Soon after the incubation begins, the yolk moves from the center of the egg to the upper end. Embryonic blood vessels are forming beneath the yolk membrane. Similar as described in [14], the extraembryonic blood can be analyzed in ovo by fluorescence spectroscopy. In order to get a first overview of the registered in ovo fluorescence spectra, the mean spectrum and standard deviation were calculated from all spectra of female and male embryos, respectively. The assignment of spectra to the female or male group was done according to the result of feather sexing. Figure 3 shows the fluorescence profiles at an excitation of λ1 = 532 nm.

Calculated mean spectra (μ) and standard deviation (δ) of in ovo fluorescence (I1) recorded at an excitation of λ1 = 532 nm for (A) female embryos (red) and (B) male embryos (blue)

All fluorescence spectra show a maximum intensity around 628 nm. While the spectral profile of female and male embryonic blood is similar, it is apparent that blood of female embryos shows in this spectral range in average a stronger fluorescence intensity than blood of male embryos. Although the different intensities suggest a potential discrimination between female and male embryos taken into consideration the standard deviation, the sex assignment based on this fluorescence signals alone would become unreliable. Figure 4 shows, similar to Fig. 3, the calculated mean spectra and standard deviation of fluorescence spectra at an excitation of λ2 = 785 nm.

Calculated mean spectra (μ) and standard deviation (δ) of in ovo fluorescence (I2) recorded at an excitation of λ2 = 785 nm for (A) female embryos (red) and (B) male embryos (blue)

Here, the blood of male embryos shows in average a stronger fluorescence intensity than the blood of female embryos. In comparison to the fluorescence signals of an excitation wavelength λ1 = 532 nm, the fluorescence signals for an excitation of λ2 = 785 nm are weaker and broader. Individual Raman bands of the inner eggshell membrane are observable. The strong slope between 800 and 820 nm arises from the transmission characteristic of an optical filter blocking the excitation laser line of λ2 = 785 nm.

From the spectra, the question arises which blood constituents are responsible for fluorescence between 560 and 720 nm and between 800 and 1000 nm. One possibility is that constituents of the heme synthesis are at least partly responsible for these signals. Within the complex pathway of heme synthesis, there are two molecules that show a substantial fluorescence. Figure 5 shows a section of the heme synthesis pathway [26]. Briefly, heme synthesis starts in mitochondria and intermediates enter the cytosol, where coproporphyrinogen (1) is generated. After transport back to mitochondria, coproporphyrinogen decarboxylates and oxidizes two propionic side chains to vinyl groups. Protoporphyrinogen is formed and further oxidized to protoporphyrin IX (2). Finally, the enzyme ferrochelatase catalyzes the formation of heme (3) by inserting Fe2+ into protoporphyrin IX.

Illustration final steps of the heme synthesis pathway from coproporphyrinogen (1) to protoporphyrin IX (2) and the terminal step catalyzed by ferrochelatase to the heme (3)

Coproporphyrinogen (1) is a high transparent component showing no fluorescence. Protoporphyrin IX (2) exhibits a comparatively strong fluorescence in the red and NIR while heme (3) shows only a very weak fluorescence in these spectral regions. The fluorescence spectra of synthetic protoporphyrin IX and human hemoglobin are plotted in Fig. 6.

Fluorescence spectra of protoporphyrin IX (dashed) and human hemoglobin (bold) for excitation wavelengths (A) λ1 = 532 nm and (B) λ2 = 785 nm. Note, for better illustration, both fluorescence spectra of hemoglobin are multiplied by a factor of 5

The spectra shown in Fig. 6A and B were registered with the in ovo measuring head (see Fig. 2). Both samples were placed in cuvettes with 10-mm path length. However, the effective path length is determined by the focus of the excitation lasers. The spectral fluorescence characteristic of protoporphyrin IX is similar to the measured in ovo fluorescence (c.f. Figure 3). In contrast to protoporphyrin IX, hemoglobin shows only very weak fluorescence in the region between 560 and 720 nm. However, it is known that concentration of hemoglobin in human erythrocytes is much higher than concentration of protoporphyrin IX. Thus, it can be assumed that in the egg, the concentration of embryonic hemoglobin is also higher than that of the precursor protoporphyrin IX so that especially between 800 and 1000 nm, hemoglobin does predominantly contribute to the fluorescence measured in ovo. Hemoglobin as the main source of the observed NIR fluorescence was also indicated by comparing the time courses of in ovo fluorescence intensity and hemoglobin content during early erythropoiesis [12]. The spectral profile in the range between 800 and 1000 nm also corresponds well with the in ovo fluorescence spectra shown in Fig. 4. However, the fluorescence of protoporphyrin IX (560–720 nm) and hemoglobin (800–1000 nm) alone does not explain the sex-specific differences of the observed in ovo fluorescence intensities I1 and I2 (c.f. Figure 3 and Fig. 4). To understand the different fluorescence intensities, the transmission of protoporphyrin IX and hemoglobin has to be considered. The transmission spectra of both components are plotted in Fig. 7.

Transmission spectra of (A) synthetic protoporphyrin IX and (B) of human hemoglobin. The dotted gray lines indicate the excitations laser lines. Gray bars represent spectral regions used for evaluation in ovo fluorescence intensities

The transmission spectrum of protoporphyrin IX (Fig. 7A) is characterized by the strong Soret band at 408 nm and the four Q-bands between 450 and 650 nm. The Soret band of hemoglobin (Fig. 7B) is also located around 408 nm. The Q-bands are merged to broad absorption band between 500 and 650 nm. The transmission spectra reveal that hemoglobin absorbs stronger the light in the green than protoporphyrin IX. This applies to both the excitation line (λ1 = 532 nm) and the fluorescence light at the maximum of the protoporphyrin IX fluorescence around 628 nm.

Table 2 lists the measured transmission values as well as the calculated extinction coefficients of both compounds for the spectral positions and ranges used for the in ovo fluorescence spectroscopy. In particular, hemoglobin has at λ1 = 532 nm a high extinction coefficient. In contrast to the green spectral range, the absorption of protoporphyrin IX and hemoglobin is only very weak in the range between 940 and 960 nm.

Consequently, the concentration of hemoglobin in the embryonic erythrocytes modulates the fluorescence intensities of protoporphyrin IX. Male chicken embryos have, in contrast to female chicken embryos, obviously at the early incubation days a higher regulated metabolism accompanied by a higher content of hemoglobin. It was recently published that ferrochelatase of the domestic chicken (Gallus gallus f. dom.) is encoded by a single nuclear gene located at the gender specific sex chromosome Z [27]. Birds have no dosage compensation system for the Z chromosome similar to the X inactivation observed in mammals [28]; higher expression levels of Z-linked genes in homozygote male (ZZ) than in hemizygote female (ZW) chicken embryos are the result [29]. Therefore, it can also be postulated that male chicken embryos produce more heme and finally also more hemoglobin than female chicken embryos. This hypothesis is supported by the study of Rahman et al. which have investigated the spectral transmission of eggs at day 3 [30]. The authors found that eggs of male embryos show a higher absorbance of hemoglobin than eggs of female embryos, indicating a faster erythropoiesis of male embryos in the first phase of incubation. Also Tagirov et al. found in a study of sexual dimorphism of the early embryogenesis that male chicken embryos grow faster than females [31]. Thus, it can be assumed that the ratio of hemoglobin to protoporphyrin IX increases with the development of the embryo.

Although not all mechanisms and their interrelationships are yet understood, it can be hypothesized that the observed sex-related differences in the fluorescence intensities are based on the embryonic heme synthesis.

Prediction of sex from spectra registered in ovo can be done using supervised classification. Methods for this have already been successfully applied in previous work [11,12,13]. Due to the complementarity between the signals of the two spectral regions, a more simple classification approach can now be performed. For the in ovo sexing, the fluorescence signals of excitation λ1 = 532 nm and λ2 = 785 nm are ratioed according to

$$F=\frac_^_\left(\lambda \right) d\lambda }(_(620\dots 640 \mathrm))}$$

(1)

The scatter plot of the fluorescence ratio F for female and male embryos is presented in Fig. 8A. Even at first glance, it is clear that female and male embryos can be very easily distinguished. For classification, a discriminator FD, i.e., a threshold, can be introduced. It is obvious that the choice of the threshold determines the accuracy for one or the other sex in opposite direction. Figure 8B shows the accuracy for female and male embryos as a function of the discriminator FD. The accuracy is defined as the proportion of correct determined embryos to the total number of hatched chicks for the female and male groups, respectively. Eggs that were in ovo sexed but from which no chick hatched were not included in the calculation.

(A) Scatterplot of the calculated fluorescence ratio F (see Eq. (1)) for all in ovo sexed eggs. (B) Accuracy versus discriminator FD. Assignment of F to female (red dots) or to male (blue dots) embryos is based on feather sexing

The intersection of the curves is approximately at a discriminator value of FD = 0.49 with a corresponding accuracy of 96% for both sexes. Classification of the in ovo spectra according to Eq. (1) and at a threshold of FD = 0.49 resulted in 795 female embryos and 786 male embryos. For female embryos, there were 759 matches and for male embryos, 752 matches between in ovo and feather sexing. The results show that with the analytical techniques of fluorescence spectroscopy, the sex-relevant molecular information can be registered in ovo. At the same time, fluorescence spectroscopy as a non-contact technique with respect to the living embryo ensures a high hatching rate of normally developed and healthy chicks.

The prediction of sex by in ovo spectroscopy and agreements and discrepancies to the subsequent feather sexing are listed in Table 3.

The reasons for misclassified embryos are not yet understood. As in all previous studies [11,12,13,14], similar deviations were observed too; it can be assumed that variations in the embryonic developmental stage or in the general biochemical profile are reasons for the misclassification. Furthermore, a closer look at the scatter plot also reveals that the in ovo signal F of a few female as well as male chicks is clearly emerged in the other, “wrong” sex group. The reasons for this clear discrepancy are not known. Measurement errors can be excluded after a detailed analysis of measurement procedure and data. As all misclassified chicks have been raised, a possible inherent sexual identity based on genuine male to female chimaeras [32] could not be found in any of the animals so far. It might be that there are discrepancies between the biochemical and genetic sex which are not yet understood.

In addition to the accuracy of the in ovo sexing approach, the hatching rate is also an important parameter of the method. For the evaluation of the hatching rate, a similarly large number of eggs were hatched with complete preservation of their native state. The hatching results of this control group and of in ovo sexed eggs are summarized in Table 4.

For both groups, the hatching rate is nearly identical. In other words, opening and spectroscopic in ovo sexing do not lead a priori to a reduced hatching rate.

It should be noted that the study was performed as a blind test. The sex of each egg was documented prior to hatching and then compared to the result of feather-sexing. The feather sexing as well as the comparison to the predicted sex was performed by independent experienced specialists from the FLI and the Agri-Advanced Technologies GmbH (Visbeck, Germany).

The here described and patented technique of the two-wavelength fluorescence spectroscopy [33] allows an early in ovo sex determination. Due to its applicability at an early stage combined with a high accuracy and especially because of the non-decreased hatching rate, it is well suited for a practical application in hatcheries in order to overcome the killing of day-old male chicks and to solve a current animal welfare problem.

At this point, it is stated that the methodology described here has been filed as a patent and the granting of the property right is in prospect.