [58], 850 bp [57]	377 bp [58], 463 bp [57]
UMFIX/RTP	1.4 kb [20], 450 bp [19]	816 bp [20], 450 bp [19], 700 bp [59]
Z7	2400 bp[21]	361 bp [21]

Effects on proteins and immunohistochemistry

As with RNA/DNA preservation, formaldehyde fixation has a deleterious effect on the resolution of proteins in formalin-fixed tissues. The use of formalin-fixed tissue in western blots demonstrates the extent of the issue where only 4 out of 23 proteins examined by Vincek et al. could be detected after 24 h of formaldehyde exposure [19]. However, as immunohistochemistry is the routine method of protein analysis, this overemphasises the problem and suggests an immunohistochemical model where heat induced epitope retrieval (HIER) did not exist. As detailed elsewhere in this Methods edition (Warford et al.), the advent of heat induced epitope retrieval (HIER) [39] has revolutionised immunohistochemistry, both in the clinical and research setting, and yet there are still protein epitopes that are resistant to unmasking and there is still a need to provide detailed advice on the proper validation of antibodies for immunohistochemistry (Howat et al., 2014). Thus the perfect fixative, that allows morphology, RNA & DNA preservation and protein visualisation through immunohistochemistry will continue to be sought.

Molecular fixatives have been studied in this regard not because they are believed to solve the inherent limitations of a cross-linking fixative, but to try to provide that Holy Grail of perfect fixative that can be used across all pathology laboratories. Some, for example UMFIX in combination with rapid tissue processing, have been adopted in some centres and have been reported to increase throughput when adopted [40], but most have not been widely adopted, partly because of the expense of transferring to a more costly fixation regime and purchase of equipment and partly because of the inevitable re-validation of antibody panels in the clinical setting [24,41].

In examination of the literature it can be difficult to extract the details of the performance of various fixatives in comparison to others due to the variability in staining procedures used and the types of antibodies. Paavilainen et al. [25] provide the most comprehensive dataset using 72 antibodies, both commercial monoclonals and in-house prepared monospecific antibodies, across eight tissue types fixed in seven separate fixatives. Analysis of this data is, by necessity, complicated and generalisations are used. However, in general, none of the fixatives outperformed NBF in immunohistochemistry with only a few exceptions. Similarly, van Essen et al. using 85 in vitro diagnostic clinical antibodies demonstrated that fixation in NBF provided good staining in 84% of antibodies, with RCL2 only providing good staining in 66%. Thus the number of antibodies where RCL2 outperformed NBF was low (10/85) compared to those that were the same or worse than NBF (51/85 and 24/85, respectively) [42]. Whilst most groups use clinically validated antibodies for testing, such as MIB-1 for Ki67, the relevance of research phospho-specific antibodies has been evaluated by only one group who determined that Streck’s Tissue Fixative gave an increase in intensity in an impressive 60% of antibodies tested [43]. However, in most other cases examined, the proportion of antibodies that are determined to be better in the alternative fixative are low. One notable exception being UMFIX where Nadji et al. show that 33% (23/70) of the clinical antibodies used performed better using UMFIX than the standard NBF protocol [44]. This may be a reflection of the experience of this group with this fixative and the use of a microwave-assisted tissue processing protocol designed to complement the fixative. Thus the process is optimised for the alternative fixative rather than for NBF. Indeed, it has been shown that optimisation can improve the staining with alternative fixatives, as demonstrated by Belloni et al. [41] where NP40 was used to increase nuclear permeability and improve the staining where PAXgene processed tissues were used.

Table 4 details an evaluation of the published literature with regard to both standard and molecular fixatives where they have been employed in a direct comparison to formaldehyde. Each manuscript has been examined and the results extracted from the tables and text provided and assessed for the type of antibody used, whether the staining was optimised for the alternative fixative and the type of scoring utilised by the authors. While the older literature, prior to HIER [45], should be treated with some caution, as there is some doubt whether the results would stand had HIER been used at that time, it is clear that molecular fixatives only increase the chances of detection of protein by immunohistochemistry in a small proportion of cases. Indeed, as many antibodies are found to be as good in formaldehyde-fixed tissue as in alternative molecular fixatives. It could be argued that an improvement in the staining with alternative fixatives is not the desired result, but that maintenance of the status quo is required if alternative fixatives are to be adopted. However, given the prevalence of formaldehyde in clinical labs and the projected costs and change to workflows and protocols that may be required to adopt an alternative, it is likely that formaldehyde will continue to be the method of choice for immunohistochemistry for some time to come.

Table 4

Comparison of number of antibodies fixed in alternative fixative compared to formalin-fixed paraffin embedded tissue.

Fixative	Publication	Better in alternative fixative	As good as NBF	Better in NBF	Antibody typea	Optimised for alternative?	Scoringb
AFA	Nietner et al. [24]	1/3		2/3	C	Y	SQ
Bouin’s	Mitchell et al. [45]	3/4	1/4		C	N	SQ
Carnoy’s	Mitchell et al. [45]	3/4	1/4		C	N	SQ
Ethanol	Gillespie et al.[59]		1/1		C	N	V
FineFix	Kothmaier et al. [51]	2/5	3/5		C/R	N	SQ
	Nykanen et al. [52]		2/2		C	N	Q
	Paavilainen et al. [25]			72/72	C/R	N	SQ
Glyo-fix	Paavilainen et al. [25]			72/72	C/R	N	SQ
Histochoice	Vince et al. [53]	4/21	15/21	2/21	C	N	Q
HOPE	Kothmaier et al. [51]	3/5	2/5		C/R	N	SQ
	Braun et al. [54]		3/3		C/R	Y	V
	Goldmann et al. [55]		4/4		C	Y	V
Methacarn	Mitchell et al. [45]	3/4	1/4		C	N	SQ
	Delfour et al. [57]		10/10		C	N	SQ
Neo-Fix	Paavilainen et al. [25]			72/72	C/R	N	SQ
PFA (4%)	Burns et al. [43]		8/10	2/10	R(ph)	N	SQ
PAXgene	Nietner et al. [24]	1/3		2/3	C	Y	SQ
	Belloni et al. [41]		5/10	5/10	C/R	Y	Q
	Staff et al. [56]		7/7		C	N	V
	Kap et al. [26]		33/33		C	Y	V
RCL2	Van Essen et al. [42]	10/85	51/85	24/85	C	N	SQ
	Kothmaier et al. [51]	1/5	4/5		C/R	N	SQ
	Staff et al. [56]		7/7		C	N	V
	Delfour et al. [57]		10/10		C	N	SQ
	Preusser et al. [58]		12/12		C	Y	V
Streck’s	Burns et al. [43]	6/10	4/10		R(ph)	N	SQ
UMFIX/RTP	Nadji et al. [44]	23/70	42/70	5/70	C	Y	V
	Nassiri et al. [40]		2/2		C	Y	SQ
	Vincek et al. [19]		29/29	1/29	C	N	V
Z7	Lykidis et al. [21]	3/3			C	Y	SQ
	Staff et al. [56]		7/7		C	N	V
ZF	Paavilainen et al. [25]			72/72	C/R	N	SQ
ZBF	Paavilainen et al. [25]	10/72	62/72		C/R	N	SQ
	Wester et al. [23]	5/9	2/9	2/9	C	Y	SQ

ZBF – Zinc based fixative.

C – clinical, R – research, (ph) – phospho.

Q – quantitative, SQ – semi-quantitative, V – visual.

Effects on histochemistry

The effect of molecular fixatives on histochemistry has not been extensively studied as they have focussed on the benefits of the fixatives to RNA/DNA analysis and analysis of protein levels. However, the aetiology of histochemistry has always been enmeshed in fixation regimes.

Histochemistry, has been utilised since the early 1900’s to allow the visualisation of many tissue components and especially prior to the extensive use of IHC. Fixation has been an important part of this genesis and the effect of the fixation is largely dependent on the stain being employed. For example, it is widely reported that formaldehyde fixation has a negative effect on trichrome stains, unless a mordant such as picric acid or mercuric-chloride solutions is used [46].

While few publications have focussed on the effect of molecular fixative and histochemistry, the few that have incorporated this into their manuscript have shown little effect on the stains employed. Kap et al. report an increase in intensity of goblet cell staining in PAS of intestinal samples in Paxgene-fixed samples [26]. Similarly, Moelans et al. reported that while FineFix, RCL2 and F-Solv gave adequate staining for PAS, PASD, Alcian Blue and a number of others, fixation in NBF and standard pre-optimised techniques were superior [27].

Technical evaluation of alternative fixatives

While a number of published studies utilise whole section staining to demonstrate the differences between fixatives, tissue microarray technology is particularly suited to this investigation. Originally envisaged by Kononen et al. in 1998 [47], this technique enables the researcher to sample multiple tissue “donor” blocks and replace into a single “recipient” block. Thus sections from the recipient block contain the representation of all of the donor blocks and controlled on a single slide (see Ilyas et al. for a recent review, [48]). To combine the TMA technology with multiple fixatives has been performed previously [25], but no comparison using this technology has been applied to histochemical stains. The following is a protocol for the construction of such a TMA and its staining with a modified Masson’s Trichrome protocols and their results.

Method

Tissue preparation

Balb/c mice were killed using a Schedule 1 procedure and the following tissues were excised to construct a gastrointestinal tract (GIT) tissue microarray: Caecum, Colon, Ileum and Oesophagus. Tissues were fixed individually in the following fixatives: Bouin’s, Carnoy’s, Methacarn, 10% Neutral Buffered Formaldehyde (Sigma, UK), 4% Paraformaldehyde, 2% Glutaraldehyde, Formol Saline and Zinc Formaldehyde. All solutions were made fresh prior to tissue fixation using standard protocols, other than 10% NBF which was purchased pre-made. Zinc formaldehyde was made using 1% Zinc Chloride in unbuffered 4% formaldehyde. All tissues were fixed for 24 h before being processed through a standardised routine tissue processing protocol. Mice were housed and maintained in accordance with UK Home Office regulations.

Tissue microarray construction

A tissue microarray (TMA) was constructed using the Beecher Manual Tissue Arrayer (MTA-1, Beecher Instruments, USA) utilising a 0.6 mm donor and recipient needles and 0.2 mm gap between cores. A single TMA core was taken from each fixation tissue block and a marker core of sheep kidney was used to delineate an asymmetrical pattern. All individual tissue blocks were orientated to allow transverse viewing of the tracts within the GIT array and thus when tissue microarray cores were taken, each represented the full thickness and tissue representation of the tract. The TMA was heated to 40 °C in a slide drying oven, before lightly compressing with a slide to flatten the top and bind the edges of the TMA core to the remainder of the recipient block. Following heating, the TMA was cooled on an embedding station cold plate, before repeating the cycle of heating and cooling.

TMA sections were cut at 3 μm and placed onto Superfrost Plus slides (Leica, UK) with the orientation core placed uppermost on the slide. All sections were oriented the same way and the entire tissue block was cut, with remaining sections dipped in wax and stored at room temperature.

Masson’s Trichrome staining

We utilised a modified Masson’s Trichrome protocol, modified in two steps. Firstly, we have adapted a standard human Masson’s protocol to murine tissue with the addition of Bouin’s as a mordant. Secondly, following Li et al. [49], we have omitted the Weigerts Haematoxylin nuclear counterstain to lighten the stain and improve the recognition of the collagen fibres for automated image analysis.

In detail, following deparaffinisation, slides were pretreated in Bouin’s fixative overnight before being washed in running tap water until the water ran colourless. Following washing, the sections were stained with Acid Fuchsin for 10 min and rinsed in deionised water. A 5 min treatment with Phosphomolybdic acid followed and then immediately stained with methyl blue before being washed, cleared swiftly through alcohols and mounted in DPX.

Image digitisation & analysis

All slides were digitised using an Aperio XT slide scanner (Leica, UK) at 20× magnification and stored in .svs file format.

20–35 data points of the thickness of the oesophageal wall were taken using the ruler tool in Aperio Imagescope v12 software across all fixatives.

Positive pixel analysis v9.1 (Leica) was used to analyse the number of blue positive pixels in the oesophageal samples, representing collagen fibres, where the positive pixels were trained according to the NBF-fixed oesophagus. Specifically, Hue value was set to 0.63, Hue width to 0.2 and Colour saturation threshold to 0.01. All other colours were delineated as negative pixels. Staining intensity was binned into three bins; 220–175 grey levels (weak positive pixels); 175–100 grey levels (positive pixels); 100–0 (strong positive pixels). The sub-mucosal area of the oesophagus was specifically isolated and analysed in each sample to ensure reproducibility and eliminate variation relative to the shrinkage of the specimen.

Statistical analysis was performed with GraphPad Prism v6 using a Mann–Whitney U test between two fixation variables.

Results and discussion

Morphology

The overall morphology of the specimens was acceptable in all cases. Image analysis of the oesophageal mucosa demonstrated a significant shrinkage in the thickness of the oesophageal wall in Carnoy’s-fixed specimens relative to NBF-fixed specimens (Fig 1) and a significant increase in the thickness of Formol Saline-fixed oesophagus compared to NBF. It is surprising that there was no significant decrease in Methacarn-fixed oesophagus, as it is likely that methanol causes the same shrinkage as ethanol, however, Puchtler et al. do report no shrinkage in their methacarn-fixed specimens compared with Carnoy’s-fixed samples [3]. Similarly, there are no reports that Formol Saline significantly swells epithelial tissue as it has been used extensively for many years and is recognised to protect the dimensions of the tissue well. To investigate whether this difference was due to an increase in the preservation of the keratinised oesophageal layer, the wall thickness measurements were repeated for NBF, Formol Saline, Methacarn and Carnoy’s-fixed tissue, removing the keratinised layer, where it could be observed. This measure confirmed the original observations and the significance level (data not shown) and, importantly, did not change the lack of significance for the Methacarn-fixed samples nor the significance of Formol Saline-fixed samples.

Fig. 1

Graph demonstrating the differences in thickness of the oesophageal wall between each fixative. Mean and SE. Significance applied at <0.05.

As reported for molecular fixatives [19], there was loss of the morphology of erythrocytes in Carnoy’s and Methacarn specimens when compared to formaldehyde fixed specimens (NBF, PFA, FS, ZF). While this study did not trial the molecular fixatives, it is unlikely that any different result would be achieved on a morphological perspective as the base solution of most molecular fixatives is dehydrating.

Histochemistry

Masson’s Trichrome

Oesophagus was chosen as an example tissue to study since the entire cross-sectional area contained a wide variety of tissue types, including epithelial and sub-mucosal connective tissue, as well as smooth muscle. It was unfortunate that the oesophageal sample fixed in Bouin’s directly was mis-cored and a sample of trachea was in its place and thus the direct comparison between a Bouin’s-mordanted NBF sample and a Bouin’s sample alone was not possible.

The modified Masson’s protocol for formaldehyde fixed tissue (Fig 2C), which is post-fixed in Bouin’s and lacks a nuclear counterstain, demonstrated the clearest separation of collagen fibres from epithelial tissue of the oesophagus and clear delineation of the fibres surrounding the smooth muscle. The standard Masson’s protocol for use with human tissue and the post-fixed protocol with nuclear counterstain were both used to stain a section of GIT TMA and were rejected in comparison (Supplementary Fig 1).

Fig. 2

Masson’s Trichrome staining of murine oesophagus following fixation in a variety of fixatives. Representative images of oesophageal wall, sub-mucosa and surrounding muscle are demonstrated. (A) Carnoy’s, (B) Methacarn, (C) NBF, (D) PFA, (E) Glutaraldehyde, (F) Formol Saline, (G) Zinc Formalin.

The upper keratinised layer of the oesophagus was clearly stained with NBF compared to any other fixative. While the other aldehyde-based fixatives, PFA, Glutaraldehyde, Formol Saline and Zinc Formalin (Fig 2D–G), all demonstrated a good level of differentiation between the individual tissue components, the clarity in separation of the epithelial component, keratin and sub-mucosal collagen was not as distinct as NBF. Carnoy’s and Methacarn were notably similar in appearance, with no differentiation between epithelium and keratin and a notable darkness in intensity of the collagen fibres thus rendering them less distinguishable from each other (Fig 2A, B). Both Formol Saline and Zinc Formalin appeared to have a lesser amount of collagen present in the sub-mucosa compared to the other groups.

Quantitative image analysis of the sub-mucosal compartment confirmed the qualitative observations. The percentage positivity, reflecting the proportion of blue collagen fibres in the sub-mucosa, was 64% and reflected staining intensity that crossed all categories, from weak to strong as was expected given that this was the tissue that the image analysis algorithm was trained on (Fig 3).

Fig. 3

Graph demonstrating the analysis of percentage positivity applied to the sub-mucosal Masson’s Trichrome stained oesophagus. Each pixel is split into three “bins” of Strong, Medium and Weak positive and calculated as a percentage compared to overall pixel number.

The percentage of pixels designated as positive in Formol Saline and Zinc Formalin samples at 18% and 14%, respectively, was clearly lower than that for NBF, but demonstrated the same degree of split across all intensity ranges. In contrast, Carnoy’s and Methacarn samples showed an increased level of positivity at 80% and 67%, respectively and the majority of the pixel intensity fell into the strong positivity category (Fig 3). PFA and Glutaraldehyde showed no difference compared to NBF. Statistical analysis was not performed on these samples as only a single area measurement was performed for the positive pixel count and thus there is insufficient power for statistical calculation.

As weaker fixatives, it is not surprising that the relative ratio of intensity was greater in the Carnoy’s and Methacarn samples and it could be assumed that molecular fixatives would behave in a broadly similar way. It is widely reported [26,40,42,43,50] that molecular fixatives, as well as Carnoy’s and Methacarn, do not require antigen retrieval for immunohistochemistry, concomitant with their weaker fixation and in contrast to NBF. Indeed, that is one of the advantages of an ethanol-based fixative. It is probable that a modification of the Trichrome to differentiate further and remove some of the intensity of the collagen stain, would provide a more suitable depth of colour of stain should Carnoy’s or Methacarn be used routinely.

The results from the Formol Saline and Zinc Formalin-fixed samples are unexpected and a visual examination of the data demonstrates that this is not an issue of algorithm validation as the algorithm has reported intensity measures that are broadly similar to NBF, but that there is simply less collagen present. Whether this represents a trend or reflection of the fixation regime requires further investigation.

It is clear that a fixation TMA approach provides a unique method to evaluate the effect of fixation on protein, RNA & DNA or histochemical stains and allows the scientist to evaluate all of the fixatives on one slide. This not only demonstrates the variation clearly and quickly but also provides the reassurance that this is an effect of the fixation itself, rather than a slide-to-slide variation. As histochemical stains are often performed manually, this is more likely to be the case. Using this technology, in so far as a single histochemical stain is concerned, we did not demonstrate any significant reason to change fixation regime from NBF-fixed tissue and that there are differences between the other fixatives in terms of intensity and depth of colour of stain, which may be ameliorated by modification of the technique specific for that fixation.

Conclusions

Fixation is a vital part of the overall life of a tissue sample and cannot be oversimplified. The type of fixative used, whether cross-linking or dehydrating, will in some way compromise the morphology, RNA/DNA extraction ability, protein evaluation or histochemical staining of the tissue and therefore the fixation regime used can be tailored to the end-result. However, neutral-buffered formaldehyde remains the fixative of choice in the majority of histological laboratories for good reason. While other fixation regimes may increase the read length of RNA, the utilisation of alternative technologies (see Cassidy & Jones, 2014) and extraction techniques may circumvent this. Similarly, while some laboratories report an improvement in the detection of epitopes with alternative fixation regimes, an equal or greater number do not. Finally, we, in confirmation of other laboratories, have not been able to demonstrate any significant advantage in using alternative fixatives over formaldehyde, when using histochemical stains as a baseline. In order to carefully examine the best fixation regime for the sample, tissue microarrays provide the easiest and most cost-effective route for the evaluation of a multiple fixation trial.