Fixation methods can differentially affect ciliary protein immunolabeling

Primary cilia are immotile, usually found singularly per cell, and are recognized for their roles in signaling and development [1]. Structurally, the primary cilium is composed of the basal body and the axoneme [1–4]. The basal body is a mature mother centriole that has docked to the plasma membrane, and it gives rise to the microtubules that form the cilium [5]. These microtubules form the shaft or axoneme of the cilium, and are arranged in a 9+0 pattern that consists of 9 doublets of microtubules arranged in a circular fashion [6]. This is in contrast to motile cilia which have a 9+2 pattern that consists of 9 microtubule doublets surrounding a central pair of microtubules [6]. Primary cilia are organelles and represent a separate compartment of the cell, meaning that the cilia-membrane and cilioplasm are distinct from the plasma membrane and cytoplasm, respectively [7]. The cilium also has its own transport system, the intraflagellar transport system (IFT), consisting of motor protein complexes that carry proteins in an anterograde and retrograde manner along the microtubular axoneme [8]. A defect in any of the proteins important for the assembly, maintenance and/or function of the primary cilium can result in a category of developmental diseases called ciliopathies [1, 2, 9].

The history leading to our current understanding of the primary cilium is partly dependent on technological advances. Primary cilia were first observed in 1898 by the Swiss anatomist, KW Zimmerman, who drew images that depicted the mother and daughter centrioles with a primary cilium protruding into the luminal space of a kidney tubule [10, 11]. Zimmerman noticed that this immotile structure was found one per cell, and named it the “centralgeissel”, meaning the central flagellum, and surmised that it had a sensory function [11]. However, it was not until the invention of the electron microscope that this organelle was verified [10, 12]. In 1985, Poole et al. speculated that primary cilia have chemical and sensory roles [13], assumptions that have now been verified by multiple labs [14–17]. Here, we suggest that an important subcategorization within the primary cilia field will lie in the study of primary cilia proteins at extraciliary sites [18].

Many cilia proteins localize to cellular sites besides the cilium [18]. For example, centrosome and spindle pole associated protein 1 (CSPP1) was initially described as a centrosome and mitotic spindle protein [19], and was later found to also localize in primary cilia [20]. CSPP1 has now also been found to localize to desmosomes [21] and kinetochores [22]. Arl13b is not only found in cilia, but also co-labels with endocytic markers [23]. Intraflagellar transport protein 20 (IFT20) is another cilia marker, and it also localizes to the Golgi [24]. Therefore, further research is needed to understand the role of cilia proteins at extraciliary sites and how this might contribute to the underlying pathologies of ciliopathies [18].

Both the microtubule and actin cytoskeleton have been shown to be critical for proper primary cilia formation/function. It is not surprising that the microtubule cytoskeleton plays a role in ciliogenesis as the cilium is a microtubule-based structure. In fact, CSPP1, a ciliogenesis protein, has a microtubule-binding domain [25]. Defects in another ciliogenesis protein, Ahi1, have also been reported to result in a disorganized microtubule cytoskeleton [26], suggesting that Ahi1 may have a role in microtubule organization. Moreover, knockdown of Ahi1 in IMCD3 cells was shown to result in a disorganized actin network as well [26]. Actin proteins are now increasingly being shown to be important for ciliogenesis. Recently, multiple labs reported that non-muscle myosin heavy chain 10 (MYH10), an actin regulating protein, is also necessary for ciliogenesis [27, 28]. Interestingly, MYH10 does not localize to the basal body or axoneme of primary cilia, but loss of MYH10 results in loss of cilia [27]. This suggests (1) that proper functioning of the microtubule and actin cytoskeleton is necessary for the construction of the primary cilium, and/or (2) that at least some cilia proteins also function as more general cytoskeletal proteins (i.e., regulators of actin and microtubules). Consequently, understanding how different fixation techniques alter the actin and microtubule cytoskeleton, as well as the primary cilium, is critical for understanding ciliopathies.

The effect of fixation on ciliary protein localization via immunocytochemistry can be demonstrated with the cilia-associated protein, CSPP1 [20]. CSPP1 was initially identified as a protein that localizes to centrosomes and mitotic spindles [19]. Subsequently, CSPP1 was found to localize to primary cilia in methanol-fixed cells [20]. The fixation process used is important because paraformaldehyde fixation does not reliably yield cilia immunolabeling when using the same CSPP1 antibody (unpublished observations). Paraformaldehyde is known to disrupt the native conformation of microtubules and can hide cilia immunostaining for some cilia markers [29]. This problem can sometimes be mitigated by use of methanol as a fixation agent; however, methanol obscures the phalloidin epitope (a widely used reagent to view actin stress fibers). Therefore, alternative fixation methods are necessary to allow for the reliable, concurrent viewing of microtubules, phalloidin-stained actin stress fibers, and cilia markers.

Preservation of the cytoskeleton during fixation has historically been achieved through the use of cytoskeleton buffers [30]. In vitro observations show that tubulin polymerized when (1) calcium is absent, (2) magnesium is added to the buffer, and (3) when the incubation occurred at 35 °C as opposed to 0 °C [31]. This led to the development and use of various forms of cytoskeletal buffers (also referred to as extraction buffers) that consisted of EGTA (a calcium chelator), magnesium, and Triton X-100 detergent for extraction [32]. These cytoskeleton buffers proved useful for studying CSPP1 when in 2014, three laboratories published papers that showed CSPP1 was a causative gene for Joubert syndrome (a neurodevelopmental ciliopathy). However, there were discrepancies with two of the laboratories reporting different immunolabeling patterns for CSPP1 [33, 34]. Both laboratories used primary human dermal fibroblasts collected from control subjects and individuals with Joubert syndrome. One laboratory showed CSPP1 localization to the centrosomes [33], while our laboratory observed CSPP1 localization also at the axoneme of the primary cilium [34]. We found that when studying CSPP1, a microtubule-stabilizing buffer similar to cytoskeletal buffer was required to reveal consistent and reliable CSPP1 labeling at the ciliary axoneme [34]. These studies indicate that careful consideration and understanding of fixation methods are important for interpreting localizations of ciliary proteins. For that reason, we explored the advantages and disadvantages of various fixation methods in a systematic and comprehensive attempt to elucidate how these different methods affect immunolabeling of popularly used cilia markers. Our results would advocate the use of cytoskeletal buffers during cell fixation, which largely preserves labeling of cilia, microtubules, actin stress fibers, and at least the two extraciliary sites we examined, mitotic figures and Golgi.

It is becoming increasingly clear that cilia proteins can also have extraciliary localizations and functions [18, 26, 37, 38]. Further work is needed to determine which ciliary proteins have extraciliary functions, and whether these extraciliary functions of cilia proteins contribute to the pathology of ciliopathies. If so, what symptoms can be attributed to defective cilia, and what symptoms are related to defects at extraciliary sites? As these questions are explored and the function of these extraciliary sites for ciliary proteins is revealed, an understanding of how fixation affects immunofluorescent labeling is warranted. As described earlier, an inadequate exploration of fixation techniques has led to contradictory conclusions by different laboratories. CSPP1 has been reported to be exclusively a centrosomal protein [33], exclusively an axonemal protein [34], or both a centrosomal and an axonemal protein [20] depending upon the fixation method used. Even widely used antibodies like anti-acetylated α-tubulin can produce differing results depending on the method of fixation utilized. The acetylated α-tubulin antibody immunolabels both cilia and centrosomes in MeOH-fixed cells, but in PFA-fixed cells, immunolabeling reveals only cilia with a few cells exhibiting centrosomal labeling. Therefore, here we wanted to conduct a comprehensive and systematic examination of the effects of fixation on ciliary marker immunolabeling.

The goal of this study was to establish a reliable and flexible fixation technique that could concurrently visualize cilia, microtubules, and actin stress fibers (as assessed with phalloidin). However, current results demonstrate that each fixation method has advantages and disadvantages, and it may be necessary to use different fixation methods when studying different areas of a cell even when studying the same protein. For example, PFA-S fixation reveals AC3 to be at the primary cilium in IMCD3 cells, but when used to look at mitotic figures, AC3 only labels centrioles. However, if the goal was to examine AC3 immunolabeling at mitotic figures and not cilia, then any of the CB fixation methods would be preferable over PFA-S as these methods reveal AC3 localization to the mitotic spindles as well as the centrosomes; a finding that is novel to our knowledge. Unfortunately, the CB fixation methods result in a loss of primary cilia staining with the AC3 antibody, so we were unable to find one fixation protocol that could reveal all of the ciliary and extraciliary localizations for AC3. Therefore, the fixation method used will be dependent on the antibody and the cellular structure that is being studied.

By utilizing the two workhorse cells in the cilia field, mouse IMCD3 and human RPE cells, we have found that for some cilia markers, immunolabeling patterns are dependent on the fixation technique used and can be cell type specific. Comparison of two different cell types allows for the examination of potential cell type differences, but attention is warranted as to how much we can extrapolate from these findings. IMCD3 and RPE cells are both epithelial type cells, but IMCD3s are mouse kidney cells whereas RPE cells are derived from human eye. Therefore, any difference we see between RPE and IMCD3 protein expression patterns could be due to cell type differences, species differences, or antibody binding differences. It is beyond the scope of this manuscript to examine more cell types to tease out these differences, especially since we wanted to use already established ciliary cell lines. However, our results indicate that each cell line utilized should undergo some degree of troubleshooting for the best fixation methodology to use with the hope that our approach provides some directed guidance.

Of the four ciliary markers we examined, Arl13b and AC3 were directed against mouse epitopes, and IFT20 and CSPP1 were directed against human epitopes. Arl13b was the only monoclonal antibody with the three remaining antibodies being polyclonal antibodies. Differences in human vs. mouse reactive epitopes among antibodies can sometimes yield staining differences, but we did not observe any in our data when comparing IMCD3 and RPE cells. For example, CSPP1 is directed against human CSPP1 protein, but it does not appear to label human RPE cells any better than in the mouse IMCD3 cell line. As a whole, our data comparing cilia protein immunostaining in IMCD3 and RPE cells look remarkably similar.

When we applied a 10-min fixation incubation time for all fixation methods to both cell types, we found overall that RPE cells maintained a more intact microtubule cytoskeleton. This may not only be due to cell type-specific differences in structural resilience of the microtubule cytoskeleton to fixation agents, but may also be due to the fact that IMCD3 cells are smaller than RPE cells. Thus, it may be necessary to empirically determine the proper fixation method and incubation period for each individual cell type and fixation method used in a study. Here, by applying the same 10-min incubation period for all experiments, we hoped to gain a general understanding of how various fixation agents affected the immunolabeling pattern for some popularly used cilia markers.

We fixed our cells at three different temperatures: −20 °C for methanol fixation, room temperature for PFA fixation techniques, and 37 °C for cytoskeletal buffer-based techniques. We chose these temperatures based on what is most appropriate for each individual fixation agent. Methanol must be used cold. PFA is traditionally used at room temperature. CB and CB-PFA were used at 37 °C because Weisenberg showed that in vitro tubulin polymerizes better at more physiological temperatures [31]. However, the temperature at which a cell is fixed may influence the resulting immunolabeling pattern, since colder fixation temperatures may benefit proteins that are susceptible to rapid degradation. While it is possible that the different temperatures of wash buffers and fixatives utilized in our experiments may affect cilia immunostaining, we avoided straying from conventional temperatures used for these fixation agents to stay consistent with what is already done in the field. However, temperature of washes and fixation could be an important variable for consideration in examining ciliary and extraciliary immunolabeling localizations.

We examined a variety of different fixation techniques to see which methods were amenable to actin stress fiber staining by phalloidin, microtubule and cilia staining by β-tubulin, and three antibodies directed towards tubulin post-translational modifications. We found that acetylated α-tubulin labeled microtubules comparably with all fixation methods used, but the MeOH-fixed cells had the distinction of having much more prominent centriolar labeling at the base of cilia. Using a CB wash before fixing cells also offers a more “washed out” cytoplasmic appearance, offering greater contrast between microtubules and a cleared cytoplasm. Phalloidin labeling was preserved in fixation methods that did not use MeOH (PFA-S, CB → CB-PFA, and CB-PFA), but was damaged in any method that used MeOH regardless of whether MeOH was used alone or after being fixed with CB and PFA. Thus, CB does not damage phalloidin staining of actin stress fibers, since they were preserved in both the CB-PFA and CB → CB-PFA fixation groups. But PFA-S did not protect the phalloidin epitope from MeOH, since the CB → CB-PFA → MeOH fixation group lost actin stress fiber labeling as assessed by phalloidin. Finally, we saw no benefit in using MeOH over PFA since PFA fixation techniques offered more replicable results and have the advantage that it does not damage the phalloidin epitope.

In this manuscript, we purposely chose commonly used and published ciliary markers that are already established by others in the field. We also are cautious to not make any functional claims about any protein. We do note the location of each protein as it differs from one fixation method to the other to illustrate that fixation methods can affect the staining pattern of an antibody. But we acknowledge that each finding in this paper must be empirically verified to determine the functional significance of these localizations through knockout or knockdown studies to draw any conclusions about the function of a protein and the specificity of an antibody.

When our assortment of fixation methods was tested against our selection of cilia markers, a pattern was quickly noticed that may be generally applicable to other ciliary antibodies. The cilia markers appear to be able to be separated into three distinct groups: (1) tubulin structural markers, (2) membrane-associated proteins, and (3) axoneme-associated proteins (Fig. 9). All three post-translational modification antibodies and the β-tubulin antibody immunolabeled cilia regardless of the fixation method used although some were more effective than others. It appears that structural proteins or at least epitopes for these structural proteins are stable against PFA, methanol, and variations of cytoskeletal buffer fixation techniques. Conversely, membrane-bound proteins (AC3 and Arl13b) were best preserved at cilia when immunolabeled after PFA fixation. Arl13b cilia immunolabeling was also observed in IMCD3 and RPE cells that were fixed with MeOH or CB-PFA. We surmise that by combining CB with PFA, to make a 4% CB-PFA fixation solution, we were able to combine the cytoskeletal stabilizing benefits of CB along with the quick fixation properties of PFA, preserving Arl13b localization at cilia before any harsh membrane-washing could take place. But fixation methods that used a CB pre-wash appear to have washed away the Arl13b and AC3 cilia labeling. It is interesting that while CB-PFA fixation was able to preserve Arl13b labeling at cilia, it could not preserve AC3 cilia immunolabeling, suggesting that this method has its limitations. Last, axoneme-associated proteins (CSPP1 and IFT20) unreliably immunolabel cilia when fixed with PFA, but are revealed as cilia markers when cells are fixed with either of the CB fixation methods. This suggests that either (1) CB stabilizes the epitope of axoneme-associated proteins, or (2) CB washes away the cilia membrane to allow antibodies to access the axoneme-associated proteins for labeling. The former hypothesis is more likely since tubulin structural markers, which reside along the axoneme, are accessible without the use of CB. In addition, when we applied our fixation guidelines to IFT88, we found that IFT88 defies our categorization as it immunolabels cilia when fixed with PFA alone, methanol, or cytoskeletal buffer-based techniques. However, we observed that IFT88 labels the cilium more distinctly when CB is used, so while CB may not be necessary to observe IFT88-positive cilia, CB can be used to help improve cilia immunolabeling.

The MeOH-fixed groups were the least replicable of the fixation methods we tested, and therefore, were the most difficult groups to characterize. Moreover, MeOH is still not an ideal fixation agent for our goals as it results in distorted DNA labeling, and obscures the epitope for phalloidin. Finally, MeOH fixation does not fit our categorization of membrane-localizing and axoneme-localizing proteins as it preserves some but not all proteins from each group.

Our goals for this paper were to demonstrate the importance of fixation on cilia protein immunolabeling, and to establish guidelines that may help others optimize their fixation protocols for studying cilia proteins at ciliary and extraciliary sites. While we could not analyze all the many extraciliary sites now known in the field to be important for cilia proteins, we did choose two sites to examine our fixation methods in (1) mitotic figures and (2) cis-Golgi. Mitotic figures are similar to cilia in that they consist of microtubules that extend from centrosomes, so it is not surprising that many proteins that localize to the cilium and mother/daughter centrioles are also found at mitotic figures. Microtubules of mitotic figures proved to be similar to microtubules of the cytoskeleton in that a tubulin structural marker, acetylated α-tubulin, was able to immunolabel them with all fixation methods tested. Cilia protein labeling at mitotic figures was observed as expected with varying degrees of distinctiveness depending on the fixation method used: CSPP1 and IFT20 labeled the microtubules, AC3 labeled the centrosomes, and Arl13b did not label any part of the mitotic figure. Interestingly, AC3 appears to label the microtubules of the mitotic figure in CB fixation groups, a finding that is novel to our understanding, and requires further investigation to verify. The methanol and CB groups revealed less background staining when compared to the PFA-S group, making the microtubule and centrosome staining more distinct.

Fixation methods also affected IFT20 staining at the Golgi. IFT20 is known to localize at the Golgi and has been observed at the Golgi in PHEM-treated cells (another version of a cytoskeletal buffer) and methanol-treated cells [24]. Our results were similar. IFT20 co-localized with GM130, a cis-Golgi marker, only when IMCD3 and RPE cells were treated with methanol or CB-PFA. Fixation methods also appear to affect GM130 staining as GM130 labeling was often lost in the CB fixation groups. Altogether, these results show that not unlike at the cilium, cilia protein localization at extraciliary sites is also affected by the fixation method used.

When looking at the totality of our data, no one fixation method can be used to preserve all groups of cilia proteins, but CB-PFA comes close. CB-PFA is suitable for use when staining for microtubules, phalloidin-stained actin stress fibers, most cilia proteins we tested, and for at least two extraciliary sites (mitotic figures and IFT20 labeled Golgi). CB-PFA preserved Arl13b (a membrane-associated protein) and CSPP1 (an axoneme-associated protein) staining at cilia, but it appears to wash away AC3 cilia labeling, and it is not a reliable method for IFT20 cilia staining. This discrepancy in cilia immunolabeling within each cilia protein group is not straightforward, but intriguing. Could AC3 be more vulnerable to being washed off the cilia membrane than Arl13b? Are there different compartments of varying stabilities on a ciliary membrane much like lipid rafts exist as distinct compartments on the plasma membrane? Finally, it is also noteworthy that while IFT20 lightly and unreliably labels cilia in CB-PFA-fixed cells, cilia staining with ITF20 is dramatically improved when a CB pre-wash is utilized. A CB pre-wash is harsher than fixing with CB-PFA alone, so this might suggest that the location of IFT20 on the ciliary axoneme requires harsher permeabilization to access than CSPP1? Such questions require further study and may be elucidated through an understanding of different fixation techniques.

As the cilia field moves forward in examining non-ciliary sites and as the field continues to expand, it is useful to consider and appreciate alternate fixation methods. Our findings have revealed a level of complexity in cilia proteins that suggests cilia proteins can be separated into different groups and each group benefits from different fixation methods. Structural proteins like β-tubulin and the various post-translational modification markers are resilient to various fixation methods. Cilia-membrane proteins are best fixed with PFA alone, but axonemal proteins are often obscured with PFA fixation and benefit from CB fixation (Fig. 9). The most universal method we found was the CB-PFA method (however, there can be exceptions; Fig. 9). CB-PFA appears to be the most versatile of the fixation methods we tested as it preserves the microtubule cytoskeleton, phalloidin labeled actin stress fibers, and most cilia labeling. This method can also easily be adjusted to reveal more distinct cilia labeling by adding a CB wash to the protocol, though this can come at a cost (e.g., loss of Arl13b labeling). Although there are caveats for using any fixation method, we advocate the usage of cytoskeleton buffer as a starting point, since the primary cilium is essentially a cytoskeletal organelle composed of microtubules extending from the cytoskeleton of the cell.