Magnetic field inhomogeneities due to CO2 incubator shelves: a source of experimental confounding and variability?

A thorough assessment of the static magnetic field (SMF) inside a CO2 incubator allowed us to identify non-negligible inhomogeneities close to the floor, ceiling, walls and the door. Given that incubator's shelves are made of a non-magnetic stainless steel alloy, we did not expect any important effect of them on the SMF. Surprisingly, we did find relatively strong distortion of the SMF due to shelves. Indeed, our high-resolution maps of the SMF revealed that distortion is such that field intensities differing by a factor of up to 36 were measured on the surface of the shelf at locations only few millimetres apart from each other. Furthermore, the most intense of these fields was around five times greater than the ones found inside the incubator (without the metallic shelves in), while the lowest one was around 10 times lower, reaching the so-called hypomagnetic field range. Our findings, together with a survey of the literature on biological effects of hypomagnetic fields, soundly support the idea that SMF inhomogeneities inside incubators, especially due to shelves' holes, are a potential source of confounding and variability in experiments with cell cultures kept in an incubator.


Introduction
According to McDonald [1], a confounding variable is a variable other than the independent variable that we are interested in, that may affect the dependent variable, i.e. the endpoint of our experiment. Hence, a key aspect of experimental design is the identification of possible confounding variables and some kind of control of them. The main message of this work is that there might be a confounding variable (almost) completely unaccounted for in biology laboratories working with incubators: the static magnetic fields (SMFs), and their inhomogeneities throughout the volume of incubators, especially on the surface of the shelves, where cell cultures are commonly kept in biological experiments.
The matter has, in fact, received some attention during the last years among magnetobiology researchers, who have discussed and measured direct current (DC, or 'static') and alternated current (AC) background magnetic fields (MFs) inside CO 2 incubators. Hansson Mild et al. [2] measured AC fields inside an incubator and discussed the importance of the ones generated by the incubator's fan, also commenting on equipment placed near the incubator being a possible source of stray fields. In line with that discussion, Gresits et al. [3] measured AC fields not only in a CO 2 incubator, but also near a thermostatic water bath and a laboratory shaker table, finding non-negligible intensities, with significant variations over relatively short distances. To the best of our knowledge, the most extensive work on CO 2 incubators was done by Portelli et al. [4], who measured both DC and AC fields inside of 21 different incubators. These fields are of particular importance in experiments of weak DC and/or extremely low frequency MFs, where the field intensities under study are of the same order of magnitude as the fields typically present inside incubators, clearly posing them as a potential confounding variable, and also making replication by independent laboratories more difficult [5].
As stated above, we suggest that background fields should be of concern to all researchers working with incubators, not only the ones devoted to magnetobiology. In the following sections, we make our case by: (i) presenting a thorough assessment of the SMFs inside a typical CO 2 incubator including, for the first time, a high-resolution mapping of the fields near several holes of a standard, stainless steel (SS) shelf and (ii) discussing our measurements in the context of an up-to-date survey of the literature on biological effects of weak SMFs, including the so-called hypomagnetic fields.

Material and methods
All measurements were performed in a HERACell 150i CO 2 Incubator (Thermo Fisher Scientific, MA; figure 1a). Metallic shelves provided with the incubator by the manufacturer were made of the austenitic SS AISI 304, an extremely common and well-known alloy. We also used an 8 mm thick 420 × 465 mm plastic shelf made of polymethyl methacrylate (PMMA). Measurements were performed with an HMC5883 L 3-axis magnetometer (Honeywell, New Jersey, NY) connected to a personal computer, as detailed elsewhere [6]. For this work, each independent axis of the sensor was calibrated against a TM75-41 magnetometer (Izmiran, Fryazino, Russia). Each of the MF maps presented in the next section was measured by manually locating the sensor in each of the points of a regular grid. Upon the assumption that the holes of the metallic shelves could be a source of distortion of the SMFs inside the incubator, we decided to use a grid with a periodicity that would match that of the shelf's holes, so that all the measurements of a given map would be performed in equivalent points of the holes' lattice. Hence, according to the dimensions indicated in figure 1b, we defined one first (coarse) grid with a unit of 33 × 20 mm 2 , which covered almost all the usable area of the shelf. Also, suspecting that the field at the centre of the holes could be different from that at the edges, or from that between holes, we measured three maps with the coarse grid, shifting its position just 5 mm to the right each time (holes' diameter was 10 mm; figure 2a-c). Measurements with the coarse grid led us to a medium grid which, in term, led us to further refinement. Locations of the grids on the shelf are shown as grey rectangular areas in figure 1b, and further details are presented in table 1, which shows that the smallest of our pixels were 1 mm 2 . While the chip that contains the three orthogonal sensors of our magnetometer has external dimensions of 3.0 × 3.0 × 0.9 mm, we considered it reasonable to assume that the sensors themselves are confined in an area smaller than 1 mm 2 , given the presence of the plastic packaging encapsulating the electronics, plus the fact that circuitry auxiliary to the sensors is also contained inside the same chip (a multiplexer, an analogue to a digital converter and a control unit, among others).    perimeter), because 'low-field' horizontal blue bands are present in figure 2e. Also, these maps strongly suggested that the used grid was not fine enough (i.e. pixels were too big) to assess the fields in full detail. Therefore, we used a medium grid with pixels of 5.5 × 5 mm 2 to explore an area (black rectangle) around the 'hot-spots'      evident how field is homogenized as distance from the shelf increases, fading into the background field. Table 2 displays minima, maxima and differences ( ) corresponding to all maps in figure 3.

Results
Having determined that the metallic shelves were indeed a source of non-negligible distortion, in order to study the eventual distortion of the fields only due to the incubator's walls, ceiling and floor, all metallic shelves were removed from the incubator, the plastic shelf was positioned at each of its 11 levels and fields were assessed using the coarse grid. Figure 4a shows the 11 MF maps, all plotted in the same colour map scale for ease of comparison (see table 3 for numerical values). First observation to be made is that maps are smooth (i.e. only slightly 'pixelated'): without the distortion from the metallic shelves,     the coarse grid turns out to be appropriate for studying the fields inside the incubator. It is clear that variations within several centimetres are relatively subtle both in each level, and between consecutive levels (which are 4 cm apart). However, differences greater than 40 µT are observed in the upper shelves. Also, even though differences between consecutive shelves are small, the first and the eleventh levels are clearly different. Lastly, we evaluated the effect of performing the measurements either with the incubator door open, or closed (figure 4b). The maps clearly show that components B x and B y (and most notably this latter, perpendicular to the plane of the door) are the most affected by the closing of the door, while B z is almost unaffected. Comparing differences for the open and closed condition (see for levels 6 and 6* in table 3), it is evident that closing the door has a slight homogenizing effect on the fields inside the incubator.

Discussion and conclusion
Given that incubator's shelves are made of a non-magnetic SS alloy (AISI 304, relative magnetic permeability µ r = 1), we did not expect any important effect of them on the SMF. Surprisingly, we did find relatively strong distortion of the SMF due to shelves. Thus, a first question to address is how CO 2 incubator shelves can display the magnetizations we measured. The answer has long been known in the metallurgy industry. Manufacturing processes like folding, stamping, drilling, extruding or punching can induce an structural transition from austenite to martensite, leading to a weak magnetization of the otherwise non-magnetic SSs [8][9][10]. Furthermore, this effect has been studied in detail in the particular alloy of our shelves [11][12][13]. It is worth noting that even though the explanation of our findings is based on facts well known by the metallurgy industry, they are probably ignored by most part of the scientific community working with cell cultures, including magnetobiology researchers. Secondly, it is fair to question the relevance of our findings: could the MF inhomogeneities that we found actually affect the outcome of an experiment through confounding or increasing the variability of a biological endpoint? While SMFs' capability of eliciting biological effects has extensively been demonstrated (e.g. see review by the World Health Organization [14]), it is also true that most of such reports dealt with fields orders of magnitude stronger than the ones we measured (i.e. from mT to several teslas). Hence, it is appropriate to retrieve here a diversity of studies with fields no greater than 415 µT (i.e. the absolute maximum among all our measurements), reporting effects on cell-free systems [15][16][17][18][19], genotoxicity [20][21][22], in vivo neurophysiological effects [23][24][25][26], in vivo sensory receptors [27], analgesia [28,29], behaviour [30][31][32], muscles [33], pineal gland [34,35], development [36], modulation of hydrogen peroxide production [37] and endothelial cell proliferation [38]. Furthermore, in their extensive review, Binhi & Prato [39] gathered and analysed over 130 articles on effects of fields between 0 and 10 µT (hypomagnetic fields). These effects were observed when compared with samples 'exposed' to the geomagnetic field, which takes values in the range of 23-64 µT, depending on the location on the Earth [40]. Moreover, an example of special interest for the present work is the study by Martino et al. [41] on fibrosarcoma and colorectal cancer cells, because the authors reported changes of proliferation upon differences of approximately 35-45 µT, a range that includes the ones we measured within a single shelf, for several shelves (see s for |B| at shelves 8-11 in table 3). This indicates that, even using plastic shelves, proliferation can indeed be significantly affected by the exact location of cultures on the same shelf.
A further detail to point out is that in standard multi-well plates, typical vertical distances from inside wells' bottom to the resting plane (e.g. the shelf inside an incubator) are of 3.0 mm (Thermo Scientific, MA) or 3.53 mm (Corning, NY), while under typical experimental design in Petri dishes cells can lie 1.09 mm (MatTek, MA) over the resting plane, or as close as 0.17 mm in case of glass bottom Petri dishes (Ted Pella, CA; Cellvis, CA). In table 3, we show that at a height of 1 mm, differences as high as 250.6 µT were measured within a few millimetres distance (hole 2), while at a height of 3 mm the difference was of 34.6 µT (hole 3).
In summary, we conclude that our measurements, along with the data retrieved from the literature in the preceding paragraphs, make it sensible to suggest that SMF inhomogeneities inside incubators, and especially at typical experiment location of cells regarding metallic shelves, can be a source of confounding and variability. Consequently, the use of non-metallic shelves, along with bearing in mind the exact location of cultures inside the incubator (even on the same shelf), could enhance in-lab repeatability of results throughout all disciplines working with cell cultures in incubators, regardless of their specialty.